© 2026

· WORK

The Workplace AI Paradox. Why AI amplifies friction in crowded companies

AI amplifies whatever it touches. In a small, strong team it amplifies output. In a crowded one it amplifies friction.

01

Solow’s line

In 1987 Robert Solow wrote that you could see the computer age everywhere but in the productivity statistics. AI at work looks the same. In 2025 Faros tracked over 10,000 developers: those using AI merged 98 percent more pull requests, review time rose 91 percent, and company delivery metrics did not move.

The economist Paul David explained Solow’s puzzle with electricity. Early factories swapped the steam engine for one big electric motor and kept the old shafts and belts. The gains came decades later, when factories were rebuilt around the motor. Most companies are at the first stage with AI: a faster motor on the old line, feeding the same approval queue.

02

AI amplifies friction

Fred Brooks observed in 1975 that adding people to a late software project makes it later. Communication grows faster than headcount: n people have n(n−1)/2 possible channels. Ten people have 45. A hundred have 4,950.

AI puts more traffic on every channel. Each person can now produce more drafts, more messages and more code for everyone else to read, review and fix. Researchers at Stanford and BetterUp call the worst of it workslop: AI output that looks finished and isn’t. In their 2025 survey, 40 percent of US desk workers had received some in the past month, and each instance took almost two hours to deal with.

So AI does not fix a crowded organisation. It turns the volume up. Where many people already create friction, AI creates much more.

03

Fewer, stronger people

In 2001, after the dot-com crash, Netflix had to lay off a third of its 120 staff. Reed Hastings kept the 80 strongest and expected morale to collapse. Instead, he wrote, teams were accomplishing more, faster. He called it talent density.

A 2006 study by Will Felps and colleagues points the same way: one person behaving badly in a small team cut its performance by 30 to 40 percent. Friction is not additive. It spreads.

In 2022 Elon Musk bought Twitter and cut it from just under 8,000 staff to about 1,500 within six months. The business paid a real price as advertisers left, and it is not a recipe. But the site kept running. It showed how much of a large company is the work of coordinating itself rather than the product.

04

Hidden gains do not spread

In 1943 Bryce Ryan and Neal Gross found that Iowa farmers adopted hybrid seed corn after watching a neighbour succeed. Everett Rogers later named observability as one of five traits that speed adoption: people adopt what they can see working.

Inside companies, AI use is the opposite of observable. Ethan Mollick calls these people secret cyborgs: they use AI and keep quiet, because it is banned, because human-looking work is valued more, or because they fear being replaced. A gain no one can see cannot spread or be reorganised around.

05

Signal, not headcount

The fix is not more tools. It is fewer handoffs, higher talent density and AI use made visible. A small team of strong people with AI will outrun a large team with AI, because the amplifier works on both and only one of them is mostly signal.

AI amplifies people. Make sure what it amplifies is worth hearing.

KEEP READING

  1. 01The World Is Not Logged. Why real-world data, not intelligence, limits AIThe world records everything, in light, heat and wear. Almost none of it reaches a model. That gap, not intelligence, is what limits AI.Deep Tech
  2. 02System 1 Is Solved. AI agents now fail at spatial reasoning, long horizons and contextFast, perceptual work is done: agents read screens, click and type as well as people. What breaks now is the slow work of holding a space, a long task and a lot of context in mind.Deep Tech
  3. 03Reverse Benchmarking. Measure AI on what the leaders neglect, not what they optimiseBenchmarks make every lab chase the same scores. The useful question runs the other way: where does the best model still disappoint?Deep Tech
  4. 04Leveling Up Intelligence. A game analogy for RL environments, agents, benchmarks and evals“RL environment” is a buzzword thrown around as if it were a new machine. The loop it names is simple. What is hard is deciding which problem the score is supposed to stand for.Deep Tech
  5. 05You Cannot Hear a Fake. Why voice deepfake detection fails, and what replaces itDeepfake detection is an arms race the detector loses. Stop asking “is this fake?” and ask “can this prove it is real?”Speech
  6. 06The Voice Came Loose. The state of speech AI: cloning, codecs and super-resolutionSpeech is now editable, and a voice is a parameter, not a person.Speech
  7. 07Deconstruct to Be Reborn. Taking the self apart to find what is load-bearingYou were born one way. You can be born again, if you let yourself be taken apart first.Philosophy
  8. 08Comparison Is the Thief of Joy. But it is also the one-sided window of potential, and only you can act on itComparison steals joy when you read it as a verdict. Read it as a window and it shows you what is possible. Everyone builds an ego out of what they compare themselves to. Some build it proud and some build it to please, and both end up serving the crowd. So measure yourself against something perfect.Philosophy
  9. 09Problem Relativity. The size of a problem depends on where you standA problem has no absolute size, only a size measured from where you stand.Philosophy
  10. 10The Issue Is the Primitive. Running companies of humans and AI agents on Linear’s modelThe best operating system for a company of humans and agents already exists: the issue tracker.Systems
  11. 11The Curse of Knowledge. Why experts explain badly, and prompts fail the same wayOnce you know something, you can’t imagine not knowing it. That ruins how you teach, write and prompt.Mind
  12. 12The Story Is Not the Wall. What high agency is, and why AI makes it matter moreHigh agency is treating the limits other people describe as a story you may edit.Agency