AI Is the Missing Half of the Improviser's Culture

Why cultures that do the impossible finally have a technology for the possible — a practitioner's view by Dzhimsher Chelidze.

Two numbers have been bothering me for months.

In the 2025 Global Innovation Index, Russia ranks 28th in the world for human capital and research. For institutions and environment — 131st out of 139. The overall result: 60th place. A hundred positions between people and systems. World-class talent, working in an environment that cannot convert it. And the talent is not folklore: Russian university teams took the world finals of the ICPC — the top coding championship — in both 2024 and 2025.

We even have a national sport: raising geniuses for other economies. Alexander Popov demonstrated radio; others capitalized on it. Vladimir Vapnik built the theory behind machine learning; it runs in other people's products. Sikorsky and Zworykin — trained here, capitalized elsewhere. The price tag is measurable: Russia's import-substitution program spent about $2 billion on industrial software and earned roughly $17 million from replicating it — I break those numbers down here.

I run digital transformation in industry and build AI products. I have watched this gap from the inside for over a decade. And I believe generative AI has just changed the equation — not only for Russia, but for every culture built on improvisation rather than process. Let me argue the case.

Two kinds of cultures

Simplifying — deliberately — the world's business cultures sit on a spectrum.

At one end: cultures of process. The East Asian economic miracle was built here. Discipline, method, flawless reproduction of a procedure. A person as a reliable function inside a well-oiled system. This is not a caricature. It is the foundation of the corporations of Japan, Korea, and China — and for decades this model beat everyone who lacked its discipline.

At the other end: cultures of improvisation. Russians call it smekalka — the knack of doing the impossible with whatever is at hand. India has jugaad. The French have Système D. Americans say MacGyvering. The pattern is the same: brilliant workarounds, breakthroughs where the manual says "impossible" — and weak follow-through once the heroics are over.

Every real country and company mixes both. But the mix matters. And AI does not treat the two ends equally.

What the perfect executor changes

Look at what generative AI actually does well. It analyzes. It follows method. It reproduces a standard flawlessly, at any hour, without fatigue. It drives a task to "done" and packages the result — into a document, a deck, a product.

In effect, the market now sells a perfect executor for tokens.

Now ask the uncomfortable question. What happens to a trump card when anyone can buy a copy of it? It gets cheaper. AI does not attack the process culture's weaknesses. It attacks its main strength. Competing with a machine on discipline is a losing game — for everyone.

To be clear: Asia understood this first. China is building its own models at frontier pace, and its "AI+" national program is the most systematic government AI strategy I know — I analyzed it in detail here. The point is not that process cultures will lose. The point is that their traditional trump card has stopped being exclusive.

Meanwhile, the opposite asset is appreciating. AI already beats the average human on typical creative tasks. But in tests of non-standard thinking, the top 10% of humans still beat every model. The machine eats the typical. The singular gets more expensive. Improvisation was always singular.

The improviser's curse

So the improviser's strength is rising in value. Why am I not celebrating?

Because I know the other side of the ledger. No systemic follow-through. No method. Ideas that never get packaged. This is not a stereotype either — it is my daily work, and it is what a hundred-position institutional gap looks like in practice.

History repeats the profile. The Soviet Union secured the atomic project and space — where there was systemic management, personal accountability, and hard deadlines. The Tu-144 — the "Concorde-ski" — flew first and wasn't sustained. The Buran shuttle flew once. And OGAS, the 1970 project of a national computer network — "the Soviet internet" — was buried in bureaucratic correspondence. China drove its analogue all the way through. That is the difference between having ideas and converting them.

The same curse operates at company scale. I was once responsible for pulling a woodworking manufacturer out of crisis. On entry: staff turnover near 90% a year, six production chiefs in two years, lawsuits worth two monthly payrolls. And at the same time — people routinely pulling off impossible orders. Overnight, heroically, with pure smekalka.

We didn't start with technology. We built order: described the processes, set up planning, removed duplication and heroics. Before any investment in equipment, productivity grew 2.5 times — measured strictly, by machine time and output. On some product lines, three to four times. Same people, same machines.

