AI in consulting: collecting data is easier than interpreting it correctly

Can a company build a strategy solely on data collected with AI?

Starting with AI is possible and worthwhile. But data collected by a model, without verification and interpretation, rarely becomes a reliable foundation for a strategy.

Preliminary research with AI is useful: within an hour you get an overall picture of the market, a list of players and the main regulatory frameworks. For an initial understanding, this is enough even when you use standard, publicly available models.

Problems start when this picture is treated as the basis for decisions. A model’s answer is assembled from what has already been published and averaged out to a level where it is right for everyone and for no one in particular. The model does not know your business, does not see your economics and, most importantly, does not carry your risks. If a conclusion turns out to be wrong, it is the company that loses money, time and reputation, not the tool.

For AI to deliver a high-quality result, it has to be used properly: understanding the structure of the industry, framing the task in the right way, requiring a structured answer from the model and then checking that answer. Without understanding what the information is made of and how a practical result is achieved, relying only on data collected by a model is risky.

What has AI really changed in consultants’ work?

The biggest change is in working with data: it can now be structured and analyzed much faster and in greater depth.

Like many consultants, we have started working with more structured data. Our own data, or data a client provides for analysis, can now be processed in days rather than weeks: combining different sources, spotting patterns, testing hypotheses on a larger volume of information. This has significantly raised the level of what can be done within a single project.

This matches what researchers are recording. In a 2026 study by the University of Münster, consultants describe AI primarily as a way to automate the processing and analysis of information, which frees up time for client work and complex tasks. Most of them do not see AI as a threat to their core value, which rests on human interaction.

Why is AI’s main problem not the data, but its interpretation?

Because a model sees a number but does not see how it came about. Without that, the same number can mean opposite things.

Energy and engineering are the sectors we work with most often, and this is where the gap between data and understanding it is particularly visible. Working with data in these industries requires two things:

First: understanding what the data is

Who counts it, what it includes, by what rules it is produced and at which stage of the chain it arises: with the developer, the EPC contractor, the grid operator or the regulator.

Second: interpreting it correctly

Understanding what the number means for a specific business in a specific market. This is the stage where models go wrong more often than in collecting data.

Here is an example from the Polish renewable energy market, which we know well:

Source: report of the President of URE (Energy Regulatory Office) for 2025, data from Agencja Rynku Energii (ARE) and the Ministry of Energy, as published by Gramwzielone.pl on 11 May 2026.

Each of these numbers is easy to misread, and that is how models often read them:

None of the surface interpretations is made up. Each follows logically from the number. But each leads to the wrong decision: where to go, whom to sell to, how much to invest. The correct reading is much easier to find when you understand the structure of the industry and know which questions to ask next: about regulatory acts, the competitive environment and project economics in that specific country.

Why does data verification matter more than speed of analysis?

Because fast analysis of unverified data only leads to a mistake faster.

Verifying, checking and validating data today is not an optional step after working with AI. It is a condition without which the output of that work is hard to rely on in decisions. Every number a conclusion rests on should be traced back to its primary source, checked for currency and, ideally, confirmed by people who work in that market.

Even specialized tools do not remove this need. A preregistered Stanford study published in the Journal of Empirical Legal Studies showed that leading legal AI services gave incorrect or unsupported answers in 17-33% of cases, albeit less often than general-purpose chatbots. Energy regulation is no simpler than legal databases, and it changes more often. This is one of the main limits of AI in consulting today.

AI does not know your business and does not carry your risks. That is why responsibility for which data goes into a decision stays with people.

Dima V. Nechyporenko, founder of the nech

Why do people remain decisive in B2B?

Because B2B deals are made between people, and trust between them does not come from an email exchange.

Market research ends with a report, but that is where business only begins. Companies need to find each other, get acquainted and, most importantly, understand each other: how decisions are made, what matters to the partner, where the limits of compromise lie. That is very hard to achieve through emails or messages on LinkedIn or messaging apps.

In energy and engineering this is felt particularly strongly. Deal cycles are long, the sums are large and the risks for both sides are measured in years. In one country the key decision is made by the technical director, in another by a procurement committee, in a third the position of the project’s bank or investor will be decisive. A model can describe these differences in general terms. Someone who has worked there is much better placed to sense them in a specific conversation.

A great deal in business remains human: relationships, presence, an understanding of the mentality and cultural specifics of markets, real-life experience. These elements are absent from a model’s answer, yet they are often what decides whether a deal happens.

How is AI changing the consulting industry?

Quickly and profoundly. We see these changes as very positive.

Part of the work that once took up most of a consultant’s time, namely searching for, collecting and initially processing information, is now done much faster. This does not make consulting unnecessary; it changes where its value lies: in understanding the industry, correct interpretation, data verification and working with people.

For us, this means being able to do much more complex things in less time and to help clients achieve results in Europe faster. The tool has become more powerful. Responsibility for how it is used has stayed where it always was.