A company AI strategy: where to start when pilots are already running
According to 2025–2026 surveys, only 26% of companies have an AI strategy — even though 97% report adopting AI. If pilots are already running, the place to start is not yet another tool but four steps in order: take stock of what you already have, diagnose management maturity across six domains, choose two or three directions with measurable impact, and assign owners with a review rhythm.
97% adopt AI. Only 26% of them have a strategy
Report adoption
of large companies use AI in at least one process — the technology stopped being a rarity long ago.
See no impact
of companies get no measurable return from their pilots: the tool is up and running, but no one is in a position to decide its fate.
Have an AI strategy
of companies can say who is responsible for AI, how it is built into decisions and how the result is measured.
By mid-2026 this is no longer a question of competitive advantage but of manageability: departments pile up a dozen scattered subscriptions and AI initiatives, and sooner or later someone has to bring them into one loop with clear rules.
The gap between “we’re adopting” and “we’re managing” is the gap between a tool and a strategy. If pilots are already running, the task is not a sixth pilot but a management framework around the first five — and that is what Executive AI navigation is about.
Take stock: what is already running, and who started it
Before you formulate a strategy, put together an honest list of what is already going on. In mid-sized companies there are usually more pilots than the top executive thinks: marketing is testing text generation, sales is transcribing calls, accounting is trying document recognition, someone in IT is experimenting with a coding copilot. Each pilot was launched by its own department, on its own budget, with no common owner.
The inventory is a five-column table: what the pilot is, who initiated it, what the budget is, what the success metric is, and who decides whether it continues. If a pilot fills three columns out of five or fewer, you are looking not at an experiment but at an expense line with no owner. Pilots like these are the ones that most often end up among the 54% that show no measurable impact.
The output is not a multi-page report but a single one-spread table that you can present to the board in five minutes. That is enough to move to the next step: understanding why some pilots are stalling while others already deliver impact.
A maturity diagnostic: six domains instead of one IT department
A pilot from the inventory is almost always held back not by the quality of the model but by the maturity of the management around it. In the QAC model there are six such domains: Power (the decision-making loop), Tech&Data (data and infrastructure), Processes (business processes), Innovations (change management), Culture (people’s readiness) and Emotions — the team’s emotional climate, a measurable management factor rather than “soft context.”
For each pilot from step 1, it helps to ask: which domain does it mainly live in, and can that domain carry the load? A data pilot runs into Tech&Data if the data is scattered and no one is responsible for its quality; a sales pilot stalls in Culture if the team sabotages the new tool. A quick answer comes from the QAC express — a questionnaire and a review within a week; a full QAC audit turns all six domains into maturity indices and a roadmap.
Along the way, the diagnostic shows how tasks naturally route: the structural domains — Power, Tech&Data, Processes, Innovations — are handled by the consulting side of the practice, while the human ones — Culture and Emotions — move into work with Galina Malysheva’s Quantum Coaching method (sessions in Russian).
Choose two or three directions with measurable impact
After the diagnostic you can see where a domain already carries the load and where a pilot is dragging the whole organization along with it. Strategy is, first of all, saying no: out of five to seven pilots, companies usually pick two or three to develop further — where domain maturity and potential impact overlap. The rest are paused or shut down, and that too is a management decision, not a failed pilot.
A “strategy” is not a list of tools
AI subscriptions, copilot licenses and a dozen parallel pilots are not a strategy but a shopping list. A strategy answers three questions: who is responsible for the decision, how it is built into the process, and how the result is measured. If even one has no answer, what you have for now is tools, not management.
Owners and rhythm: who is responsible and how often you review
Each chosen direction needs one owner — not “IT in general” but a specific executive whose own KPIs include the pilot’s metric. More often this is the head of a business function rather than the CIO: the chief commercial officer owns AI in sales, the COO owns AI in production.
The second condition is a review rhythm. Maturity indices are not calculated once and filed away: a sensible practice is to revisit them once a quarter and check whether the domain has grown and whether the impact has held up. A strategy without a review rhythm turns back into a list of pilots within six months — only a more expensive one.
What people ask most often about AI strategy
How is an AI strategy different from a list of AI tools?
A list of tools records what has been bought and launched. A strategy answers three questions: who is responsible for an AI-driven decision, how it is built into the process and how the result is measured. Without an answer to even one of them, a dozen subscriptions and pilots don’t add up to a strategy.
Where do we start if we already have five unconnected pilots?
Start with an inventory: the initiator, budget, success metric and person responsible for each pilot. Then run a maturity diagnostic across the six QAC domains to see which pilots rest on a mature domain and are ready to scale, and which are dragging the whole organization along.
How long does it take to develop an AI strategy?
An express diagnostic and a draft of priorities take one to two weeks — enough to see whether it’s worth going further. A full QAC audit with a roadmap for two or three directions usually takes 2–4 weeks, after which implementation support begins.
Written by Sergey Parshakov, a C-level practitioner with 33 years of management experience and the author of the QAC model. Author’s page → (in Russian)
Start with a diagnostic, not a sixth pilot
30–40 minutes with a partner: we’ll map your current pilots across the six QAC domains and see where you already have a managerial lever and where it’s just spending.