AI in business strategy: avoiding similar answers

Category: Decision-Making | AI & Strategy

Reading time: 7 minutes

AI in business strategy entered organisations with a clear promise: to accelerate analysis, generate alternatives and support complex decisions.

But there is a less frequently discussed risk. When different organisations use similar tools, ask similar questions and lack a deliberate process of constructive challenge, they may end up considering answers that are more alike than they realise.

The problem is not that artificial intelligence produces a wrong answer. It is that it produces an answer that is plausible, fast and sufficiently good to prevent the organisation from looking for less obvious alternatives.

In this article, I use the term algorithmic isomorphism to describe this risk: the possibility that standardised use of AI may narrow the range of strategic hypotheses considered by different organisations.

When AI accelerates the answer but does not improve the question

There is an important difference between producing faster and thinking better.

An organisation can use AI to summarise markets, design value propositions, identify segments, build growth plans or generate strategic scenarios. All of this can be useful. The problem emerges when the request is too predictable and the result is accepted as an end point rather than used as a starting point.

In practical terms, asking several tools for a “growth strategy”, a “differentiated value proposition” or a “talent retention plan” without challenging the underlying assumptions tends to produce recognisable solutions: improve processes, invest in leadership, understand the customer, strengthen communication, automate tasks and create indicators.

None of this is necessarily wrong. But nor does it constitute, by itself, a distinctive strategy.

What is algorithmic isomorphism?

The concept of isomorphism describes the tendency for organisations to become more similar over time. In institutional theory, this convergence results from regulatory, professional and imitative pressures: organisations observe what appears legitimate, adopt recognised practices and move closer to the dominant forms in their sector.

Generative AI may introduce a complementary mechanism. Organisations that do not observe one another may still receive increasingly similar answers when they use similar tools with similar requests.

This does not mean that all strategies become identical. It means that the way an organisation questions AI influences the variety of hypotheses, actions and perspectives that reach the strategic discussion.

Algorithmic isomorphism is therefore a lens for examining a contemporary risk: convergence may no longer depend solely on imitation between companies, but also on how AI systems generate probable answers.

What an exploratory study found

As part of my PhD in Management at the University of Évora, I conducted an exploratory study on the convergence of AI-generated strategic content.

Forty strategic proposals were produced for the same fictitious company, using two AI models and two prompting conditions.

• Standard search: equivalent requests made in independent sessions.
• Differentiated search: an iterative protocol with explicit instructions to explore opposing perspectives, less obvious alternatives and distinct strategic paths.

The indicator used was lexical similarity between documents, measured through the Jaccard coefficient applied to the relevant vocabulary in each proposal.

Under the standard condition, average similarity was 0.307. Under the differentiated search condition, it fell to 0.196. This represented a relative reduction of 36.3% in lexical overlap between outputs.

In an exploratory replication using prompts in English, the pattern remained in the same direction. Across the two models tested, no statistically significant difference was detected in the effect of the prompting condition.

Methodological note. This study assesses lexical overlap between strategic outputs produced in a controlled scenario. It does not directly measure strategic quality, commercial performance, competitive advantage or semantic equivalence between proposals. The results should be interpreted as exploratory evidence concerning content variety, not as proof that different companies will inevitably adopt the same strategy.

Why does this matter for business strategy?

Because an AI-generated answer can be clear, well written and apparently sophisticated without being genuinely distinctive.

When the same tool is used for market analysis, positioning, service design or growth plans with questions that are too similar, the organisation may reduce its search space without realising it.

The risk does not lie in using AI. It lies in confusing production efficiency with diversity of thought.

Research on the jagged technological frontier reinforces this point: AI can significantly improve performance in certain tasks, but it does not produce uniform value across every type of problem. The management question is not merely whether to use AI. It is to understand where AI adds quality, where it requires validation and where it may limit reflection.

This is also the question explored in Artificial intelligence in decision-making: greater clarity or greater error?: the tool can support a decision, but it does not replace a clear decision architecture.

Three practices for using AI without narrowing the options

1. Diversify the question before requesting the answer

Do not simply request “a strategy”. Request alternatives that contradict the initial hypothesis, explore different audiences, test assumptions or begin from opposing business logics.

For example, instead of asking “create a differentiated value proposition”, try:

• Which assumptions behind this proposal may be wrong?
• Which alternative would a competitor with half the resources adopt?
• Which solution would make sense if this market stopped valuing price?
• Which perspective from a seemingly distant sector might challenge this approach?

2. Separate the generation of possibilities from strategic judgement

AI can generate hypotheses, organise information and create scenarios. But priorities, risk, context, feasibility and implementation continue to require human judgement.

Using AI as an interlocutor is different from using it as the final arbiter. In relevant decisions, someone should have explicit responsibility for questioning what is missing, what may be biased and what has been assumed without sufficient evidence.

3. Introduce constructive challenge before closing the decision

A team should not accept the first plausible answer as the best available answer.

Before turning an AI recommendation into a decision, it is worth creating a moment of constructive challenge: someone should search for the scenario in which the recommendation fails, the variable the model did not consider and the perspective that is absent.

This is where leadership shifts from rapid validation to the ability to improve the quality of collective thought. As we explain in Why training managers is not the same as developing leadership, developing leaders is not only about providing tools. It is about creating the conditions for better decisions in ambiguous contexts.

Four questions before using AI in a strategic decision

• Which specific decision are we trying to support?
• Which assumptions did we include in the question before requesting an answer?
• Which relevant alternative is not being considered?
• Who will challenge the output before the decision is made?

These questions do not make the process slower as a matter of principle. They make it more deliberate. They may also prevent a statistically probable answer from being mistaken for the best decision in the organisation’s specific context.

The advantage of AI in business strategy does not lie solely in the tool

As AI becomes more accessible, technology alone ceases to be a sufficient source of differentiation.

The difference will lie in the quality of the questions, the perspectives included in the process, the decision criteria and the ability to recognise what the model still cannot see.

At VAGAR, this work does not begin with choosing a tool. It begins by understanding how the organisation makes decisions, where information is lost, which criteria are absent and how leadership, structure and processes can use AI without delegating judgement.

Explore VAGAR’s solutions for organisations or book an initial conversation to examine your context.

Book an initial conversation

Evidence base

Lebovitz, S., Lifshitz-Assaf, H., & Levina, N. (2022). To engage or not to engage with AI for critical judgments: How professionals deal with opacity when using AI for medical diagnosis. Organization Science, 33(1), 126–148.

Pires da Silva, L. (2026). Algorithmic Isomorphism: AI-Generated Strategic Content Convergence and the Mitigating Role of Differentiated Search. Working paper, PhD in Management, University of Évora.

DiMaggio, P. J., & Powell, W. W. (1983). The iron cage revisited: Institutional isomorphism and collective rationality in organizational fields. American Sociological Review, 48(2), 147–160.

Dell’Acqua, F., McFowland III, E., Mollick, E. R., Lifshitz-Assaf, H., Kellogg, K., Rajendran, S., Krayer, L., Candelon, F., & Lakhani, K. R. (2026). Navigating the jagged technological frontier: Field experimental evidence of the effects of artificial intelligence on knowledge worker productivity and quality. Organization Science.