Halfway through an intense game of Wavelength, I checked my phone and noticed City were losing with 30 minutes left in the game. Unless City came back to win, Arsenal would win the league title. I instantly turned on the game. The game of Wavelenght continued around me. Nobody else was watching.

City ended up equalizing in stoppage time, but that wasn’t enough. Arsenal became champions for the first time in 22 years. The last time they won the league, I was in preschool, George Bush was president, and people were still using MySpace. The result didn’t feel real, and it took a few minutes to set in. Something I had wished and prayed for so long had finally happened. I stepped away, called my sister, texted my fellow Arsenal fans, and spent an hour thrilled by the result. Then I went back to my afternoon.

A few days earlier, I graduated from my MBA program. On graduation day, I woke up, went to the gym (and listened to ICEMAN), got dressed, went to the graduation ceremony, took pictures with family and friends, went to dinner with my family, and then celebrated with my friends. It was a truly wonderful day. But in many ways, just another day.

I had been dreaming about both of these moments for years. Both arrived rather quietly.

The younger version of me had a different outlook on this.

I had destination addiction, which is the belief that happiness lives at the next milestone. It is the belief that you will only be happy when you finish your degree, get your dream job, or when you “make it”. The problem is that this “when” keeps moving. With each milestone you hit, you celebrate and feel happy for a brief moment, and the clock on the next destination has already started.

My destination addiction was strongest in my first year of undergrad. I had a rule for myself back then: be in the library at 9 am every morning. I would often get there before the librarians and wait for it to open. A few times that year, I cut off my social life to focus on my studies. I told myself I would sacrifice and suffer for a few years and that I would be happy once I had a perfect GPA and my dream job offer. This mindset was effective and maybe even necessary at the time. I did well in school. I got the internships I needed and lined up the consulting job. But it was unsustainable. I was constantly stressed and unhappy. I was living entirely in anticipation of a future goal and destination.

Share

The Courage to Be Disliked, a great book I read just after finishing undergrad, frames this as a mountain climb. If your goal is to reach the top of the mountain, and you tell yourself you’ll be happy once you get to the top, you’re bound to be disappointed: reaching the top is only a brief moment in the overall climb. The peak is always short. The climb is where most of your time is spent.

In Drake’s introspective track “Is There More?”, he reflects on this theme: being unsatisfied and searching for more despite reaching the top of the proverbial mountain.

“Is there more to life than goin’ on trips to Dubai?

Yachts on the 4th of July, G5 soarin’ the skies

Is there more to life than all of these corporate ties

And all of these fortunate times”

The crystal merchant in The Alchemist (one of my favorite books) faces the same problem from the other direction. He has dreamed of travelling to Mecca his whole life, but never goes, afraid that arriving there would leave him with nothing to live for. This is destination addiction inverted: instead of racing toward the milestone, he preserves it by refusing to arrive. Either way, the present disappears. When I first read this in my senior year of high school, I was sure the lesson was about courage: follow your dreams, don’t be the merchant. What I didn’t see was how much I had in common with him.

I understood this lesson conceptually, but my MBA year was the first time I put it into practice.

There’s a philosophical distinction that helped me name what I was doing.

Telic activities have a natural endpoint and are defined by their completion. Writing an essay. Getting a degree. Landing a job offer. On the other hand, atelic activities have no natural endpoint. Learning. Building relationships. Improving at a skill. Atelic activities have no finish line because they are ongoing activities.

The mistake I made in undergrad was treating my entire life as a telic activity. Structuring your life this way produces a permanent sense of not-quite-there. You spend years in the run-up to a milestone, reach it, tick it off, and immediately begin searching for the next one.

The solution is not to stop having goals, but rather to shift your relationship to your goals. Completing a degree is telic. The process of becoming educated is atelic. Landing a job offer is telic. Building a fulfilling career and expertise is atelic. The goals still matter because they provide much-needed direction, but they are no longer the only measure of satisfaction.

By the time I started my MBA, I’d made this mindset shift. The library-at-9am version of me would never have spent a weekend organizing dinner for 50 classmates at an Ethiopian restaurant in Brooklyn. There was no resume line and no grade attached to this, just a long table of people I wanted to spend time with. This year, I made a conscious effort to invest in these experiences and enjoy them. When graduation came around, it was the natural conclusion to a wonderful year.

The approach I took as an MBA student carried over to my football fandom.

Everything I said about beating destination addiction assumes some agency. You can choose how to approach your MBA. You can decide to be present, to invest in the people around you, and to focus on the process rather than an outcome. Supporting a football club gives you none of that. I have not kicked a single ball for Arsenal. I cannot affect the result of games, the fitness of players, or the decisions of the manager. Worse, the structure of being a fan forces destination addiction on you. The title is the only acceptable outcome. 38 games over 10 months build up to answer one question: did your team win?

For most of my life, the answer was no. Arsenal spent nearly two decades being mediocre and the joke of the league. Then, over the last three seasons, the team had a resurgence. In each of those seasons, Arsenal led the league for long stretches. The hope would build early in the season, and I would let myself believe that this was finally our year. I would faithfully tune in every weekend and watch us get closer to winning the long-awaited league title. Then, somewhere around February or March, it would fall apart.

I watched those collapses in horror. The 4-1 loss to City in 2023 (which I took the afternoon off of work to watch). The shocking loss to Villa the season after. Each collapse arrived with a wave of messages from friends who support other clubs. I’ve spent more hours than I’d like to admit arguing with delusional United fans about which team had a better season. I won’t be having that debate this year.

Those years were really painful. Destination addiction in a context where you have no control is a particularly bad deal. You’ve staked your emotional state entirely on an outcome you cannot affect, so every wobble on the way lands as a small catastrophe. A single bad result could ruin a Sunday.

So this year, I tried to watch Arsenal the same way I approached my MBA. As a fan, you cannot change a result. The only thing you can change is what you let a result do to you.

I enjoyed each game on its own. The last-minute equalizer against City early in the season. Both North London derbies. Last week’s crucial match against Burnley. I watched the season as a season rather than a 10-month trial for a single afternoon.

When the title came, it was a wonderful end to a season I had enjoyed, regardless of the outcome. It landed the way graduation had. A good hour, and then the afternoon carried on.

Arsenal is the better teacher. The MBA was an easy test for the lesson. I had agency, and I could shape the year. Football handed me the lesson without any of that. Most of what you actually care about in life is closer to watching a team than running a project. The fact that the lesson held up under those conditions is what makes me trust it.

There’s an obvious objection to all of this.

“Enjoy the journey” is what some people say when they haven’t arrived anywhere. It’s the cope of some who never quite make it, repackaged as wisdom. If you actually wanted something and got it, you’d feel euphoric. Some would argue that a muted arrival is just a lack of ambition in disguise.

That’s not what I’m advocating for. Five years ago, I made a plan: get a job in consulting, get promoted, and get an MBA. Alhamdulillah, five years later, I’ve done almost everything I set out to do. The atelic mindset didn’t make me less ambitious or teach me to want less. It simply made the process more enjoyable and sustainable.

The telic approach burns bright and tends to crash. You see it in people who grind toward one thing really hard, get that thing, and then stall. The atelic approach is a bit slower and more consistent by nature, but it compounds. You keep going because the process of learning and improving never ends.

A week after graduation, I’m already planning out my next five years. Not because I’m unsatisfied, but because the marathon continues.

In Hustle Harder, Hustle Smarter, 50 Cent describes life as an endless tunnel: no destination, no “happily ever after,” just a never-ending series of new challenges around the corner. I read that book around the same time as The Courage to Be Disliked, just after finishing undergrad. Both books made similar points, but it took a few more years for me to understand and implement their lessons.

That Tuesday in Red Hook, we stayed at the waterfront for a few more hours. We talked, played more games, and took some pictures.

Most of us didn’t know when we’d all be in the same place again. The title was everything I’d hoped for, for about an hour.

As the sun set, the celebrations were starting elsewhere in the city. We didn’t go.

I moved to New York for my MBA with the full intention of going back to Canada. However, that day was one of the first days I realized I might want to stay in New York.

Before New York, I was a Montreal person. I grew up in a suburban compound in Saudi Arabia (the kind of place where everyone knew everyone). When I arrived at McGill for undergrad, Montreal felt enormous. It was the first real city I had ever lived in.

Over time though, Montreal became home. I grew to love the summers, the parks, the culture, and the laid-back pace. Montreal is a one-of-a-kind city. I settled into it in a way I hadn’t expected. By the time I left for New York, Montreal felt like the right-sized city for my life. It was big enough to support my career and interests, and small enough to be close to everything I needed and not be overwhelming. I lived within walking distance of my office, my closest friends, the gym, and my barber. What a life. The American mind can’t comprehend this.

