نبذة عني
I never wanted to be an entrepreneur.
Getting fired two weeks into my internship…
Getting fired two weeks into my internship…
النشاط
٥٦ ألف متابع
-
Usman Sheikh شارك ذلكUber exhausted its AI services budget four months into the year. Its CTO is now talking about going back to the drawing board on AI strategy. Eleven percent of backend code comes from AI agents, and engineers are ranked on leaderboards by tool usage. Adoption looks impressive on every dashboard. No one has a clear answer for whether the output was worth what they paid. This is what happens when tokens become the new vanity metric. The billable hour had decades of estimation infrastructure behind it. A token has none of that yet. Buyers are running blind, boasting about consumption the way firms once boasted about headcount. Tokens consumed is not the metric. The metric is whether each engagement leaves behind a system that makes the next one better. McKinsey's public boast of a hundred billion tokens with OpenAI is the loudest version of the same pattern. Adoption proof standing in for delivery proof. The firms winning today are building different infrastructure: encoded playbooks the system can apply, and the telemetry to track which ones are working. From the outside the output looks similar. The substrate is not. The question is no longer how to adopt AI. It is whether the work you're doing today makes tomorrow's work more efficient to deliver. Link to the full essay in the comments.
-
Usman Sheikh شارك ذلكMcKinsey says a quarter of its revenue is outcome-based, and PwC's CEO has publicly committed to move the firm off hourly billing. Yet when you look at public equities in the consulting and BPO space, margins and revenue‑per‑head aren’t moving the way that story implies. Outcome-based pricing has two versions, and the debate conflates them. Version 1: Changes the invoice. The firm prices on outcome, the client pays against it, and the next engagement starts from scratch. Version 2: Changes the work. Each engagement captures which parts followed a pattern and which required judgment, so delivery hours per outcome fall over successive engagements while the outcome price holds steady. Only Version 2 compounds. The tell is operational and lives in the firm's own data: on a repeatable service, do delivery hours fall across successive engagements? If yes, the work is changing. If not, only the invoice is. However, Version 2 has a procurement problem. A compounding judgment library cannot be benchmarked and cannot be exited cleanly, and procurement refuses what it cannot price. So the real move is happening one layer down. Accenture's 2025 Telstra deal was a $700M AUD seven-year equity JV, not an SOW moved to outcome pricing. A different contract form entirely. The invoice is not what's changing. The contract it sits on top of is. Link to the full essay in the comments.
-
Usman Sheikh شارك ذلكGallup's latest report says low engagement cost the global economy nearly $10 trillion in lost productivity last year. The data is from 2025. The signals existed twelve months earlier, in conversations managers were having. The measurement arrives annually while the productivity loss is felt daily. A founder I work with is building in that gap. His conversational AI surfaces disengagement signals daily that annual surveys would catch next year. In one case, his platform detected a disengagement pattern in a department nobody was watching, weeks before the first resignation appeared. There is pressure on him to charge for the outcome: retention. He can't, because the outcome is entangled. When someone stays, the reason is a combination of their manager's response, a compensation adjustment that happened independently, a team restructure nobody planned and several factors no platform ever measured. We will probably never prove that a specific intervention caused a specific employee to stay the way you can prove a contract clause was reviewed correctly. The opportunity isn't at the outcome level where attribution is permanently entangled. It's at the atomic signal level, where individual detections can be verified against observable outcomes. Most organizations evaluating these tools see an engagement platform and run procurement like they'd buy SaaS. These tools behave like signal infrastructure, and the organizational self-knowledge they accumulate is what a competitor can never reconstruct. The full essay breaks down where the real moat forms and why it's not where most people are looking. Link to the full essay in the comments.
