For thirty years, competitive intelligence in B2B software had a familiar shape. Once a year, an analyst firm published a two-by-two chart. Vendors briefed the analysts, supplied customer references, and argued about their dot. Buyers downloaded the chart, built a shortlist from the upper right, and issued RFPs. Everyone involved understood the rules, and the rules were stable.
That shape is changing, not because analyst firms have become less rigorous but because buyers have added a step before the analyst report, and often instead of it. They ask an AI assistant. The assistant produces a shortlist in seconds, with commentary, drawn from everything it has read. And the shortlist it produces is, in many categories, not the one the analysts published.
This article is about what that means for competitive intelligence: what analyst reports measure, what AI-generated market maps measure, where the two diverge, and why the divergence is the most useful thing a competitive intelligence team can track.
What an Analyst Report Actually Measures
An analyst quadrant is a considered expert judgment. It is built from vendor briefings, product demonstrations, reference customers, and a published set of evaluation criteria, and it is produced by people who spend their careers on a category. Its strengths are real: depth, access to information vendors do not publish, and a defined methodology.
Its limits are also real, and they are structural rather than a matter of quality. It is published annually or less. It covers the categories the firm has chosen to cover, typically a few hundred, out of the thousands buyers actually evaluate. It reflects the judgment of a small number of analysts. And it is influenced, legitimately, by the vendors that engage most actively with the process, which favours large firms with analyst-relations budgets.
AI-Powered Competitive Intelligence — The systematic measurement of how AI systems describe, rank, and recommend vendors in a category, across multiple models and over time. Unlike analyst research, it does not evaluate vendors directly; it evaluates the recommendation layer that buyers consult, which is itself a synthesis of everything published about the vendors.
What an AI Market Map Measures
When QuadrantX scores a category, it asks five production AI models the questions buyers ask, repeatedly, and measures two things about each vendor: Narrative Dominance, how prominently and consistently the vendor appears in the answers, and Sentiment, how favourably it is described. The result is a quadrant, but it is a quadrant of a different thing. It does not tell you which vendor an expert believes is best. It tells you which vendor a buyer will be told is best, by the tools buyers increasingly use first.
The distinction has three consequences.
Coverage. Analyst firms cover the categories large enough to justify an analyst. AI models will answer a question about any category a buyer can name. QuadrantX tracks Canadian credit unions, arena scoreboard manufacturers, and Oracle Eloqua implementation partners alongside CRM and ERP, because buyers ask AI about all of them and nobody else is measuring the answers.
Frequency. An analyst report is a snapshot. AI recommendations shift as models are updated and as the public narrative about vendors changes. We regenerate every category on a rolling schedule and can watch a leader's position change edition by edition. In our data, roughly half of categories have a leader so stable it appears in more than nine editions out of ten; the other half move, and the movement is invisible to anyone reading an annual report.
Perspective. An analyst report reflects the analyst. An AI market map reflects the accumulated public record, weighted by five different systems with five different training sets. It is not a better judgment. It is a measurement of a different, and increasingly decisive, thing.
The analyst tells you what the experts think. The AI market map tells you what your buyers are being told. When those two disagree, you have found the gap that is costing you deals.
Where They Diverge
In mature enterprise categories, analyst placement and AI recommendation often agree at the top. Salesforce leads CRM in every edition we have generated and for every model. SAP leads ERP. AWS leads cloud infrastructure. These vendors are dominant with analysts and dominant with AI because both are reading the same overwhelming public record.
Below the top, and outside the largest categories, they diverge in ways that matter.
AI rewards the documented, not the evaluated
A vendor with a strong product, a small marketing footprint, and a good analyst placement can be nearly absent from AI recommendations, because AI models learn from what is written and little has been written. Conversely, a vendor with a loud content presence can appear prominently in AI answers despite a middling analyst view. Neither AI nor the analyst is wrong. They are measuring different things, and the buyer is now consulting the one that measures documentation.
AI resolves category boundaries differently
Analysts define categories precisely and place vendors deliberately. AI models infer categories from language, and different models infer differently. In our crisis communications category, two models treat it as a public relations discipline and three as a software category, and they recommend accordingly. An analyst would never make that ambiguity; a buyer typing a question into an assistant encounters it without knowing.
AI has no reference calls
The things analysts learn from customers under confidentiality, about implementation pain, support quality, and roadmap risk, do not appear in AI answers unless customers have written about them publicly. Sentiment scores in AI recommendations track public reputation, which is a related but different quantity. A vendor with quiet, satisfied enterprise customers and no review-site presence will be under-described.
AI moves first
When a vendor's narrative shifts, whether through a product launch, an acquisition, or a run of press coverage, retrieval-based AI models reflect it within weeks and training-data models within months. The next analyst report may be a year away. Competitive intelligence teams that watch the AI layer see the shift before their competitors' analyst-relations teams do.
Using Both
The practical question is not which source to trust but how to use them together. Analyst research remains the deepest evaluation of product capability available, and analyst placement is itself one of the inputs AI models learn from. AI market maps add three things analyst research cannot: coverage of every category, measurement on a continuous schedule, and a direct reading of what buyers are told.
A competitive intelligence program built on both looks like this:
- Map the gap. For each category you compete in, compare your analyst placement with your AI quadrant position. Where AI under-represents you relative to the analysts, the problem is documentation and consistency. Where AI over-represents a competitor, study what they publish.
- Track the models separately. Five models produce five views. A competitor gaining ground on the retrieval-based model first is a leading indicator; a competitor strong on one training-data model and absent on the others is an artefact of that model's sources. Pooled scores hide both signals.
- Watch the categories analysts do not cover. Adjacent and emerging categories are where AI is often the only structured measurement available, and where the narrative is still being written.
- Read the sentiment, not just the placement. How AI describes a competitor, including the caveats it attaches, is the language your buyers are hearing. It is also the earliest public signal of a reputation problem.
Treat the analyst report as the expert view and the AI market map as the buyer's view. Most competitive surprises in the coming years will happen in the gap between them: the vendor the analysts rate highly that AI never mentions, and the vendor AI recommends that the analysts have not yet evaluated. A competitive intelligence function that measures both, continuously, will see those surprises coming.
The Bottom Line
Analyst reports are not going away, and they should not. What has changed is that they no longer describe the whole competitive landscape, because the landscape now includes a recommendation layer that buyers consult first and that behaves differently from expert judgment. Measuring that layer, across models and over time, is what AI-powered competitive intelligence means, and it is the part of the picture that a single model, or a single annual chart, cannot show you.