New York was a shock and incredibly overstimulating. I was overwhelmed by the noise in Midtown, the dirt and rats in the subways, and the never-ending list of things to do and places to see. In my first few weeks in the city, I told myself this would do for a year, but I was ready to go back to Canada.

Over time, just like I did nearly a decade ago in Montreal, I grew accustomed to and started to enjoy the pace of life in New York. One thing I particularly liked was the diversity of people I met. In Montreal, almost all of my friends were engineers or consultants; I enjoyed it. We had similar backgrounds, common interests, and parallel goals. However, in New York, I was exposed to a much broader group of people: founders, lawyers, finance bros, and artists. I made some new friends and many more acquaintances.

Over 50 years ago, a Stanford sociologist published the foundational paper on acquaintances, which he called weak ties. His finding was that your acquaintances are more valuable for career opportunities and new ideas than your close friends. The reason is simple: your strong ties mostly know what you already know. Your weak ties carry information from worlds you don’t have access to. A 2022 study of 20 million LinkedIn users found the same thing. Weak ties were more effective at transmitting job opportunities than strong ones, even now. The bridges between networks turn out to be more useful than the networks themselves.

New York, more than any other city I’ve spent time in, is built for weak ties. This isn’t because it has the most talent (it doesn’t, in any single field). It’s because the talent is genuinely diverse across sectors in a way that most cities aren’t. San Francisco may have more engineers per square mile, but New York has engineers, entertainment lawyers, quant traders, and gallery owners all working on the same block.

Think of a city as a physics problem.

Every city has potential energy: the raw material of what could happen given the people, institutions, and ideas in it. A small town with three employers has low potential energy. New York, with its density of talent across finance, media, law, healthcare, tech, and everything else, has extraordinarily high potential energy. New York generates more possible collisions per unit of time than almost anywhere else.

Potential energy, though, doesn’t convert to kinetic energy on its own. A battery left on a shelf stores charge, but doesn’t do anything with that energy.

The city puts you in proximity to people from entirely different worlds. Whether you actually reach across to them is a different question. Most people default to their immediate network: MBA friends, industry peers, people who already think like them. That’s not a New York problem; that’s a human problem. New York removes this structural barrier.

Share

The city gives you more to work with, but what you do with it is still on you.

Your weak ties can only think of you when an opportunity arises if they have a clear mental model of your interests and goals. If your weak ties only know you vaguely (smart, works in business, interesting guy), that’s not enough.

There’s a formula I keep coming back to: L = D × T. Luck equals Doing times Telling. Most people cover Doing quite well, but are less intentional about Telling. This ultimately limits their Luck surface area. The key is not only to Do more things, but also to Tell people about what you’re doing.

Public writing is the mechanism for Telling. It’s not just about building an audience, but also about giving the 200 people around you an accurate picture of what you care about and the problems you’re interested in solving. When the right opportunity arises, you want to be the first name people think of.

I didn’t understand this when I started writing publicly. I thought it was about sharpening my thinking, and maybe building a following. This understanding came later, when friends started referencing posts in conversation or mentioning my writing to people I’d never met.

I’ll be honest: I haven’t seen a tangible payoff yet. If this bet doesn’t pay off, it’ll be because the writing wasn’t specific enough, or I stayed too close to my existing circle. It will be an experimental failure rather than an operational failure.

Where you live is a choice about what kind of environment you want to marinate in. It’s both a lifestyle and a career decision. Your weak ties can only surface opportunities for you if they have a clear mental picture of what you’re doing. Your environment can only influence you if you let it. Both require some intentionality, which sounds obvious and turns out to be genuinely hard to sustain.

One thing I keep thinking about is what prompted me to sign up for my first race, and later my first half-marathon. It was the conversation about sovereign capital with someone who managed more money than I’ll probably ever see, and my day at the park watching people who looked like running was their second job fly past me.

What actually got me to the starting line on race day was the people I’d met during the year. Classmates, new friends, and the loose connections the city keeps generating.

New York doesn’t make you better. It makes certain versions of yourself feel more available, and puts people around you who are already living those versions. That turns out to be enough.

I’m staying.

Subscribe now

Swaniker’s key example was his firm Sand Technologies, which leverages African engineers to build innovative solutions for global clients. Swaniker made the point that Africa has the talent needed to do challenging technical work, and that the rest of the world, particularly the US and Europe, were lucrative markets that African companies should build for and try to capture value in. I thought to myself: How can Ethiopia replicate this model?

A few weeks later, I listened to a panel about physical AI at another conference, and the answer started to come together.

The Wrong Bet?

Ethiopia’s answer to Swaniker’s question appears to be Business Process Outsourcing (BPO). At Elevate Africa 2026, Ethiopian Investment Holdings announced a $25.56 million, 3,000-seat BPO facility, designed to anchor Ethiopia as a regional outsourcing hub. The instinct behind this bet is right, deploying patient state capital to develop an export sector. This is the same broad logic that built Ethiopian Airlines, and the logic I argued Ethiopia should double down on in “Ethiopia Has the Playbook.” However, this might be the wrong sector.

The highest-volume parts of the $300B BPO market, especially customer support, are already being replaced by AI. a16z recently framed the BPO sector as a prime target for AI disruption. Decagon, one of the leading AI customer support agents, is now achieving 80% deflection on queries that used to require a human. When deflection rates move that fast, the economics change: fewer human seats are needed to deliver the same service level. As AI becomes more intelligent and affordable, the margins and volume in BPO will likely compress, which means the work, the jobs, and the export revenue all migrate somewhere else.

This process might take five or ten years, but it feels inevitable. Betting on BPO in 2026 risks looking like betting on horse-drawn carriages in 1910.

That raises an important question: if not BPO, then what?

Thesis: Ethiopia should treat physical AI data and infrastructure as the next export sector: not just supplying labor, but owning part of the intelligence layer.

The Physical AI Revolution

The LLMs we know and love today were primarily trained on text scraped from the internet. The internet is currently being scraped faster than humans write. Physical AI is a completely different beast. Physical AI is AI that learns from and acts in the real world, which means the bottleneck isn’t compute. It’s high-quality data from real environments. You can’t easily “scrape” the data needed to train physical AI models. Cars, robots, drones, and other tools can’t be trained on Wikipedia and Reddit. They have to be trained through interaction with the physical world, which requires people to build and maintain it.

For the unfamiliar, physical AI is not limited to robotics (here is a great primer on the world models that are needed for physical AI). It encompasses any system that operates in or learns from the physical environment: sensors, collars, drones, IoT devices, and computer vision deployed in real conditions. There has been an influx of physical AI companies in the last few years, with an impressive amount of funding raised. One example is the New Zealand-based Halter, which builds solar-powered GPS collars for cattle. Underneath the hardware, they’ve built an AI model trained on 7B hours of cattle behavior data collected from over 600K animals. They call it the “Cowgorithm”. Peter Thiel’s Founders Fund just led their latest round, investing $220M at a $2B valuation. The company’s moat isn’t the collars, but rather the data and the model.

Halter is one of the clearest examples of the rise of physical AI, but far from the only one. More than $5B has flowed into a handful of other notable physical AI and world-model companies in recent years, including Figure AI, Skild AI, World Labs, AMI, and Physical Intelligence.

These model companies rely on a significant amount of real-world data, and companies are rushing to generate and provide that data. DoorDash recently rolled out Tasks, which pays its 8M gig workers to film themselves doing everyday tasks to “help AI and robotic systems understand the physical world”. This is the same model Scale AI used to build a $30B company providing training data for LLMs, applied to the physical world.

That physical equivalent of Scale AI hasn’t been built yet, but it will be soon.

Ethiopia’s unfair advantages

Ethiopia has 3 unfair advantages that can help it win in this space.

1. Cattle and agriculture

Ethiopia has the 4th largest cattle population in the world at ~70M (7x as large as New Zealand), spread across more diverse terrain, climate, and breed conditions. The training dataset that could be generated from Ethiopian livestock alone would be much larger and richer than anything Halter has built. That’s before you factor in crops, arable land, and the cost of running operations.

2. Abundant and affordable energy

Ethiopia has some of the cheapest energy in the world. The Grand Ethiopian Renaissance Dam, inaugurated in September 2025, added 5,150 MW of hydroelectric capacity and doubled the country’s total electricity generation in a single project. Physical AI infrastructure runs on cheap power, and Ethiopia’s abundant sustainable energy provides a solution to AI companies and projects facing increasing scrutiny regarding energy consumption.

3. Rapidly growing workforce

Ethiopia also has a rapidly growing working-age population. While more developed countries’ working-age populations are shrinking, Ethiopia adds roughly three million people to its working-age population every year. The median age in the country is 19, and nearly half the population is under 30. The workforce that would deploy this infrastructure is young and keeps getting larger.

