Transformation
Between the AI tool and the bottom line, one layer is missing
Semantic transformation means organising a company's names, processes and data so that people, language models and board reports understand them the same way. 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.
Licences are no longer the problem. It has become the work nobody has described
Language models produce ever more content. The board needs numbers. Between the two sits the layer we build: names, owners and definitions that make work with AI measurable.
The model has three elements, and the middle one is what the company pays for: AI generates → semantics organises → BI accounts. Without the layer in the middle, the growth in content does not turn into a number the board can ask about. We describe this path in more detail on the From AI to BI page.
The gap the research shows
According to McKinsey, The State of AI 2026, 8 in 10 employees see time savings from GenAI; only 4 in 10 companies see an impact on operating profit (EBIT). In the same study, among the 6% of respondents seeing real returns, almost 3 in 4 have fundamentally redesigned workflows.
Part of the difference is eaten up by corrections. A study by BetterUp Labs and the Stanford Social Media Lab from September 2025 found that 4 in 10 office workers received AI-generated content in the past month that they had to fix or reject.
The problem also concerns data. Gartner predicted (February 2025) that by the end of 2026 companies will abandon 60% of AI projects not supported by AI-ready data, and that 63% of organisations either do not have, or do not know whether they have, the right data management practices.
5 reasons why licences do not turn into results
- Licences without use. The company pays for dozens of seats; a month later most employees have forgotten about the tool.
- A tool instead of a process. Employees receive an assistant, but no description of what to use it for.
- More content, less meaning. Communication speeds up while its value falls; work comes back to the sender as corrections.
- Knowledge in 2–3 heads. Procedures, exceptions and customer history exist, but not in a form a new employee or a model can understand.
- No measurement. Success is measured by course ratings, not by time recovered in a specific process.
Each of these reasons comes down to one thing: between the business goal and the tool, the company’s work has not been described.
Two layers, one method
The translation layer describes processes, knowledge, rules, data and decisions in a form that people and models understand the same way. In most companies it lives in the heads of two or three people. It is maintained by process translators: employees of the departments, not IT specialists.
The semantic layer is the company’s shared vocabulary: the same names across departments and tools, data owners, definitions of measures. Language models and BI systems draw on it at the same time.
A company that has both layers can choose any model and any tool, because the process description does not depend on the vendor.
One semantics, two directions, one report
Inwards: models help employees do the company’s work, because processes are described and knowledge sits in the system, not in people’s heads.
Outwards: models describe and recommend the brand correctly to customers, dealers and purchasing agents. We measure this through brand research in language models.
Both directions draw on the same vocabulary and both feed a single AI value report for the board. A company that organises one gets the other at lower cost.
The 4 levels: in what sequence
| level | what happens | who decides (budget) | what stays in the company |
|---|---|---|---|
| 1. competence | people understand how GenAI works and what the rules on data and security are | HR, competence development | personal certificates, library of templates, usage measurement after 30 days |
| 2. naming | workshops in departments: processes named and described, process translators selected | department directors and the decision-maker on the board | process cards, departmental vocabularies, GenAI communication standard |
| 3. linking | the company’s shared vocabulary, data owners, definitions of measures; the board decides who is accountable for results where departments meet | board, CIO, COO | semantic layer, data flow map, brief for IT and the BI partner |
| 4. value measurement | quarterly report to the board: value, cost, quality, compliance, external brand | CFO, board, supervisory board | report with interpretation and recommendations; dashboard on the client’s side |
Training is no longer the goal. It has become the entry to the first level. Process automation is an application on the third level: first the process is named and linked, then comes the agent.
The entry point to the ladder is the 30-day readiness assessment: it shows which level the company is on and where to start.
6 areas: what we do
| area | what the company receives |
|---|---|
| readiness assessment | readiness scorecard in three dimensions (people, knowledge, processes), a prioritised list of processes, the specification of one process for a pilot |
| competence | workshops based on the company’s own tasks, measurement before and after, documented measures to support staff AI literacy |
| process translation | processes described in a form that people and models understand; a single source of knowledge that withstands staff turnover |
| automation after mapping | a pilot of one process with measurement before and after; model and tool chosen for the task |
| GenAI communication standard | a library of templates for departments, rules for checking answers before they are sent, a list of data that must not be processed outside the approved environment |
| measurement | measures with owners, a board report based on data from the company’s systems |
The levels show the sequence in which we work. The areas show the full range of what we can do on each of them.
What we do and what stays with the company
| role | who | accountable for |
|---|---|---|
| decision-maker | member of the company’s board | goal, priorities, the decision to scale up and how recovered time is used |
| process owner | department director or manager | how the process runs, input data, acceptance of the result |
| measure owner | department director | definition and quality of the measure in the report |
| architect and moderator | AI Business Center Foundation | method, assessment, process description, choice of model, brief for IT, measurement |
| technical and BI contractor | the company’s IT or an appointed partner | integration, permissions, data model, dashboard |
| process translators | department employees | maintaining the descriptions and the standard |
We do not implement BI tools. We define what to measure and how. The Foundation is vendor-neutral: we do not sell licences or implementations. We act on the board’s side — we help choose the contractor, prepare requirements and criteria, run the procurement procedure or tender, and assess bids and contractor quality. The board makes the decision.
5 risks the project has to avoid
| risk | how we avoid it |
|---|---|
| more content, less meaning | communication standard and checks before sending; we measure corrections, not the number of pages |
| resistance: “faster means more work” | a narrative of recovered time, line managers taking part in workshops, HR as owner of the change |
| data and permissions | review of access rights before an agent goes live, the principle of least privilege, sensitive decisions always with a person; more on this: data sovereignty and private AI instance |
| dependence on a single vendor | process description not tied to any tool, model chosen for the task |
| regulatory compliance | documented support for the development of staff AI literacy (Art. 4 AI Act) and transparency towards customers (Art. 50); scope agreed with the company’s legal adviser |
What the rules require in communication and sales, and what they do not, is set out on the AI Act in communication and sales page.
Who this work is for
We work with companies that have bought licences for GenAI tools and cannot say what they get from them; that have several departments or sites which do not speak a common language; that have someone on the board ready to take responsibility for the change; and that want to build the competence in-house.
This is not work for companies looking for a turnkey implementation without involving their own people, a one-off tool training session, or a declaration of compliance instead of a real programme.
Templates, criteria and scales are the methodological property of the programme’s authors. Completed documents belong to the company and remain with it after the engagement ends.
Frequently asked questions
Why do GenAI roll-outs not show up in financial results?
Because the tool lands in an unchanged process. Employees write faster, but nobody has described what they should use the model for, who is accountable for the result and how to measure it. Minutes saved at the desk do not turn into money on their own. A described process with an owner and a measure does that.
How does semantic transformation differ from tool training?
Training teaches people to operate a tool. Semantic transformation changes how the company's work is described and leaves people in the organisation who maintain that description. The effect of training is measured by course ratings; the effect of transformation by process time, the number of corrections and throughput, using data from the company's own systems.
Does the Foundation implement BI tools?
No. We define what to measure and how, build the shared vocabulary and the owners of measures, and prepare a brief for the IT department. The dashboard and the data model are built by the company's IT department or a partner it appoints.
Which level should a company start from?
From the one it is actually on. The 30-day readiness assessment shows this: the scorecard covers three dimensions, people, knowledge and processes, which map onto the first three levels. A company with a well-trained team can start by naming its processes.
Let’s start with an assessment
30 days, a readiness scorecard and one board decision: implement, organise first, or stop. A good starting point is one process in sales, customer service or proposal preparation. Let’s talk about the assessment.