How we help

What language models say about your company when a customer asks them for a supplier

The AI Business Center Foundation (AIBC) is a think tank and advisory partner that helps the boards of construction, industrial and technical-services companies adopt artificial intelligence so that it delivers measurable results and builds competitive advantage. Brand research in language models measures how models describe and recommend a company.

A search engine is no longer the only place where customers meet a brand. Model answers have become that place too

A procurement department looking for a supplier of facade systems, HVAC installations or a contractor for an industrial hall increasingly starts with a question put to a language model. The model replies with a list of companies and a short description of each. A company that is missing from the list, or described wrongly, may lose before the sales conversation begins.

The problem we solve

Most companies do not know what language models say about their brand, products and people. Nor do they know which sources those answers come from, or why the model recommends a competitor. Without measurement, none of this can be managed.

What we research

Visibility at each decision stage: whether the company appears when the customer is only looking for a solution, when comparing suppliers and when checking a specific brand.
Who is recommended instead of the company, and why: which competitors appear in the answers and on the basis of which sources.
False statements: incorrect information about the company, its owners and its products, and their likely sources.
Missing documents: what models look for but do not find in the company’s content: parameters, technical data sheets, references, answers to common questions.
Other languages: the same protocol run in the languages of the markets where the company sells or where its customers come from.
Category and sector: how models understand the category in which the company competes, and whether they assign it to the right sector.
Comparison with competitors: the same measure for several companies in the category, calculated with the same protocol.
Change over time: how results shift between measurements, after changes to the company’s content and data.

We carry out the research with the Semantio tool and interpret the results together with the company’s team. Our recommendations concern content and data that the company itself controls: its website, product descriptions, documentation and expert publications.

An example from our research

In a Foundation study for a manufacturer of building-services products (2026), we put 26 question scenarios to five AI systems in two categories, AI search and generative assistants: 130 probes in total. In 30 answers to questions that did not name the brand, the company did not appear once. In 100 answers to questions that named the brand, it appeared in 92. 14 of the 130 answers contained false statements about the company, its people or products; after verification we classed 9 of them as material.

Study by the AI Business Center Foundation, 2026, anonymised data. This is not a market ranking or market share.

The problem is not the company’s reputation but its association with the category: models know the brand when the customer names it, but do not suggest it when the customer asks for a solution.

Applications in the real economy

Specifications co-written by models. Designers, general contractors and contracting authorities increasingly start a technical description from an AI assistant’s draft. We measure what models write into such drafts: which products they name as the reference product and how often, which parameters they treat as the default, and where errors in standards and classes or restrictive provisions appear. We organise the results into an error register and help correct the errors at source. We do not promise a place in a specification; we describe the phenomenon in the analysis Specifications drafted by language models.
Partner network and local visibility. Whether models point to the company’s authorised dealers, installers and service providers when a customer asks for a solution in their region.
Profiles of counterparties. How models describe the companies the company works or competes with; we research companies, not individuals.
Versions in other languages. How the brand appears in answers in the languages of foreign investors’ and general contractors’ teams.
Recurring monitor. The same measurement repeated at a fixed interval, to see the effect of changes to the company’s content and data.

Why scale and repeatability matter

A model’s answer is not fixed: the same question asked twice can produce two different lists of companies. That is why a single question put to a single assistant proves nothing. We put the same prompt to many assistants, several times, record the full text of every answer and apply the same protocol in subsequent measurements. Only then is a difference between measurements a change rather than chance.

Why this is part of the same method

The company’s vocabulary decides two things: whether a model can correctly recommend the company to a customer, and whether it can carry out the company’s work. That is why brand research is the outward-facing direction of semantic transformation, that is, organising a company’s names, processes and data so that people, language models and board reports understand them the same way. The research result is the external measure in the AI value report for the board, in the “external brand” group.

What stays with the company

a research report with the models’ answers and their interpretation,
a list of errors in how the company is described, with recommended corrections,
a baseline for subsequent measurements,
an “external brand” measure for the management report.

For Foundation Partners

Foundation Partners receive access to the Semantio tool on preferential terms and earlier access to new features through the early access programme for Partners.

Frequently asked questions

Why would an industrial or construction company need brand research in language models?

Because buyers, investors, designers and procurement departments increasingly begin their search for a supplier with a question put to a language model. If the model describes the company wrongly or does not know it, the company loses its place on the shortlist before anyone visits its website.

Which tool is used for the research?

The Semantio research tool, which puts sets of questions matching real customer queries to language models and analyses the answers: brand presence, how the brand is described, recommendations, competitors and errors. The Foundation's team interprets the research.

Is research into model-generated specifications meant to help the company win tenders?

No. We study what language models write into draft specifications, and we help correct errors concerning the company's products and parameters. The content of a specification belongs to its author, and the choice of bid to the contracting authority. We describe the same phenomenon for contracting authorities and designers who want to avoid incorrect provisions.

On what terms do Foundation Partners use the tool?

Foundation Partners receive access to the Semantio tool on preferential terms and earlier access to new features through the early access programme for Partners. Other companies use the research as a separate service.

Let’s start with an assessment

The first step is research based on the set of questions the company’s customers actually ask. Let’s talk about the research.