None of that is sufficient on its own. Affordable land and labor describe a lot of countries. What separates Ethiopia is the state’s ability to execute long-horizon, capital-intensive state-led projects. Ethiopian Airlines is one of the best carriers in Africa, built and maintained as a state enterprise through decades of instability. The GERD–Africa’s largest hydroelectric dam and one of the 20 largest in the world–was conceived, financed, and completed domestically against significant geopolitical opposition. The new Bishoftu airport outside Addis Ababa will be the largest in Africa upon completion. Ethiopia has an impressive track record on executing large-scale projects, which is exactly what a physical AI bet requires.

Subscribe now

Capturing the Value

Scale AI built a $30B company using Kenyan laborers earning a few dollars an hour. Many of them didn’t know they were working for Scale — they thought they were working for a platform called Remotasks. When Scale pulled out of Kenya in early 2024, those workers were left with nothing: no transferable skills, no equity, and no platform.

The Kenya case illustrates a structural problem that isn’t unique to Scale AI. Winning the activity is not the same as winning the profit. Countries that host production can still lose the economics to whoever owns the intelligence layer above it. The question isn’t whether Ethiopia can attract physical AI activity. It’s whether, a decade from now, it has built an industry rather than just a workforce.

For example, Ethiopian Airlines started by flying planes, and over decades built the operational expertise to become the best carrier on the continent. The ideal arc for Ethiopian physical AI is the same: start by deploying sensors on the world’s fourth-largest cattle herd, accumulate the domain knowledge that outside firms can’t replicate, and move toward owning the data intelligence layer. A curated dataset of 70 million cattle across Ethiopian breeds, terrain, and climate is something no Silicon Valley firm can buy its way into. Sub-Saharan cattle and crops look more like Ethiopia’s than it does New Zealand’s.

The mechanism for capturing that value already exists. EIH (Ethiopia’s sovereign wealth fund, with co-investment mandates across 40 state enterprises) should take equity stakes in physical AI companies that want access to Ethiopian environments, the same way a national mining company takes positions in any extraction projects. Any training dataset generated on Ethiopian land, with Ethiopian labor, should be jointly owned, with Ethiopia holding a perpetual license to use and commercialize it. Foreign companies should embed their engineers in local AI development teams, not just deploy local operators. The goal is technology transfer: building the capacity to supervise, validate, and extend AI outputs instead of just collecting them. For early movers that accept these terms, Ethiopia offers something no regulatory environment in the developed world currently can: clear access, fast approvals, and exclusive rights to specific data environments.

Zipline’s work in Rwanda is another useful example. In 2016, Zipline asked for airspace to deliver blood by drone. Rwanda said yes, and by 2026, that decision had generated a 61% cut in blood delivery times, 350 local jobs, and an in-country R&D facility. What it did not generate was Rwandan IP, platform equity, or a Rwandan company that could eventually compete with Zipline. Ethiopia should replicate Rwanda’s willingness to move fast and push even further on the terms. The ceiling isn’t an R&D facility, but rather an Ethiopian company that owns the agricultural AI platform for sub-Saharan Africa.

Why Foreign Companies Will Come on These Terms

The obvious question is whether any physical AI company would actually agree to this. The answer is yes, because the alternative is worse.

The physical world can’t be moved to where the capital is. You can’t offshore 70M cattle the way you can offshore software development. The developed world may have the models and the money, but it keeps running into the same wall: the environments where physical AI needs to learn are expensive, heavily regulated, and increasingly hostile to large-scale sensor and drone deployment.

In Europe, spraying crops with agricultural drones is banned by default. You need a member-state-specific permit and must prove that no viable alternatives exist. The US FAA only published its proposed rules for autonomous drones flying beyond visual line of sight in late 2025, with final regulations still pending. California and New York have added privacy restrictions on outdoor camera and sensor deployment. High labor costs make the DoorDash-style gig collection model expensive to run at scale in developed countries.

This is a structural bottleneck. Regulation moves more slowly than the technology, and land and labor costs only go one direction in developed markets. For a company trying to train models on genuinely diverse agricultural conditions, the choice isn’t Ethiopia versus a better alternative. It’s Ethiopia versus waiting.

What This Means

For Ethiopia, the window is narrow and closing. The physical AI value chain is forming now. No incumbent has locked up the agricultural data platform, no standard has been set for training data from sub-Saharan environments, and no company owns what Ethiopia could own. The structural advantages are already in place. Manufacturing, once the engine of East Asian growth, has been declining as a share of global employment for decades, even in China. The traditional development ladder is narrowing for new entrants. Physical AI data curation is one of the few explicitly AI-native sectors where human judgment remains essential and can’t be automated away. Being first to a specific environment creates an advantage that compounds over time. Korea treated semiconductor manufacturing in the 1970s as the foundation on which everything else would be built: a deliberate positioning in the part of the chain where the value would eventually concentrate. The logic for Ethiopian physical AI is the same.

For anyone watching from outside Ethiopia, the general lesson holds. The most defensible positions in any emerging industry go to whoever moves before the structure solidifies. That’s true for both countries and companies. Whoever owns the agricultural AI training layer in 2030 will likely still own it in 2040.

Closing

Swaniker’s question at HBS is still the right one. What does “build for the world” mean at a national scale?

For Ethiopia, the answer is bigger than agriculture. Ethiopian Airlines has been flying and maintaining planes for over 80 years. That’s 80 years of operational and maintenance data that aviation AI companies can’t easily get anywhere else. GERD is another story: Africa’s largest dam, operational since last year, generating physical environment data at a scale that’s difficult, if not impossible, to replicate in a lab. Ethiopia has been building great projects: a world-class airline, a powerful dam, and a massive airport in Bishoftu. Physical AI presents an opportunity for the country to better leverage these assets.

This is what “build for the world” actually looks like.

Share

Leave a comment

Further Reading:

[1] EIH BPO facility announcement — StockMarket.et, “Ethiopian Investment Holdings Unveils $25.56 Million Flagship BPO Project at Elevate Africa 2026,” 2026. https://www.stockmarket.et/ethiopian-investment-holdings-unveils-25-56-million-flagship-bpo-project-at-elevate-africa-2026/
[2] BPO market size ($300B) and AI disruption thesis — Kimberly Tan, Andreessen Horowitz, “Unbundling the BPO: How AI Will Disrupt Outsourced Work,” February 2025. https://a16z.com/unbundling-the-bpo-how-ai-will-disrupt-outsourced-work/
[3] Decagon deflection rates — Decagon.ai, product documentation and case studies.

https://decagon.ai/

[4] Halter Series E ($220M, $2B valuation, Founders Fund) — BusinessWire, “Halter Raises $220M in Series E to Accelerate Global Expansion of Virtual Fencing,” March 24, 2026. https://www.businesswire.com/news/home/20260324036414/en/
[5] Halter Cowgorithm and 7 billion hours of training data — Yahoo Finance / Inc.com, “Peter Thiel Bets $220M On AI Cow Collars That Control Herds With An App — The ‘Cowgorithm’ Is Trained On 7 Billion Hours Of Cattle Behavior,” 2026. https://finance.yahoo.com/sectors/technology/articles/peter-thiel-bets-220m-ai-140103545.html
[6] DoorDash Tasks app launch — TechCrunch, “DoorDash launches a new ‘Tasks’ app that pays couriers to submit videos to train AI,” March 19, 2026. https://techcrunch.com/2026/03/19/doordash-launches-a-new-tasks-app-that-pays-couriers-to-submit-videos-to-train-ai/

[7] EU agricultural drone spraying ban — European Parliament, Parliamentary Question E-000920/2024 on drone spraying legislation, 2024. https://www.europarl.europa.eu/doceo/document/E-9-2024-000920_EN.html; Eurofins, “Drone spraying regulation and opportunities in Europe.” https://www.eurofins.com/agroscience-services/about-us/latest-news/drone-spraying-regulation-and-opportunities-in-europe/
[8] FAA BVLOS proposed rules — FAA, “Beyond Visual Line of Sight (BVLOS),” and Federal Register, “Normalizing Unmanned Aircraft Systems Beyond Visual Line of Sight Operations,” January 28, 2026. https://www.faa.gov/newsroom/beyond-visual-line-sight-bvlos; https://www.federalregister.gov/documents/2026/01/28/2026-01644/