-
Usman Sheikh شارك ذلكAsk a senior partner what makes a good client memo and you'll get a confident answer. Ask them to write those criteria down so a system can score against them and the conversation will dramatically slow down. That gap is the most critical problem in professional services right now. AI collapsed the cost of producing work. It didn't collapse the cost of knowing whether the work was any good. Every firm has quality systems, but they were designed for a world where a senior person could keep pace with what a junior person produced at human speed. That ratio is broken today. Output is moving at machine speed while judgment is trying to keep up at human speed. Firms like Crosby who just raised $60m are raising it on the back of these libraries of encoded judgment. A competitor can launch tomorrow using the same models but they can't reconstruct that library and the sequence it was built on. Most firms are betting better models will close the quality gap. The other bet is that the gap gets closed by the firms that define what "good" means in their own domain and build the scoring infrastructure around it. Link to the full essay in the comments.
-
Usman Sheikh شارك ذلكResolution pricing will never work for professional services. I’ve heard that more than a few times as the NewCo thesis has sharpened Earlier this week Crosby raised $60 million to process standard contracts in under an hour. They charge per document. Traditional law firms bundle pattern work and judgment into the same billable hour. Crosby separated them. Recognizable clauses go through AI and lawyers who carry personal liability take the judgment calls. When the bundle breaks, the pricing follows. Every LegacyCo is angling for a deal with a Frontier Lab right now. PE-backed OpenAI partnerships, Accenture acquiring AI implementation firms. The thesis: embed AI into existing organizations and make them leaner. Leaner is not the same as restructured. The test is whether the pricing model changed. If the client still pays the same way for the same bundle, nothing was restructured. The cost of operation may have become more efficient but that is a temporary advantage until everyone catches up. We are entering a world of unlimited labor: output without wages, scale without headcount. In that world, the question is whether firms built to distribute can learn to reinvest before firms built to compound make it irrelevant. Link to the full essay in the comments.
-
Usman Sheikh شارك ذلكThe people training AI models today are often the same ones the models will replace. Verge recently covered Mercor and other AI data training labs showing displaced professionals writing rubrics that define what "good" looks like. When the model can handle everything they throw at it, they are done. Their judgment has been encoded and the contract ends. What made that piece dark was a subtle distinction. Many of these people thought the skill was their ability to use the tools, to build the models or translate correctly. That was scaffolding around the skill, not the skill itself. The real skill was knowing what the output should look like and the impact it needs to deliver. The models are learning the scaffolding, not the thing underneath. The path from junior to senior used to build that instinct. Years of error-correction, sitting with work before you understood why it mattered. That path is shrinking from both ends. Companies can't justify the investment in juniors given the pace of change. Mid-level process-heavy professionals are having their skills encoded one contract at a time. AI isn't replacing judgment; it's replacing the path that builds it. Link to the full essay in the comments.
-
Usman Sheikh شارك ذلكMarc Andreessen recently shared his experience of working at IBM at the height of their power. He counted twelve layers of management between him and the CEO. Each layer put a little spin on what they reported upward. Not with malice, but because each person wanted to look good. One layer of spin is tolerable. Two or three and it starts to compound. Twelve layers, and the CEO has no idea what is happening inside his own company. His story explains why AI pilots keep failing at enterprises and why the solution is always to "transform" the business. Clean up the data, add context and suddenly all of our problems will be solved. An elegant story. One selling billions of dollars of consulting work today. What it misses is the gap between what organizations say they value and what they actually measure. Inside that gap are tireless humans who absorb contradictions daily, holding two realities at once. Technology can't absorb those contradictions. It optimizes for the signals and instructions encoded. Between the idea and the reality, between the motion and the act, falls the shadow. That shadow is where most organizations live. This essay is about what happens when it becomes visible. Link to the full essay in the comments.
