AI
consulting
AI consulting
Deciding which of your company's problems AI can solve today, in what order to tackle them and what data you need, before spending a single euro on development.
AI consulting starts with the business, not with the technology. We look at how your company works. Where the hours go on repetitive tasks, which decisions are made on incomplete data, which questions nobody can answer with confidence. It is an AI readiness assessment of the business as much as of the data. Out of it comes an inventory of use cases, each with its estimated impact, its effort and the data it needs.
The difference from a talk on trends or a course on generative AI is that you leave with a roadmap you can execute. The first project defined, the budget bounded and the criteria for measuring whether it worked. It is not training, and it is not a ‘transformation’ programme. It is a plan to put AI to work on real tasks in your company, coordinated by a team that has been putting AI projects into production since 2020.
What we are asked in the first meeting, answered in two lines.
The first step is deciding what is worth building. A development project solves the problem you hand it; consulting checks that it is the right problem and that you have the data to solve it.
Yes, if you have processes with volume: hundreds of emails, orders, invoices or support tickets a month. Below that, AI rarely repays the effort, and we will tell you so in the first session.
The roadmap takes four to six weeks. The first use case up and running, depending on which one is chosen, takes two to four months after that. We start with the one that hurts most, not the flashiest.
Six deliverables that come out of the consulting work. Each one answers a question nobody can answer with confidence today, and together they make up the roadmap.
Everything AI could do in your company, ranked by economic impact and effort. Out of twenty ideas come three worth starting now and ten that don't pay off.
What gets built, when, and what has to be ready beforehand. A calendar that leadership can approve, budget for and track.
What data exists, in which system, at what quality and who controls it. It decides whether a use case is viable in three months or needs a data platform first.
Data platforms
Cost to build, cost to run (licences, model usage, maintenance) and expected savings, with the assumption written down so it can be checked later.
Closed scope, success metric, architecture and chosen model. If it is an AI agent, it leaves here ready to be built.
AI agents
An internal usage policy, who approves what and how the decisions AI makes are recorded. Plus what the EU AI Act requires in your case. Just enough not to slow you down.
Five steps over four to six weeks. The first one isn't about AI.
Interviews with leadership and with the people who do the work every day. We look for repetitive tasks with volume, decisions made blind and knowledge that lives in two people's heads.
Systems, documents, emails, spreadsheets. What exists, at what quality and who controls it. Without this step, any roadmap is fiction.
Not every step you need to take depends on AI alone. We review the technology you already have, propose the improvements it needs and represent you in front of your suppliers.
Each use case is scored on impact, effort, available data and risk. Three candidates come out, not thirty, and the order is argued with numbers.
The roadmap goes to leadership, with the business case for each project. The outcome can be ‘we start with this’ and it can also be ‘not yet’: if AI doesn't pay off, we say so.
Consulting decides; the other products build. This is where the roadmap comes from, with the use cases prioritised and split between those our products can cover and those your technology suppliers should handle.
If the winning use case is a task a person does step by step today, we build an AI agent. If it is answering questions about internal documents, an AI knowledge base (RAG). If it is a complete process that crosses several systems, AI workflow automation. If the diagnosis shows the data isn't ready, a data platform or a Business Intelligence dashboard comes first. And if you need to predict demand, bad debt or breakdowns, Machine Learning.
That is why consulting is not a report that gets filed away. It is the step that stops you building what you don't need, and it makes sure every AI implementation has data, an owner and a metric. Typically a roadmap produces two or three projects of three to six months each, and the success of the first one makes the next ones easier.
Scattered ideas
Roadmap
Supplier invoices that record themselves, bank reconciliation and replies to claims. This is where repetitive hours pile up most in a mid-sized company, and where the return is easiest to measure.
An assistant that answers with your company's information and escalates what it doesn't know. It also shows which questions keep coming back and which documentation is missing.
A proposal put together in minutes from the price history, the customer's terms and the current catalogue. And an early signal of which customers are buying less, before they leave.
Demand forecasting, stock calculation and shift or route planning with real data. This is work we have already delivered.
Procedures, regulations, technical sheets and contracts available with a single question. It is the first thing companies ask for when they depend on two or three people who ‘know everything’.
The ones that come up on the first call. If yours is missing, write to us and we will add it.
Let's talk about your projectIt depends on how many areas and systems need analysing and how many people need interviewing. We give a fixed price after a first introductory session: that session is enough to see the scope.
No. Diagnosing its state is part of the work. If it isn't ready, the roadmap puts a data platform first: we don't build an agent on data nobody trusts.
No. There are sessions with the team to explain what will be built and why, but the deliverable is a plan of projects, not a course.
The roadmap, the business cases and the architecture are yours, and they are documented. You can build with us, with another supplier or with your own team.
OpenAI, Anthropic, Google and open models, depending on the case: cost, confidentiality, language and answer quality weigh differently in each project. We don't choose a provider before we understand the problem.
Almost all the use cases in a mid-sized company are minimal or limited risk, and they only require transparency. Those that affect people carry extra obligations; we classify every one of them.
Seven industries have their own page: construction and engineering, manufacturing and food, travel and hospitality, yachting and marine, professional services, healthcare and logistics and supply chain. And we work with clients in many more. The diagnosis is the same in all of them; what changes is where you start, and that is set out on the page for each industry.