[9] Ethiopia cattle herd size (Africa’s largest, ~68-70 million) — Statista, “Africa: countries with largest cattle population,” citing FAO data, 2022. https://www.statista.com/statistics/1290046/cattle-population-in-africa-by-country/
[10] GERD inauguration, 5,150 MW capacity, doubling Ethiopia’s electricity generation — Al Jazeera, “Ethiopia inaugurates GERD dam amid downstream tensions with Egypt, Sudan,” September 9, 2025. https://www.aljazeera.com/news/2025/9/9/ethiopia-inaugurates-gerd-dam-amid-downstream-tensions-with-egypt-sudan
[11] Ethiopia electricity prices among Africa’s lowest — GlobalPetrolPrices.com, “Ethiopia electricity prices,” 2025. https://www.globalpetrolprices.com/Ethiopia/electricity_prices/
[12] Ethiopia median age (~19) and youth demographics — Worldometer, “Ethiopia Demographics 2024.” https://www.worldometers.info/demographics/ethiopia-demographics/

[13] Zipline Rwanda launch (2016), 61% reduction in delivery times, 350 local jobs, R&D facility — Lancet Global Health, “Effect of unmanned aerial vehicle (drone) delivery on blood product delivery time and wastage in Rwanda,” 2022. https://www.thelancet.com/journals/langlo/article/PIIS2214-109X(22)00048-1/fulltext; Gavi, “Rwanda launches world’s first national drone delivery service powered by Zipline.” https://www.gavi.org/news/media-room/rwanda-launches-worlds-first-national-drone-delivery-service-powered-zipline

[14] Scale AI / Remotasks Kenya pullout (2024) — Rest of World, “Scale AI’s Remotasks is booting workers with no explanation,” March 2024. https://restofworld.org/2024/scale-ai-remotasks-banned-workers/

I’m much busier these days, and most weeks, I spend fifteen minutes before the deadline looking through my team before making a gut decision. My FPL results have suffered accordingly. I studied engineering in undergrad. I like correct answers, clean models, and systems that work. And yet here I am, managing something I care deeply about on nothing more than vibes.

I’m not the only one with this problem. Around 12 million people play FPL every season, and many of them are doing exactly what I do: checking their team for a few minutes before the deadline, making an instinctive decision, and hoping for the best.

This may sound trivial, but FPL is incredibly important to me. A good gameweek genuinely improves my mood for days. A bad gameweek can easily ruin it.

One of my new year’s resolutions this year was to win all of my FPL leagues. With only a few weeks left in the season, I’m not at all on track, so I decided to build a tool that would help me stop relying on intuition alone, use data to get back to the top 0.1%, and win all of my leagues.

what I built

I’ve wanted to build this tool for a long time, but hadn’t for a simple reason: I couldn’t code (or at least not well enough to build what I had in mind). That changed a few months ago when AI coding tools made it possible to go from idea to working prototype in a few hours. I built the first version during a week-long trip to Panama, and refined it over the next few weekends.

The idea for the tool started with a single question: what data do I actually need to make the optimal FPL decision each week? I was spending my fifteen minutes before each gameweek deadline jumping between different websites, trying to piece together injury news, fixture difficulty, recent form, projected points, and advice from FPL influencers on X (yes, this is a real thing). My initial instinct was to centralize and aggregate all the data I used regularly into a single dashboard.

Each player’s form, fixtures, and expected points over the next few gameweeks. How each player was under- or overperforming their expected points over the last few gameweeks. How my team stacked up against my mini league rivals, who we all had in common, and where the differentiators were.

The logic was pure Moneyball: better data would lead to better decisions and more points. I was convinced this would work.

what broke

After some initial refining, I was ready to use the tool to make my FPL decisions a few weeks ago. I loaded up the site excitedly, but something felt off almost immediately. I had all this data in front of me and some very cool-looking dashboards, but I wasn’t sure what to do with all of the data.

One of the key problems was expected points, the stat the tool was built around. xP is a stat commonly used in FPL to project the number of points each player will score in the upcoming gameweek. Although xP does correlate with actual points scored, the Pearson correlation across active player gameweeks over recent seasons is only about 0.60. Meaningful, but far from decisive. Football is a fundamentally random sport decided by moments. A player can have an expected output of 6 points and score 2 or 15. Over a single gameweek, the noise often overwhelms the signal.

I tried to add more data to the dashboards in hopes that it would give me clarity on my weekly decisions. One such metric was an over/under metric - how much a player was over- or underperforming their xP over the last 3 or 5 gameweeks. I figured that if a player was consistently underperforming their expected output, this might indicate they were likely to regress to the mean in the coming weeks. Although that’s true sometimes, it’s still not reliable enough to base transfers on. Some players consistently overperform their xP, and other players consistently underperform their xP. More data was useful, but it didn’t give me the decision-making certainty I was hoping for.

This doesn’t mean the data is worthless. Expected points over the next few gameweeks, form trends, and mini league rival comparisons are all valuable data to have in one place. The problem was my assumption that better data would automatically lead to better results. I’d built myself a more informed version of the same uncertainty I started with. The dashboard was useful, but it wasn’t sufficient.

the reframe

One of the classes I’m taking this year introduced me to the jobs-to-be-done framework. The core idea, from Anthony Ulwick, is that people “hire” products to get a specific job done. Nobody buys a drill because they want a drill. They buy a drill to make a hole in the wall.

I started asking myself: what job am I actually hiring this tool to do? My initial answer was obvious: maximizing the points my team scores. However, after building the tool and hitting the limits of the data, that framing stopped making sense. I realized that no tool can reliably maximize your points when the sport itself is this unpredictable. The outcome is outside your control. You can make the “right” call on every transfer and still have a terrible gameweek. The randomness isn’t a solvable problem; it’s a permanent feature.

While a tool can’t control the outcome, it can control the process. Particularly, the speed of the process. It can help you make a more informed decision faster and feel confident in it. The job for many FPL managers isn’t “make my team score more points”, but rather “help me feel informed and decisive in five minutes on a Friday afternoon.”

That reframe clarified who the tool is actually for. There are FPL managers who want to spend hours deep in the data. The dashboard already works for them. Then there are people like me on most weeks: busy, competitive, wanting a good answer before the deadline. That’s the larger group, and they don’t need more data. They need a decision.

So I built the 5-minute brief. You open it, it pulls in your team and your mini league position, and gives you a transfer recommendation with a short explanation. Something like: “Transfer out Player X (underperforming, tough fixtures next 3 weeks) for Player Y (strong form, favorable run). This gets you closer to Rival Z, who has 15 points on you in your work league.” No dashboard. No charts. Just a decision and a reason, in the time it takes to drink a coffee.

what i learned

I could not have arrived at the 5-minute brief without building the dashboard first. If I had started by interviewing 50 FPL managers and asking what they wanted, most of them would have said exactly what I initially believed: “I want a tool that helps me score more points.” They would have described the outcome they want rather than the job they need done. I would have built a fancier dashboard, loaded it with more data, and ended up with a tool that didn’t serve the job most FPL managers actually need done.

The insight for the 5-minute brief only came from building the dashboard, seeing that it didn’t fully solve my problem, and then being forced to take a step back and think about what I needed to fix this. Building the wrong version of the tool replaced the research.

The lesson goes beyond FPL. Everyone talks about how AI makes it easier to build things, and that’s true. I couldn’t code, and I built a working tool in a couple of weekends. However, the more interesting consequence is what cheap building does to the learning process. When building a first version costs months of work and thousands of dollars, it makes sense to spend weeks carefully planning and researching before you start. When it costs a few hours or a weekend, the math flips. You learn more by making something and discovering its limits than by trying to anticipate said limits in advance.

This is how judgment develops in practice. Not by reading extensively or studying frameworks, but by building something, discovering what’s wrong with it, and then fixing it. Building the wrong version trains your instincts for what the right version should be.

closing

This Friday, when the deadline closes in a few hours, I’ll open the app. I won’t jump between six websites trying to piece together what to do. The 5-minute brief will tell me which transfer to make and why, based on my team and my league. I’ll read through the brief, and if the recommendations make sense, calmly make the decision and go back to enjoying my day.

I still don’t have a tool that guarantees more points. Football is too random for that, and I’ve made my peace with it. What I have is a faster way to make a decision I feel good about.

If you’ve been sitting on a side project idea, don’t wait until you have the time to plan it properly or the skills to build it right. Go ahead and build it. The first version will probably suck and won’t be what you end up with, but that’s the point.

Share

Learn more about the tool and sign up for the beta waitlist here! Or watch a demo of me using FPL Edge here!

Most of the career advice we get is focused on the next 2-3 years. Which company, which role, which city? That’s useful, but I’ve been more interested in a longer-term version of the question. If I had to make a 50-year bet on what to build my career around, what would I bet on?

After spending the better part of this year observing how quickly the world is changing, I’ve landed on two structural shifts that I believe will define the next 50 years, and a framework for how to position yourself for both.