-
Usman Sheikh شارك ذلكI gave a talk at the University of Toronto last week. Eighteen minutes about a pattern I've been noticing in myself for twenty years: a voice that shows up whenever things go quiet. It sounds like me, uses my vocabulary and its only job is to fill the silence with something productive. After the talk, a student raised his hand and asked: "So what do I actually do with this?" He had just watched me describe a pattern that turns everything into a prescription for what to do next. His first instinct was to ask for one. He wasn't doing anything wrong. He was doing what every high-performer in the room would do, what the culture trained all of us to do. Turning any insight into instruction. I grew up at a dinner table where one message ran through every gathering: nothing worth having comes without sacrifice. That message became the driving force which built my career. It also resulted in sacrifices that included both relationships and health. It became an engine that didn't come with a way to stop. When a business I was running fell apart, the engine stalled. At a kitchen table in Karachi my mother said something she'd been saying my whole life and for the first time I actually heard her. A few years later when things were going well, she tried again at a restaurant in Singapore. I couldn't hear her that time, the engine was back. For a long time I thought this was just my story. Only recently have I started noticing the same pattern accelerating. Dawkins called ideas that spread this way memes, before the internet stripped the word of meaning and seriousness. Ideas don't replicate because they're true. They replicate because they're easy to pass along. "Nothing worth having comes without sacrifice" survived in my family because it compressed decades of immigrant struggle into a sentence that sounded like wisdom. Now the same selection pressure operates through algorithms built to surface whatever is easiest to share. "Start something" "Build your personal brand" None of these are evaluated for whether they're actually true for the person hearing them. The only thing close to the truth is a quiet feeling that something resonates or doesn't, before you can explain why. That feeling is the thing worth paying attention to. But you can only hear it when the engine isn't running. Full essay linkedin in the comments.
-
Usman Sheikh شارك ذلكWPP traded close to $120 in 2017. This week it's nearer to $17. Dentsu tried to sell its international business, $4.5 billion in net revenues and nobody bought it. You could blame management. But Accenture, one of the best-managed services firms on the planet, is down 40% from its high. Capgemini is down 32%. Even Publicis, which invested in data infrastructure years ahead of the curve, is down 14% after the market reacted to its 2026 growth outlook. Omnicom closed its IPG acquisition and immediately cut 4,000 positions and retired three agency networks. You can't dismiss all of that as bad management. The market is repricing the structure, not cherry picked operators. Performance marketing has become the economic engine inside these holding company structures. When the performance layer automates, that base erodes. Brand strategy has to justify itself independently. The deeper problem sits across all of professional services: amnesia as a service. When billing is time-based, institutional memory becomes a cost center. The firm that remembers gets punished and the one that chooses to forget gets rewarded. The best agencies did build proprietary knowledge. Behavioral science practices and research methodologies. But that knowledge was organizational, not systematic. Too dependent on specific people to compound independent of them. Nobody built a resolution library because amnesia paid more. The billable hour model financially rewarded rediscovery and penalized institutional memory. That trend is starting to reverse. The firms best positioned to make the move are the ones most incentivized not to, because amnesia is still more profitable than memory. The clock isn't AI's rate of improvement. It's how long clients take to notice what they're actually paying for. Full essay has where I believe the industry could pivot too. Link in the comments below.
Usman Sheikh قام بالتعليق على منشور
١ يوم
https://framebreak.com/posts/tokens-are-the-new-headcount-04302026
Usman Sheikh قام بالتعليق على منشور
١ أسبوع
The way I see NewCos sidestepping is they don't arrive as one big contract. It will look like Crosby where they take a wedge, pull the data, build trust, expand. Same as the land-and-expand strategy professional services has always used, only this time we're placing agents at key points.
I wrote about this in my article last week on the entanglement problem.
https://framebreak.com/posts/the-entanglement-problem-04162026
-
Usman Sheikh أضاف إعجابًا إلى ذلكUsman Sheikh أضاف إعجابًا إلى ذلكLet's be honest, building Qaflah hasn't been easy! But I can make your startup journey much easier, more in the video! Startup Syndicate / Urooj Zia / Ahmed Ayub / Usman Sheikh / Mahmood Mirza / Farzal Ali Dojki
التراخيص والشهادات
-
Certified Psychometric Consultant
Psychometric Centre, University of Cambridge
تم الإصدار في -
Certified Neuro Linguistic Practitioner
Adam Khoo Learning Technologies
تم الإصدار في
اللغات
-
English
إجادة اللغة الأم أو إجادة لغتين إجادة تامة
-
Urdu
مستوى إجادة محدود
التوصيات المستلمة
13شخصا قدموا توصية لـUsman
انضم الآن لعرض