Then came the second half of the lesson. Leadership relaxed back into the old style. Within months everything rolled back: delays, attrition, heroics. Order that lives only in people leaves with them.

The thesis

Here is the main thing I want to say.

For the first time in history, the improviser's weakness is closed by a technology. Reread the list: system, method, follow-through, packaging. That is exactly what AI does best.

Every previous wave of technology demanded order as a precondition. Taylorism, lean, ERP — first build the discipline, then collect the effect. That is why classic automation took root so painfully in improvisation cultures. Generative AI is the first technology that helps create order. It writes the regulation from the debrief. It keeps the minutes. It drives the task to "done." It packages the idea into a document, a pitch, a product.

The formula is simple: AI closes what the improviser lacks — and multiplies what the improviser has.

I test this daily on myself. My personal management contour — journal, planning, decision analysis, preparation for hard conversations — runs on an AI assistant. Decisions are still mine. But order no longer costs me my evenings. This practice grew into a product my team now builds — a personal operating system for managers: it keeps the memory of decisions, drives tasks to completion, automates routine planning, and holds the focus on strategic goals. One line about it, and back to the argument.

There is a leverage effect on top. One strong generalist with AI now covers the task pool of a whole department. Breadth of vision used to be a "career weakness" against narrow specialization; AI lets you rent the missing depth. The bottleneck stops being headcount. It becomes the number of people who can frame a task and verify a result.

The window will not stay open

Two warnings before the recipe.

First, the low-base effect works only while the base is low. Process cultures are already pairing AI with their discipline — China's "AI+" sets penetration targets of 70% by 2027 and 90% by 2030, with an "intelligent economy and society" by 2035. The complement-versus-substitute asymmetry is a head start, not a guarantee.

Second, June 2026 showed everyone whose hand is on the switch. The United States cut access to Anthropic's frontier models for almost three weeks — including for allies: Canada, France, Germany, Japan. Access was restored — on Washington's terms. Access to someone else's model is a switch in someone else's hands. Building a strategy on it means having no strategy. The architectural conclusion is universal: the model is a component, not a foundation.

What to do with the advantage

For leaders of improvisation cultures — a country, a company, or a team of brilliant firefighters — the play rests on three pillars.

Management. A systemic approach, not a campaign. Here the benchmark is, ironically, China: "AI+" is a grounded ten-year framework that pushes AI down into industries, with named accountability and measurable targets. Improvisation cultures need exactly this kind of boring decade.

People. Bet on the pairing "talent + AI systematizer" rather than on drilling process discipline into heroes. Teach two skills above all: framing tasks and verifying results. And deliberately grow your generalists — rotations, adjacent domains, work at the intersections. Erudition plus rented depth is the new scarce profile.

Infrastructure. You do not need a sovereign super-model to start. Light, specialized models wired into real processes are cheaper, faster, and controllable. A factory doesn't need a billion-dollar frontier model. It needs a trained assistant for its own process — and clean data underneath.

And keep the order you create in systems — regulations, data, IT — not in people. Otherwise it leaves with them.

One more thought, half a step ahead of this article. Brains and conveyors complement each other. Improvisation cultures generating breakthroughs, process cultures scaling them — that is a stronger pair than either alone. But you enter such a partnership as a subject, with your own schools, models and niches. A "brain" seat is not granted. It is taken. Otherwise you simply trade one switch for another.

Will we use it?

I return to my two numbers — 28 and 131 — and to our national sport.

The hundred-position gap was always read as a verdict. I read it now as a window. The strength it measures is appreciating; the weakness it measures has, for the first time, a technology that closes it. This is true for Russia, and it is true for every improviser culture — and for every company of strong people and weak processes anywhere in the world.

Windows like this have been missed before. Twice, in our case.

Cultures that do the impossible finally have a technology for the possible. The only question left is the old one. Will we use it — or keep raising geniuses for other economies?

Dzhimsher Chelidze is a digital transformation executive (CDTO) in industry, author of books on digital transformation and AI, and founder of BAEOS, an AI-based operating system for managers.