Shift 1: AI is redesigning knowledge work

AI will fundamentally redesign knowledge work, and nobody knows what it looks like on the other side.

Two years ago, the best AI models could independently handle software engineering tasks that take a human about 30 minutes. By late 2025, that number rose to around 5 hours. As of this month, it is roughly 14 hours. The curve is exponential.

With that said, adoption (outside of software engineering) is lagging far behind the technology. 84% of the world has never used an AI tool. Only 0.3% actually pay for one.

History suggests that productivity gains from major new technologies grow slowly before taking off once adoption reaches critical mass, and we are clearly still in the slow phase. Around 80% of companies are exploring or piloting AI tools, but only 5% of integrated AI pilots have extracted positive returns. The primary issues are not the models themselves, but poor integration with existing business processes, lack of data readiness, and focus on the wrong use cases.

The historian Paul David documented this pattern with the adoption of electricity. Factories had electric motors for decades before productivity moved, because the real gains required redesigning the entire factory floor around what electricity made possible, not just swapping out the steam engine. I wrote about this recently in the context of consulting: firms are investing heavily in AI, but most are bolting the technology onto existing workflows rather than rethinking how work gets done. The firms seeing real results are the ones redesigning their delivery models from scratch. The same pattern will play out across every knowledge-work industry. The question is not whether AI will change how we work. It is how quickly, and who will be ready when it does.

Shift 2: The world is going multipolar

The second shift is geopolitical. The US-led world order that has defined the last 50 years is giving way to something more distributed.

In January, Canadian Prime Minister Mark Carney took the stage at Davos and delivered what became one of the most talked-about speeches in recent memory. His core message was blunt: “We are in the midst of a rupture, not a transition.” The rules-based international order that the West built after World War II, and that the US has underwritten for decades, is breaking down. Great powers are weaponizing economic integration, and the institutions designed to prevent that are failing to keep up. He received a rare standing ovation.

What made the speech resonate was the prescription, not just the diagnosis. Carney argued that “middle powers” can no longer rely on the US to set the rules. They need to build their own coalitions. He recently proposed connecting the Trans-Pacific Partnership with the European Union to create a trading bloc of 1.5 billion people, which would be the largest in the world by far. His phrase captured the urgency: “If we’re not at the table, we’re on the menu.”

Ray Dalio’s framework for the rise and fall of empires helps explain why this is happening now. According to Dalio, the decline of a dominant power generally coincides with the rise of a challenger. The indicators of decline are recurring: rising inequality, increasing internal conflict, and growing external aggression. The US is showing all three. This doesn’t mean the US is collapsing, but it does indicate that the era of unchallenged American hegemony is over.

A corresponding rise is happening in China. A decade ago, Chinese technology companies were largely seen as knockoffs. Today, BYD has overtaken Tesla as the world’s largest electric vehicle maker, selling over 2.2 million all-electric vehicles in 2025, while Tesla’s deliveries declined. BYD is now entering the North American market, with plans to rapidly expand in Canada this year. In AI, Chinese labs like DeepSeek, Alibaba’s Qwen, and Kimi are producing models that match frontier Western performance at a fraction of the cost. Qwen has overtaken Meta’s Llama as the most downloaded open-source model family on Hugging Face. China is no longer just catching up; in several domains, it is competing at or ahead of the frontier.

The structural advantages indicate this trend will continue. China generates more than twice as much electricity as the US and has been growing power generation at roughly 6% per year, with over half from clean sources. As AI moves from chatbots to physical AI over the next few decades (e.g., robotics, autonomous vehicles, and industrial automation), China’s manufacturing base and energy infrastructure give it an edge that compounds. They are also exporting this playbook, building AI and energy infrastructure across the Middle East and Southeast Asia simultaneously.

Beyond China, demographic and economic momentum is shifting globally. Africa’s population is projected to double by 2050, reaching 2.5 billion people with a median age of around 20. India has already overtaken China as the world’s most populous country, and Southeast Asia continues to grow. Meanwhile, Europe’s population is shrinking, and the US, Canada, Japan, and South Korea all face aging workforces and declining birth rates. Economic gravity tends to follow people, and the people are increasingly not in the West.

Capital and talent are already moving accordingly. In 2024, the UAE attracted 173K highly skilled workers (a 21% increase), and overtook Canada and the UK as the second most popular destination for STEM talent globally. Saudi Arabia is experiencing similar growth. South-South trade has surged from $0.5 trillion in 1995 to $6.8 trillion in 2025. New centers of economic gravity are forming around the globe.

For someone planning their career over the next few decades, the implication is straightforward. Previously, focusing your career on the US was a safe bet. The growth and opportunities were there, and you didn’t need to think much beyond its borders. That is becoming less true. As power is distributed across multiple regions, companies and capital will need people who can move between them.

Subscribe now

The framework: become a global, AI-native generalist

So if these are the two defining shifts of the next 50 years, the natural question is: how do you position yourself for both? My answer, and the framework I’m building my own career around, is to become a global AI-native generalist.

AI-native: the foundation

What does being AI-native actually entail? It is not using ChatGPT to rewrite an email. It means treating AI fluency like a language you practice daily. Building workflows, experimenting with new tools as they come out, and developing judgment about what AI is good at and where it falls short. It means being the person who, when a new problem comes up, instinctively reaches for an AI tool before defaulting to the manual approach.

It matters because the technology is improving so fast that the gap between those who keep up and those who don’t widens every month. A recent study scored 50 professionals across top strategy firms, Big 4, and VC/PE on their AI capabilities. These are driven, high-performing people who use AI daily. The median score was 35 out of 100, with only 7 out of 50 scoring above 50. Two-thirds scored below 20 on automation and AI agents, the tier where productivity actually compounds. Nearly everyone was stuck at basic chat.

The few who pushed past that level saw dramatic results. A growth equity partner who invested time in learning AI tools automated his entire deal pipeline, achieving 5x his previous throughput. An incoming consultant with zero technical background spent a few weeks practicing and cut 80% of the time on a client deliverable, outperforming senior colleagues with years of experience who simply hadn’t learned the tools. Meanwhile, a consultant at a different firm that promised clients AI-powered timelines without investing in the skills ended up absorbing the gap manually and went on medical leave due to burnout. Same industry and same tools available, but completely different outcomes.

The difference was not talent or access. It was whether they invested the time to stay at the forefront of a technology that is evolving week to week. Most people are waiting for someone else to figure it out first. That wait is getting more expensive by the month.

The generalist advantage

If being AI-native is the foundation, being a generalist is where the real advantage comes from. This goes against conventional career wisdom, which preaches early specialization and becoming an expert.

That advice made sense in a world where learning a new skill took years, and the cost of being mediocre at something was high. If it took you five years to become a competent financial analyst, you couldn’t afford to also spend five years learning product management. You picked a lane and went deep because breadth was expensive.

AI changes that math. It collapses the skill gap in any given domain enough that a curious person can become genuinely useful in it quickly. I’m not an engineer, but I built a working software tool in a matter of weeks. Not because I suddenly learned to code, but because AI handled the parts I didn’t know, and I brought the judgment about what needed to be built and why. That same pattern applies across domains. A generalist with AI fluency can now create a financial model, build a prototype, write a marketing strategy, and analyze a dataset. The economics of being a generalist are changing because the cost of being useful in a new domain is dropping toward zero.

David Epstein made this case in Range before the rise of genAI. He found that in complex, unpredictable environments, generalists consistently outperform specialists. Expert predictions actually decline in accuracy as experience increases, because deep specialists tend to force new problems into familiar frameworks. Interdisciplinary teams outperform specialist teams on complex problems because they draw on more diverse mental models. The generalist advantage has always existed in “wicked” environments where the rules are unclear, and the problems don’t repeat cleanly.

What’s new is that AI supercharges this advantage. Previously, being a curious generalist meant you knew a little about a lot but couldn’t do much with any of it. You were an interested amateur. Now, curiosity plus AI fluency equals someone who can actually build, analyze, and ship across multiple domains. AI is not just one of three things to develop alongside being a generalist. It is the mechanism that makes being a generalist a viable career strategy.

The counterargument is fair. Specialists benefit from AI too. The doctor who used Claude Code to build a patient care tool did so precisely because of his deep domain expertise. AI made his specialist knowledge more powerful, not less. This is not a zero-sum argument. Specialists will continue to thrive, especially in fields where deep expertise is irreplaceable. The argument is not that specialists are doomed. It’s that generalists have been systematically undervalued, and AI is tipping the scales in their favor for the first time.

There’s a harder question behind the specialist-vs-generalist debate: if AI makes being a generalist accessible to more people, what separates the best generalists from everyone else?

