New contributor rates take effect January 2027. Review the contribution guidelines for full details.

AI in Management Decisions

AI makes options and scenarios cheap, so the judgment behind each decision matters more

AI in Management Decisions
Idea In Short

Most companies now use artificial intelligence [AI] somewhere, but few get measurable value from it, and the gap is mostly a management problem. In strategic decisions, AI is most useful for widening the set of options, stress-testing plans against several futures, arguing against a proposal and assembling evidence quickly. It is least reliable on the one-off, hard-to-reverse bets that define a strategy. Leaders who get results redesign how decisions are made, keep a named person accountable for every AI-informed choice and train managers to question model output rather than accept it. The decision in front of a leadership team is where to add AI to its own decision process first, and who will check what it produces.

What does AI change about strategic decision-making?

It makes generating options, scenarios and counter-arguments cheap, so the limiting factor becomes judgment: deciding which output to trust and taking responsibility for the call.

Which management decisions should AI not make on its own?

Decisions that are hard to reverse and have little relevant historical data, such as acquisitions or entering a new market. AI can inform them, but people should decide and own them.

Is AI in management paying off yet?

For most companies, not reliably. McKinsey's 2026 survey shows near-universal adoption, but only a minority of firms report any measurable financial impact.

What do companies that see results do differently?

They redesign workflows around AI instead of adding a tool to an old process, and their senior leaders visibly own the change.

How can managers use AI to challenge their own plans?

Ask it to argue against a proposal from several viewpoints or to run a pre-mortem, and do so before revealing a preferred option, since models tend to agree with the person prompting them.

Why is AI output risky in high-stakes decisions?

Its accuracy varies across tasks that look alike, and a fluent answer reads the same whether it is right or wrong. Every figure in an AI-drafted evidence pack needs checking against its source.

What should leaders learn about AI?

Where models are strong and weak, how to question an output, where the data comes from and which risks the company carries.

Where should a leadership team start?

Pick one recurring decision, map how it is made today and add AI to a single step, usually option generation or the pre-mortem. Then compare the quality of the debate with previous cycles.

Will AI replace managers?

The evidence points to a change in what managers do rather than their removal: less time gathering information, more time framing questions, checking evidence and owning outcomes.

Which mistakes should managers avoid when using AI for decisions?

Treating a polished AI summary as checked analysis, showing the model the preferred answer first, automating a flawed decision process instead of redesigning it, buying tools before deciding which decisions they should improve and leaving middle managers out of the redesign.

Most writing on AI in management stops at automation: meeting summaries, faster reporting, a chatbot answering human resources questions. Those gains are real, but they sit at the bottom of a manager's job. The harder question is what AI does to the top of it, where leaders decide where to compete, what to fund and what to stop, and how they keep authority over choices a model helped shape.

This article focuses on that harder question. It covers which decisions AI helps with, five practical uses with examples, what the research says about results, and how the leadership role shifts once AI sits in the planning room.

Which management decisions AI helps with, and which it doesn't

A practical way to sort management decisions is along two lines: how much relevant data exists, and how easily the decision can be reversed. The four combinations give AI different roles:

  • Data-rich, easy to reverse (weekly pricing tests, shifting campaign budgets): AI recommends and, within set limits, acts
  • Data-rich, hard to reverse (closing a plant, consolidating suppliers): AI builds the evidence base and models scenarios, and leaders decide
  • Data-poor, easy to reverse (piloting a new service line): AI generates options and pre-mortems, and small experiments settle the question
  • Data-poor, hard to reverse (acquisitions, entering a new market, a major pivot): AI is a sparring partner at most, and judgment and accountability stay with people

Strategic decisions cluster in the bottom two rows. A model trained on historical data is weakest exactly when the decision depends on a break from the past, which is what most strategic bets are. So the most useful applications of AI in strategy widen and test the options a leadership team considers, while the choice itself stays with the team.

Five ways to use AI for strategic decisions, with examples

1. Widen the set of options before anyone commits

Planning discussions tend to settle on the first credible idea, often the one the most senior person proposed. Asking a model for fifteen possible responses to a competitor's price cut, or ten ways to reach a growth target without discounting, takes a few minutes and changes what the group debates.

