What leadership teams want from AI in strategy

How the data was collected

In the weeks before the NAG Executive Forum on 3 September 2026, eight executives from Finnish companies took part in a short AI-led interview run on Insight Bridge, NAG’s AI platform for strategy execution. The group included three CEOs, two CFOs, a managing director, a strategy and transformation lead and a CHRO.

They described how their companies use AI in strategy work today, what it has changed, their biggest barrier and what they want next. Prof. Tomi Laamanen of the University of St. Gallen mapped the answers onto six stages of the strategy process (sense, interpret, challenge, commit, mobilize and learn) and presented the results in his keynote. The sample is small, so the numbers are indicative rather than representative.

1. AI sits at the front end of the strategy process

Most companies use AI where strategy work begins: scanning markets, megatrends and regulation, and summarizing evidence. Four of eight also use it to challenge their thinking through scenarios, option iteration and cross-checking. Only one uses AI at the commitment stage, and only to make it faster. No company yet closes the loop from a decision to its outcome and on to the next decision.

2. Speed for many, a changed decision for few

Six of eight report proven analytical gains: the same analysis in less time, more inputs without a bigger team, or earlier and deeper preparation for strategy discussions. Only one could point to an AI insight that led to a concrete strategic move. Customer data revealed customer groups the company had not seen before, and the strategy answered with a new key account model.

“Most likely we’d come up with the same decision, only slower without AI.”

Prof. Laamanen proposed a simple test for every use of AI: what would we have decided without it? If the answer is “the same, a week later”, AI has delivered efficiency. If it is “something we could not see”, it has delivered strategy.

3. The barrier is readiness, not technology

Six of eight named a readiness gap as their biggest barrier, and none named the tools. The barrier also changes as maturity rises. Companies that are just starting name skills, resources and habits, which is a training problem. Companies that are scaling name company-specific context, trust in AI outputs and working habits, which is an organization-design problem that no software vendor can solve for them. The most advanced respondent named only one barrier left: imagination about how to use AI.

4. Leaders want a challenger, not a decider

Asked what they want next, six of eight want AI as a challenger, sparring partner or scenario engine. None wants AI to take the decision. The verbs are changing: today AI is used to summarize, search and speed up; tomorrow leaders want it to challenge, spar, simulate and find risks outside the traditional risk map.

What the research says

In his keynote, Prof. Laamanen set these findings against peer-reviewed experiments and company cases. Three lessons stood out:

• AI can make wrong answers more convincing. In a preregistered experiment with 758 BCG consultants, those using GPT-4 produced work rated more than 30% higher in quality on tasks within the model’s capabilities. On a strategy problem outside those capabilities, correctness fell from 84.5% without AI to 70.6% with it, and the wrong answers read as more coherent. (Dell’Acqua et al., “Navigating the Jagged Technological Frontier”, Organization Science.)
• AI widens the options but can weaken the choice. In a field experiment with 776 Procter & Gamble professionals, individuals using AI matched the quality of two-person teams without AI. Yet AI users chose their own best idea about 37% of the time, against about 50% for teams without AI. (Dell’Acqua et al., “The Cybernetic Teammate”, Organization Science.)
• Irreversible commitments need limits. Zillow’s algorithmic home-buying business wrote down $304 million of inventory in the third quarter of 2021 before it was closed. Forecast error was amplified by acquisition speed, renovation capacity and capital exposure. (Zillow Group SEC filings, November 2021.)

Five design rules for AI in the strategy process

Prof. Laamanen closed with five design rules:

1. Ground before you generate. Connect AI to the evidence only you hold.
2. AI widens, humans choose. Aggregate several AI evaluations and keep a human selection gate.
3. Test tasks, distrust polish. What AI does well varies task by task, so verify its work independently.
4. Match autonomy to reversibility. The less reversible the commitment, the more AI should challenge and the less it should decide.
5. Build a responsive organization. Habits, champions, short planning cycles and evaluations that keep running.

His suggested starting point fits in 90 days: name one consequential decision, define what “better” means beyond speed, ground AI in the relevant evidence, set evaluations, AI’s role and no-go bounds, and record the decision so you can learn from the outcome.

Where Insight Bridge fits

One theme ran through the whole forum: a company’s own employees are an underused source of insight, both in strategy work and in execution. Insight Bridge, which NAG used for these pre-forum interviews, runs AI-led conversations across a whole organization at once, with the speed of a survey and the depth of an interview. Every finding can be traced back to what participants actually said.