The answer is taste. Paul Graham wrote about this years ago: “The recipe for great work is very exacting taste, plus the ability to gratify it.” AI dramatically expands your ability to gratify and to execute across domains quickly. Taste is the part that remains entirely human. It’s knowing which problem is worth solving, which solution is elegant versus just functional, and what needs to be changed to improve the solution.

When ideas are cheap, and anyone can generate a passable version of almost anything, the bottleneck shifts from production to curation. The person who can direct the fifth iteration, who knows the output isn’t there yet and can clearly explain why, becomes more valuable than the person who produced the first draft.

The good news is that taste is learnable. Graham argues it develops through cross-domain exposure, relentless iteration, and the habit of studying the best work across fields. That description fits how generalists already operate. Breadth is not only how you build skills, it’s also how you build judgment. Every new domain you explore adds to the reference library your taste draws from.

The global layer

I grew up in Saudi Arabia, where many of my friends’ parents worked at multinational corporations like Aramco. Many of them were the people who sat between cultures, translating context, not just language, helping bridge teams, companies, and markets that would struggle to understand each other directly. I didn’t have a name for it at the time, but I was watching these bridge people my entire childhood.

I saw the same pattern again earlier this year during my MBA trek to Japan and Korea. The people we met had built careers as intermediaries between Asian companies and Western ones. They navigate business norms, regulatory environments, and cultural expectations on both sides. They help deals happen that wouldn’t happen without someone who understands both worlds.

In a unipolar world dominated by the US, you didn’t need many of those people. Most business flowed through American institutions, American capital markets, and American cultural norms. As the world becomes multipolar, the demand for bridge people multiplies. Instead of every company needing one person who understands the US market, they increasingly need people who can navigate across multiple regions. Someone who understands how business works in the Gulf, in East Africa, and in East Asia becomes more valuable as each of those regions grows in economic importance.

So what do you actually do?

The most common response I get when I talk about this framework is “okay, but where do I start?” The answer is that the three parts are not really separate steps. Being AI-native is what makes being a generalist achievable. Following your curiosity into a new domain and using AI to become useful in it is how you become a generalist. Doing that across cultures and geographies is what makes you global.

My own trajectory has been a mix of intention and luck. I read Range in 2021, and it influenced my decision to start my career in consulting, a role that is fundamentally generalist. You work across industries, problem types, and functions, and you develop the ability to get up to speed on something new quickly. I feel vindicated about that decision now. I chose a tech-focused MBA program specifically to spend the year immersing myself in AI and becoming AI-native, which is already paying off in ways I didn’t expect. The cross-cultural piece was less deliberate. Growing up in Saudi Arabia, visiting family in Ethiopia every summer, moving to Canada for university, and spending time in East Asia and the US gave me a set of experiences I didn’t fully appreciate until I started thinking about what the next few decades might look like.

If I had to give someone a starting point, it would be three things.

1. Study the latest AI tools: Take courses like those freely available at Anthropic Academy.

2. Build something you’ve always been interested in with AI: Use AI to bring your ideas to life or explore new interests. I’ve been building a tool that uses data science to optimize my Fantasy sports team, a system to manage my notes and help me write more consistently, and a Career EV Analyzer to model different career paths financially. These have all sharpened my judgment about what AI is good at and where I still need to bring my own thinking.

3. Resist the pressure to specialize too early: Explore broadly, develop range, and trust that the range pays off. The pattern recognition you build across domains is what becomes taste.

4. Expand your worldview: Spend time in a different country and build relationships across borders. Learn a new language. The world is getting more distributed, and the people who can move between its different parts will have an enormous advantage.

Closing

I started this essay with a question: if I had to make a few 50-year bets on what to build my career around, what would I bet on? I don’t know exactly what the world looks like in 50 years. Nobody does. I do know the direction: AI is redesigning how we work, power is distributing across the globe, and the pace of change in both is accelerating.

The people who thrive in that environment will be the ones who are curious about many things, fluent in the tools that turn curiosity into capability, and comfortable operating across borders. That is the bet I’m making.

Subscribe now

Further reading:

1. . MIT, “The GenAI Divide: State of AI in Business 2025”

2. Paul David, “The Dynamo and the Computer: An Historical Perspective on the Modern Productivity Paradox,” American Economic Review

3. Mark Carney, Davos 2026 Speech

4. Ray Dalio, Principles for Dealing with the Changing World Order

5. Brookings Institution, “How Will the United States and China Power the AI Race?”

6. UNCTAD, “10 Trends Shaping Global Trade in 2026”

7. “The Widening AI Gap”

8. David Epstein, Range: Why Generalists Triumph in a Specialized World

9. Paul Graham, “Taste for Makers”

December 2017, my first winter in Montreal. I was sitting in the library with my linear algebra notes spread across the table, frantically trying to learn what eigenvalues and diagonalization are.

I had failed the midterm a few weeks prior, and the final exam was now worth 90% of my overall grade. The pressure was on.

48 hours into my cram session, the spiraling started.

I should have listened to Baba’s advice about taking uni seriously. I should have attended the lectures and started studying earlier. I’m going to fail this exam and this class. It’s going to ruin my GPA. I’m going to be a disappointment and a failure. I’m not as smart as I thought. I should have just been a business major.

In a new city and country, I didn’t have close family or friends nearby. The support system I had built over the first 18 years of my life was not around. It was one of the first times I had felt this level of anxiety, and I had to face it alone.

The only respite I had from my mid-cram nervous breakdown that day was the prayer room on campus.

That experience was almost a decade ago, but it has shaped how I think about faith and anxiety.

You’ve probably seen the data by now. Anxiety has been rising globally, particularly among Gen Z.

I don’t think that’s surprising to anyone who’s been paying attention. Social media is frying our brains. AI is replacing entire teams in real time. World War 3 might have already started. The career playbook most of us were handed (get a degree, land a stable job, buy a house) feels less reliable every year. The world is increasingly chaotic, and people can feel it.

Historically, faith was one of the primary ways people dealt with this kind of uncertainty. Prayer, ritual, community, and a relationship with something bigger than themselves helped people make sense of a world they couldn’t control.

However, over the past few decades, particularly in the West, we’ve started to move away from faith and replaced it with secular wellness. The secular wellness toolkit, which includes therapy, journaling, mindfulness meditation, and exercise, has merit; I’m not dismissing it. I do most of these things and find them very helpful (except for going to therapy). With that said, most of the secular wellness toolkit focuses on managing symptoms. What faith offers that these tools generally don’t is a framework for interpreting uncertainty rather than just enduring it. A way of orienting yourself when everything around you is shifting.

We have let go of this critical tool at a time when it is needed more than ever. There are a few ways this critical tool has helped me in my life.

A different relationship with uncertainty

Anxiety is a series of tricks your mind plays on you. One of the most powerful tricks anxiety plays on you is convincing you that control is possible if you just worry hard enough. This illusion of control drives the overthinking and rumination many of us are prone to. You run the scenario on repeat in your head, each pass slightly more catastrophic, and none of them produce the exit or solution you’re looking for.

The concept that changed how I think about this was tawakkul, roughly translated as reliance on God. This is the practice of giving your full effort to what you can control, and acknowledging and releasing what is out of your control.

Tawakkul is not passive fatalism, but rather active trust in a higher power and divine decree. The Islamic framework asks you to work as if outcomes depend on you while trusting as if they don’t.

This concept is not new. One of the core pillars of Stoic philosophy was to focus on what you can control and let go of what you cannot. Modern psychologists have popularized the Zone of Control practice: dividing your concerns into what you can act on and what you can’t, focusing energy on the first list, and consciously releasing the second. These are different vocabularies for the same concept. The faith-based version of this adds an important element: someone to hand the second list to.

“The chief task in life is simply this: to identify and separate matters so that I can say clearly to myself which are externals not under my control, and which have to do with the choices I actually control.” — Epictetus

Ritual provides structure when your mind has none

There are moments when I can feel my mind starting to spiral, and I know from experience that no amount of reasoning will stop it. My chest tightens, my breathing gets shallow, and my thoughts start looping. The problem isn’t the first anxious thought. It’s that there’s no off switch.

Structure can help interrupt these anxious loops. Psychologists call them anchoring statements: short, memorized phrases that you repeat under stress to reactivate the rational part of your brain.

If you’ve read or watched Dune (a universe heavily inspired by Islam), you’re already familiar with the concept. The Litany Against Fear is essentially dhikr: a memorized phrase, repeated under stress, designed to pull your mind back to center. In Islam, those phrases are specifically intended to remind us of God.

Salah, the five daily prayers, takes this further. Salah is not a single anchoring statement, but a structured sequence of them, repeated five times a day, with physical grounding built in. You stand, you bow, you prostrate. Your body is doing something structured and repetitive while your mouth is repeating phrases designed to pull your attention out of whatever loop you’re in.