The strongest evidence here comes from a field experiment with 791 product professionals at Procter & Gamble, run in 2024 by researchers from Harvard Business School, Wharton and P&G. Individuals working with a GPT-4-based tool produced ideas of the same quality as two-person teams without AI. Teams using AI were three times as likely as individuals without it to produce ideas in the top 10%. The study also found that without AI, research and development [R&D] staff proposed technical solutions and commercial staff proposed commercial ones, while participants using AI produced ideas that combined both. The tool partly offset the functional silos that shape so many strategy discussions.

2. Stress-test the strategy against several futures

Traditional scenario planning is expensive, so many companies run it once a year and file the slides. Generative AI makes it cheap to build scenarios, revise them as conditions change, and keep them in use. Think Insights has looked at how generative AI remakes scenario planning at Coca-Cola, JPMorgan, and Shopify, and the AI-enhanced strategy and scenario simulation playbook sets out a step-by-step process.

A simple prompt pattern works well for a first pass. Give the model the strategy and the three assumptions it depends on most, then ask for a scenario in which each assumption fails, plus the early signals that would show it is failing. The output that matters is the list of signals, because it tells the team what to monitor after the decision is made.

3. Ask AI to argue against your plan

AI makes a useful critic because it has no stake in the plan and no reason to defer to the person who proposed it. Ask it to attack the proposal as a skeptical investor would, then as a competitor's strategy team, then as a regulator. Or run a pre-mortem: "It is two years from now and this initiative has failed. Write the most likely explanation."1

These exercises target two biases that distort leadership decisions: giving too much weight to senior voices (authority bias) and to whatever happened most recently (recency bias). They work best as part of a wider effort to reduce bias in how an organization decides.

Language models also tend to agree with the person prompting them. If a manager states a preferred option first, the critique softens. Ask for the case against the plan before revealing a view, or ask in separate conversations for the strongest case for each option.

4. Track the strategy between planning cycles

Strategies usually drift during execution, slowly and between quarterly reviews. AI can read sales call notes, support tickets, competitor announcements and industry news each week and flag where events diverge from the planning assumptions. Teams that run OKRs can have AI summarize progress against each key result and explain why the off-track ones are slipping.2 This shortens the time between a warning signal and a conversation about it, which is where the gap between strategy planning and execution tends to open.

5. Prepare the evidence for high-stakes decisions

For board-level choices, assembling the evidence often takes longer than debating it: market sizing, competitor filings, customer research, internal performance data. AI can draft an evidence pack in hours instead of weeks. The manager's work then shifts to checking it, which means tracing every number to a source, asking what is missing and spotting where the model filled a gap with a plausible guess. A well-formatted pack can still contain invented figures, so the check matters most when the document already looks finished.

What the research says about results

McKinsey's latest State of AI survey, fielded in May and June 2026 with 1,719 respondents, found that nearly nine in ten organizations use AI regularly in at least one business function, yet only 37% of respondents attribute any earnings before interest and taxes [EBIT] impact to AI, and the "high performers" who see significant value make up just 6% of the sample.

What separates those high performers is mostly managerial. Nearly three-quarters of them have fundamentally redesigned workflows because of AI, and they are twice as likely as other respondents to say their senior leaders show commitment to AI initiatives.

IBM's 2025 CEO study points the same way. Of the AI initiatives the surveyed chief executive officers [CEOs] reported, 25% delivered the expected return and 16% scaled across the enterprise. 64% of the CEOs said the risk of falling behind drives some of their technology investments before they have a clear sense of the value.

The most useful single study for managers is the 2023 experiment with 758 Boston Consulting Group consultants, led by researchers from Harvard, Massachusetts Institute of Technology [MIT], Wharton and Warwick. On tasks suited to GPT-4, consultants using it completed 12.2% more tasks, worked 25.1% faster and produced work rated more than 40% higher in quality. On a task designed to fall outside the model's abilities, consultants using AI were 19 percentage points less likely to reach the correct answer than those working without it. The researchers called this the "jagged frontier." For managers, the practical problem is that tasks inside and outside that frontier can look alike, and the output does not show which kind it was.

