AI Does Not Produce Intelligence. Skilled People Do

Yaakov Chaliotis:

AI can accelerate monitoring, research and reputation analysis. But in strategic communication, speed is not intelligence, and fluency is not truth.

For communication professionals, the attraction of AI-powered intelligence and research platforms is obvious. The modern communication function is expected to monitor narratives, interpret stakeholder sentiment, advise leaders, prepare crisis responses, understand policy shifts, track media coverage and explain reputational risk at a pace that would have been unrealistic only a few years ago. A board wants to know what is happening. A chief executive wants to know whether an issue will escalate. A public affairs team needs to understand who is shaping a debate. An agency has to move from noise to counsel before the next client call.

It is therefore no surprise that the market is becoming crowded. Every week, a new product promises faster research, deeper analysis, cleaner visualisation and better decision support. Many of these tools are genuinely impressive. They can scan large document sets, generate structured outputs, identify patterns, compare narratives and produce polished briefings in a fraction of the time traditional research would require.

But beneath the impressive interfaces and confident demonstrations, two essential questions are still not being asked seriously enough. First, how skilled is the person using the system? Second, how verified, client-ready and resistant to hallucination is the final analysis?

These questions matter because access to a powerful AI platform is not the same as possessing intelligence capability. This distinction is particularly important for communication leaders because their work sits at the intersection of information, interpretation and consequence. Communication is not simply the distribution of messages. At its best, it helps organisations understand risk, make sense of external pressure, align stakeholders and act with credibility. That work cannot be reduced to the production of attractive outputs.

A sophisticated system in the hands of an unskilled user can still produce a sophisticated-looking mistake. AI can summarise documents, compare datasets and generate plausible explanations very quickly, but it cannot guarantee that the original question was the right one. If the user frames the problem poorly, embeds an untested assumption, ignores a sensitive stakeholder group or overlooks a critical source of evidence, the system may still return an articulate and persuasive answer. It may look complete, sound authoritative and be formatted beautifully. That is precisely why the risk is so serious.

For communication professionals, the danger is not only hallucination in the narrow sense. The greater danger is false confidence. An obvious error can usually be challenged. A polished answer that sounds credible is far more likely to pass unexamined into a presentation, executive note, holding statement, investor briefing, employee message, campaign plan, public affairs strategy or crisis response. Once that happens, the weakness is no longer technical. It becomes organisational.

The reputational implications can be significant. A hallucinated quotation can become a public correction. A weakly sourced statistic can undermine a campaign. A misunderstood regulatory development can lead to poor advice to the leadership team. A sentiment analysis that overweights a noisy but unrepresentative online community can distort priorities. In a crisis, a plausible but unverified AI output may feel like clarity. It may in fact be risk in a well-designed format.

This is where many AI intelligence products remain weak. They focus heavily on speed, design and output generation, but far less on whether the user knows how to interrogate the problem properly in the first place. The platform can accelerate the work, but it cannot supply judgement where judgement is missing. It can surface material, but it cannot decide which evidence matters most for the decision being made.

Prompting is often presented as a technical trick, as if the value lies in finding the right sequence of words to make the machine perform better. That definition is too narrow. In serious communication, research and intelligence work, prompting is an advisory discipline. It requires the user to break the problem into meaningful parts, challenge assumptions, ask for competing explanations, identify missing evidence and test the strength of the conclusion.

A skilled communication professional does not simply ask an AI system: “What should we say?” They ask: what decision are we trying to support? Which audiences will interpret this differently? What evidence supports the current view? Which sources disagree? What would make our assessment wrong? Which claims are factual, which are inferential and which are judgement calls? What is known, what is uncertain and what should not yet be said publicly?

That is not merely better prompting. It is better counsel.

The user must also understand source quality. AI may retrieve, summarise or generate information from a wide range of materials, but availability is not the same as reliability. A claim repeated across multiple websites may still originate from one weak source. A recent article may be less reliable than an older primary document. A confident social media thread may reflect a small activist community rather than a wider public mood. A dataset may look precise while resting on definitions that do not match the question being asked.

AI can accelerate the search, but it cannot remove the obligation to judge what has been found. The professional burden does not disappear. It moves upstream into question design and downstream into verification.

The second question is just as important: what exactly is being delivered to the client, the board or the leadership team? A raw AI output is not an intelligence product. Nor is it automatically a communication strategy. Senior leaders do not need a stream of machine-generated possibilities. They need analysis that has been checked, structured and prepared for decision-making. They need to know what is solid, what is uncertain, what is contested and what should be treated with caution.

That requires human review, source validation, contradiction testing, editorial discipline and strategic judgement. Claims must be traced. Dates, names, figures and quotations must be checked. Uncertainty must be communicated clearly. The final product must distinguish between fact, inference and judgement. Where there are gaps in the evidence, those gaps should be acknowledged rather than hidden behind confident language.

This is the gap we have built the methodology of Group of Verified Intelligence around. Our process begins not with the platform, but with the question. We design the prompt architecture and research structure first: defining the decision context, identifying the evidence required, exposing assumptions and building in competing lines of inquiry. AI is then used to accelerate discovery, comparison, analysis and scenario testing, but it is never treated as the final authority.

As part of the initial verification layer, we also use customised, semi-proprietary agentic AI verification tools to help test claims, surface contradictions and strengthen the evidence base. But the final decision on whether the intelligence is credible, client-ready and suitable for use is made by people. Before any work becomes client-ready, it passes through human verification, source validation, counter-hallucination checks, strategic interpretation and final editorial refinement.

Communication leaders should care about this now because AI is not only changing how communication work is produced. It is changing how organisations interpret the world around them. Media summaries, stakeholder maps, social listening reports, policy briefings, internal message testing and crisis scenarios will increasingly be AI-assisted. That is inevitable. The question is whether they will also be evidence-led.

The organisations that benefit most will not simply be those that buy the most impressive platform. They will be those that develop skilled users, rigorous methodologies and a culture in which fluency is never mistaken for truth. This means setting standards before the work begins. What decision is this analysis supporting? Which claims require primary evidence? Who is responsible for verification? How should uncertainty be expressed? Who has challenged the conclusion before it reaches the boardroom?

These questions should become part of the professional discipline of modern communication. They are as important as tone, timing, channel strategy or message architecture. A message built on weak intelligence is not a strong message. A campaign based on untested assumptions is not strategic. A crisis response that depends on unverified analysis is not agile. It is exposed.

Robust AI products are becoming indispensable tools for serious research, communication planning and decision support. Used well, they can help communication professionals move faster, see patterns earlier, test narratives more intelligently and advise leaders with greater confidence. But the real value comes from the combination of technology, expertise, verification and judgement.

For a profession that has long argued for a stronger role in strategic decision-making, this is also an opportunity. Communication professionals can become the people who understand not only how messages move, but how claims are tested; not only how narratives form, but how evidence should be weighed; not only how stakeholders react, but how leaders should interpret uncertainty before acting.

The future of intelligence will not belong simply to those with the best machines. It will belong to those who know how to ask better questions, who understand the difference between output and insight, and who refuse to trust an answer until it has been verified. For communication leaders, that may become one of the defining professional skills of the AI era.

+++

Yaakov Chaliotis is the Founder and Managing Director of Group of Verified Intelligence, an AI-assisted strategic intelligence and research consultancy. He advises organisations on decision intelligence, verified research, reputation risk and the responsible use of AI in analysis. He previously held senior communications and analytics roles at Shell International and the UK Department for Education, where he worked across digital communications, insight and strategic decision support.