Gratitude redirects attention

Another trick anxiety plays on your mind is hijacking your attention. It points everything at the perceived threat, making it impossible to see anything else. What’s going well, what’s already been given, and what’s still possible all disappear. Psychologists call these patterns cognitive distortions. Your mind doesn’t just focus on the perceived threat. It exaggerates it, catastrophizes it, and filters out all that contradicts it.

This matters because what you pay attention to shapes your lived reality. If your mind only shows you evidence that things are falling apart, that becomes your experience, whether or not it’s accurate. It’s no coincidence that the most important breakthrough in modern AI was a paper called “Attention Is All You Need.” Researchers found that what a system focuses on determines the quality of everything it produces. The same is true for the human mind. Cognitive distortions are so powerful because they corrupt the attention mechanism itself, selectively feeding you information that confirms the worst-case scenario while hiding everything else.

Gratitude counteracts anxiety by refocusing your attention. The brain cannot sustain both states simultaneously because they recruit competing neural networks. Psychologist Barbara Fredrickson’s research on positive emotions found that gratitude literally broadens your cognitive field. Anxiety narrows your thinking to the threat, and gratitude reopens it. Her studies showed that people experiencing positive emotions like gratitude recovered faster from the physiological effects of anxiety (elevated heart rate, blood pressure) than those who didn’t. Neal Krause’s research takes this a step further. It shows that people who practice gratitude specifically toward God report higher levels of stress resilience, with adults who felt more grateful to God scoring higher on stress-tolerance measures than those who didn’t.

In Islam, this practice is called shukr. Counting your blessings is a trained discipline of appreciating the good, regardless of what else is happening. The Quran says, “If you are grateful, I will surely give you more” (14:7). A. Helwa synthesizes the concept simply in Secrets of Divine Love: since what you focus on becomes bigger, when you focus on your blessings, you find more for which to be grateful. Gratitude doesn’t change your circumstances. It changes what your mind lets you see.

Gratitude is built into the structure of Islamic worship. Every prayer begins with “Alhamdulillah,” all praise belongs to God. Gratitude comes before any requests. Before you ask for help and before you bring your anxieties, you start by redirecting your attention toward what has already been given. Five times a day, the first move is gratitude. This consistent daily practice of gratitude is an effective antidote for the anxiety we may experience throughout the day.

Anxiety isn’t going anywhere. The world is going to keep changing rapidly (faster than ever before), and most of us are going to keep feeling the weight of that. The question isn’t how to make anxiety disappear. It’s what we reach for when anxiety kicks in.

For me, it’s been faith. I think we were too quick to let it go.

Further reading:
1. https://www.anxiousgeneration.com/research/the-evidence

2. https://news.gallup.com/poll/697676/drop-religiosity-among-largest-world.aspx

3.https://www.heritage.org/conservatism/report/the-great-falling-away-the-decline-religious-services-attendance-the-united

4. https://pmc.ncbi.nlm.nih.gov/articles/PMC1693418/

5.https://www.researchgate.net/publication/230728587_Feeling_grateful_to_God_A_qualitative_inquiry

6. https://www.goodreads.com/en/book/show/52207852-secrets-of-divine-love

7. https://www.goodreads.com/book/show/53138016-nervous-energy

The Prophet’s Masjid, Medina. Shot by nadiem


A few months later, GPT-4.5 dropped. My manager at the time tasked me with building Power BI dashboards for a client, which required writing code in Power Query (a data processing language I had never touched). He figured it would take at least a few days, and maybe even a week, for me to get up to speed and deliver something.

I had my first version of the dashboards ready that afternoon.

He was both bewildered and excited when he asked me, “How did you finish this so quickly?” I told him I’d had some help. I realized two things that day:

1. AI is going to be incredibly useful, and 2. AI may one day kill consulting and take my job.

Later that year, I worked on a project to build and scale an enterprise-wide GenAI tool across the firm. I watched the tools become more capable month over month, and usage grow across teams. During that project, I spent a lot of time considering how GenAI tools could increase employee productivity and the broader implications for consultants like myself.

Fast forward to today. I’m wrapping up my MBA and preparing to go back to consulting in September. The AI tools have become significantly smarter and more autonomous, and they’re now integrating directly into Excel and PowerPoint (the two lifelines of all consultants). The question I’ve been pondering since that day with the Power BI dashboard has only gotten louder: is AI going to take my job, or could it possibly enhance it?

Here’s my take now: AI won’t kill consulting. It will expand demand for it. With that said, it will also carve out the bottom of the consulting pyramid and change what it means to be a consultant.

How Consulting Firms Actually Work

Consulting firms are built around what Professor David Maister called the “leverage model.” The consulting pyramid (partner, principal, engagement manager, associate, consultant) exists because each layer below the partner generates surplus revenue. A consultant might cost the firm roughly six figures a year, but clients will pay two to three times that for the consultant’s time. Scale that across a five-person team, and the math gets compelling.

Clients are willing to pay premiums for the partners’ experience: pattern recognition built from decades of engagements, relationships with decision-makers, and the credibility to walk into a boardroom and be heard. Consultants, on the other hand, sell analytical capacity and executional horsepower.

A helpful analogy for consulting firms is a plane. The partners are the pilots; they set the team’s direction and change course when needed. The rest of the team is the engine. They provide the energy the plane needs to travel to its destination. The plane goes where the pilot points it, but without the engine, the plane doesn’t go anywhere.

The distinction matters because when people ask, “Will AI replace consultants?” they often assume this includes the pilot and the steering. I don’t see AI replacing that. The engine, however, is a different story.

AI Shrinks the Bottom of the Pyramid

Say I’m a manager or associate staffing a project, and I can choose between having three entry-level consultants on my team or having two consultants and relatively unlimited AI (Claude) credits. A year or two ago, the first option was the obvious choice. With the latest improvements in AI tools, this question is no longer as clear. If AI tools keep improving at this pace, what is the answer a year or two from today? What about five years from today?

GenAI tools increase consultants’ productivity. One major firm reported that 75% of its consultants use their GenAI tool regularly, and 20% of entry-level consultants’ work has already been automated. The result is that consulting firms can now either: 1. Get the same amount of work done with a smaller team, or 2. Get more work done while keeping the team size constant.

The hiring data points to consulting firms going with the first option. Revelio Labs workforce data shows that the number of entry-level consultant roles at leading firms is down from its 2023 peak, while the opposite is happening for senior consultant roles.

The bottom of the pyramid is getting thinner. That doesn’t mean consultants disappear entirely. The role will shift, and the entry-level consultant will become less of a creator and more of an intermediary: reviewing AI outputs, shaping them for the client, and serving as a quality-control layer between AI output and partner judgment. If you’re a first-year consultant today and your pitch is “I can do research and build slides”, you have a problem.

The pyramid isn’t just shrinking. It is slowly reshaping into something closer to an obelisk: fewer layers and smaller teams, with humans organized around AI tools rather than manual output. At the base, an AI engine (agents) will handle production work. Above it, AI-enabled associates will direct and verify that output. Engagement leads will continue to frame problems and orchestrate delivery. Partners will continue to focus on what they’ve always done best: client relationships, trust, and strategic judgment. The AI-enabled consulting firm of the future will be leaner and more agile than what we see today.

Why AI Increases Demand for Consulting

You might expect all of this to mean AI is “killing” consulting as a viable career. The demand numbers suggest otherwise.

The US consulting market is projected to grow 7% in 2026, the fastest pace since the post-COVID boom. Three years into the AI revolution, demand for consulting isn’t falling. It’s at near-record highs. The AI revolution is the biggest structural reorganization of knowledge work in a generation, and companies will almost certainly need consultants to navigate it. A survey by Source Global Research found that 90% of companies plan to hire consultants specifically to help them implement AI.

Implementation Is the Bottleneck

History suggests that productivity gains from major new technologies grow slowly before taking off once adoption reaches critical mass. Most companies are still stuck in the early stages of their AI journey and are struggling to find success in implementing and keeping up with the rapidly improving technology. MIT research shows that despite ~80% of companies exploring or piloting ChatGPT or Copilot, only 5% of integrated AI pilots have extracted positive returns. Anthropic’s recent labor market report (see screenshot below) highlighted the massive gap between the theoretical potential of AI and its current usage across numerous industries. The technology works, but successful implementation is lagging. The primary issues are not the models themselves, but rather poor integration with existing business processes, lack of data readiness, and focusing on the wrong use cases.