How leadership changes when AI is part of the decision

Make ownership of AI-informed decisions explicit

When an analysis is partly AI-generated, accountability blurs easily. Every significant decision should have a named owner, a record of what AI contributed and a note of what a person verified. Some organizations add independent oversight at board level; Think Insights has made the case for independent AI assurance boards as one model.

Train leaders for judgment, separately from tool skills

Managers do not need to write code. The AI skills that drive business growth at leadership level are about judgment rather than any particular tool: knowing where models are strong and weak, how to question an output, where the underlying data comes from and which risks the company is taking on.3 That is a different curriculum from the tool training most employees receive, and it keeps its value as the tools change.

Several business schools now run short programs built around this kind of judgment, aimed at managers rather than technical staff. The main executive options differ widely in time and cost: Wharton's AI for Business runs four to six weeks online for $850, Harvard Business School [HBS] Online's AI for Leaders takes 16–25 hours for $1,949, and MIT Sloan's five-day Leading the AI-Driven Organization program in Cambridge costs $12,900, according to a comparison of AI corporate training programs.

Show the team how you use it

If the leadership team uses AI openly, shares its prompts and says when it overruled a model's recommendation, people learn calibrated trust. If leaders hide their use or treat outputs as verdicts, teams learn either to conceal their own use or to defer to the tool.

A worked example: deciding whether to enter a new market

The example below is illustrative. Consider a mid-sized business-to-business [B2B] software company deciding whether to expand into Germany.

In the traditional process, the strategy team spends six weeks on market sizing and competitor research, develops one recommended plan, and presents a single deck to the board.

With AI built into the process, the same decision runs in three weeks:

  1. AI drafts a market and competitor pack from public sources and internal sales data. Analysts check every figure against its source and correct the errors
  2. The team asks for every plausible entry route and narrows them to four: direct entry, a partner-led launch, acquiring a local competitor, or waiting a year
  3. Each option goes through a pre-mortem and a red-team review written from the viewpoint of the strongest local competitor
  4. The team names the three assumptions that matter most (data-residency requirements, the local competitor's response, the length of the sales cycle) and builds a scenario in which each one fails, with the early signals for each

The board now debates four options with their failure modes instead of one recommendation, and leaves with a list of signals to review each quarter. The CEO still makes and owns the call.

Common mistakes in using AI for management decisions

A common mistake is treating a polished AI summary as checked analysis. Another is showing the model the preferred answer first, which weakens any critique it gives.4

Other mistakes are organizational. Companies automate a flawed decision process instead of redesigning it, which is the reverse of what McKinsey's high performers do. They buy tools before deciding which decisions those tools should improve, the pattern behind IBM's 64% figure. And they leave middle managers out of the redesign, although that group makes most of the operational decisions AI will touch first.

Summary

Artificial intelligence [AI] changes management decisions mainly by making options, scenarios and critiques cheap to produce. Its most useful roles in strategy are widening the set of choices, stress-testing plans, arguing against a proposal and drafting the evidence for board-level decisions. Research on consultants and product teams shows clear gains on suitable tasks and a drop in accuracy on tasks just outside a model's abilities, so every AI-informed decision still needs a person who checks the evidence and owns the result. Companies that see financial value redesign how decisions are made and have senior leaders who back the change. A practical starting point is one recurring decision: add AI to a single step and compare the result with previous cycles.

References

    Citation

    Cite this article

    Sridharan, M. A. (2026, November 16). AI in Management Decisions. Think Insights. https://thinkinsights.net/community/ai-management-decisions (Accessed [[ACCESS_DATE]])

    Author
    Mithun A. Sridharan Thought Leader
    I'm Mithun A. Sridharan, Founder of this website - Think Insights - on Strategy, Management Consulting, Leadership, Digital Transformation, and Data Literacy. Follow me on social media or connect with me on LinkedIn for updates.