There’s a historical pattern in the adoption of major new technologies. They are first adopted as point solutions. For example, when the steam engine arrived, factories used it as a drop-in replacement for water wheels, keeping the same floor layouts and workflows. The power loom was slotted into existing production lines. The real productivity surge in both cases came later, when manufacturers redesigned the entire factory floor around the new capability. AI is following the same arc. Most companies are still at the point-solution stage, using chatbots and copilots to automate individual tasks within existing workflows. The real gains will come from redesigning the entire organization around AI rather than plugging it into current processes. That reorganization hasn’t started yet. When it does, it’ll be the most complex transformation most organizations have ever attempted. That’s a textbook consulting problem.

The AI model companies themselves reached the same conclusion. Rather than try to cut consultants out, OpenAI and Anthropic announced multi-year partnerships with consulting firms. Their reasoning aligns with what I’ve been arguing: the bottleneck to enterprise AI isn’t model intelligence; it’s implementation (workflow redesign, change management, organizational restructuring). Consulting firms have been doing that work for decades. These are firms that have built deep, long-term relationships with their clients through previous waves of transformation (digital, post-merger, restructuring). When companies face the biggest organizational shift of a generation, they will turn to the partners they already trust.

What This Means for Consultants

So what does this mean for someone like me who is returning to consulting in September?

In my view, it means a smaller number of consultants will eventually do an increasing amount of consulting work. AI will change not only the economics of consulting firms but also the consulting experience. The parts of the job that no one enjoys (late nights formatting slides, weeks spent on manual research, the tedium of building an Excel model from scratch) are the tasks AI will automate first. What will remain is the work that drew most people to consulting in the first place: solving increasingly challenging problems, advising senior leaders, and shaping strategy.

The consultants who thrive will be the ones who can combine a deep understanding of AI tools with the softer skills a model can’t replicate. Managing client relationships, navigating organizational politics, and influencing change will become even more important in the future consulting toolkit.

If you’re early in your career, the implication is pretty direct: your value has to become judgment rather than just output. To achieve this:

• Treat AI like a sharp analyst you can’t fully trust: give it tight instructions, sanity-check the outputs, and build small repeatable workflows so you’re not starting from scratch every time

• Move up the stack faster: get good at framing the real question, spotting the hidden assumptions, and turning messy discussions into a clear decision with tradeoffs

• Develop taste: what belongs on the page, what belongs in the appendix, and what doesn’t belong at all.

My own bet is straightforward. I need to be the person in the room who understands these AI tools better than anyone else in my practice, not as a technologist, but as someone who can connect the technology to the client’s actual problem. I also need to keep developing the soft skills that become more valuable as the technical work gets automated.

Three years ago, I walked into my first consulting job the same month ChatGPT launched. I was excited and terrified. I still am, more of both, honestly. With that said, I’m optimistic. AI will not kill consulting. For those who master these AI tools and become comfortable operating in the chaotic new normal, AI will usher in a golden age of consulting.

What this golden age looks like in practice, and how the day-to-day of consulting will change, is something I plan to explore further in a future piece.

Subscribe now

Further reading:
1. David Maister, *Managing the Professional Service Firm* (1993) — leverage model and consulting pyramid economics

2. McKinsey’s internal AI platform “Lily” — 75% consultant usage, 20% junior analyst work automated. Via “The Rise and Fall of Corporate Consulting” analysis

3. Revelio Labs, “The View from 30,000 Feet: Tough Times Ahead for Consulting” — consultant-role demand down ~40% since 2023 peak. [reveliolabs.com]()

4. David S. Duncan, “AI Is Changing the Structure of Consulting Firms”

5. Stephen Foley, “Consultancies Set for Fastest Growth in Years on Back of AI Boom,” *Financial Times*

6. MIT, “The GenAI Divide: State of AI in Business 2025”

7. Anthropic Labor Market Report

8. OpenAI, “Introducing Frontier Alliances” — multi-year partnerships with BCG, McKinsey, Accenture, Capgemini.

9. PwC & Anthropic, “PwC and Anthropic Collaborate to Advance Enterprise Agent Deployment” — Anthropic-PwC partnership

Originally from Ethiopia, I grew up in Saudi Arabia, and spent most of my adult life in Canada. This January, with support from the Cañizares Center for Emerging Markets and the Foster Cunningham Fund, I traveled to Japan and Korea as part of a Cornell MBA trek with Mark Milstein, clinical professor and academic director of the Center for Sustainable Global Enterprise. This trip not only opened my eyes to East Asian business and culture but also gave me new insights into Ethiopia’s economic development.

What Ethiopia can learn from East Asia

Joe Studwell’s “How Asia Works” clearly defines the playbook behind Japan and Korea’s explosive capital-heavy growth. By directing capital toward strategic industries, Japan and Korea achieved rapid industrialization. Both countries protected early-stage companies in strategic industries until they were ready to compete in global markets. By tying government support to export performance, the countries created companies that competed globally and built the industrial capabilities needed to power long-term growth at home.

This pattern was present in many of the companies we visited. Take Sumitomo and CJ Group, both large, diversified conglomerates that grew within the East Asian development framework. Both companies relied on government support, which came in the form of patient capital and protection. This protection came with the expectation that the companies would compete internationally and bring valuable knowledge back home. This strategy succeeded, and both companies grew into large corporations that now span dozens of markets and industries. Government support was essential, but so was each firm’s ability to think in decades and execute accordingly.

This long-term orientation was evident in how these companies operate. At Sumitomo, the head of corporate development explained the company’s deliberate, consensus-driven decision-making process designed to align the organization around plans that would take years or even decades to play out. Whether with brand reputation, product quality, or international expansion, Sumitomo oriented itself toward long-term sustainable growth over speed. Smaller companies we visited across Japan, such as Kewpie, Iichiko, and Yanmar, echoed a similar mindset: Across the board, they emphasized sustained investment, long-term horizons, and global competition.

As I thought about how this framework might apply to Ethiopia, one company stood out immediately: Ethiopian Airlines. Much like East Asian conglomerates, Ethiopian Airlines benefited from early government backing. Today, Ethiopian Airlines offers flights to 140 destinations across five continents, operating at global standards while prioritizing long-term sustainable growth. More importantly, it has built capabilities that extend beyond a standard airline itself. This includes training 1,000 aviation professionals each year; running one of the region’s largest maintenance operations; and supporting smaller carriers across Africa. Ethiopian Airlines has evolved from a national airline to a platform for Africa’s aviation ecosystem. The secret was being pushed to compete internationally and responding with long-term execution.

Building companies like Ethiopian Airlines requires patient capital, which, in turn, requires governments to retain control over how capital is allocated and when domestic companies are exposed to competition. Opening markets too early can limit a government’s ability to provide sustained support for riskier industrial companies and projects.

This is one of my biggest concerns for Ethiopia. As part of an ongoing reform program supported by the International Monetary Fund, the country has liberalized its currency and started opening key sectors to foreign competition. On paper, these steps signal modernization and deeper integration with global markets. However, timing is incredibly important. Liberalization works best once domestic firms are strong enough to compete internationally. Opening too early often leads to capital flowing toward consumption, real estate, and short-term returns rather than long-term industrial investments. We have seen this pattern play out before — in Latin America during the 1980s and ’90s; in Southeast Asia before the 1997 financial crisis; and in various African economies under structural adjustment programs. Rapid liberalization in these countries brought short-term growth but left domestic industry underdeveloped and vulnerable to volatility.

Ethiopia has already demonstrated the power of patient capital with Ethiopian Airlines. It needs to develop many more homegrown global companies if liberalization is to deliver broad-based growth.

The precondition for growth

Ethiopia’s 10-year development plan lays out an ambitious path for industrialization, private-sector growth, and institutional reform. The strategy is progressive, but even the best plan requires long-term stability to execute.

Stability has proven difficult to secure in Ethiopia in recent years. In the first two decades of the 21st century, Ethiopia was one of Africa’s fastest-growing economies, growing at roughly 10% per year. Growth slowed to the mid-single digits in the early 2020s as the pandemic and prolonged conflict softened investor confidence. Industrialization takes decades and is not possible without the stability that makes long-term investment possible. Japan and Korea’s decades of relative stability from 1950 to 1980 laid the foundation for their meteoric rise.

The broader lesson for emerging markets

Japan and Korea are sometimes treated as unrepeatable miracles, products of unique cultural or historical circumstances. However, the companies I visited succeeded because their governments gave them time to develop, demanded that they compete globally, and allowed them to reinvest in capabilities rather than extract short-term profits. Ethiopian Airlines followed the same playbook and became a world-class institution.

The question for Ethiopia and for other emerging markets facing similar pressure is whether they will have the patience and stability to build long-term industrial growth. History suggests that countries that resist premature liberalization and maintain the capacity to direct capital strategically will outperform those that don’t.

In a world eager for quick returns, thinking in decades may be the most radical strategy of all.

Originally published at Cornell’s Emerging Markets website