AI consulting for companies: the diagnosis

AI consulting for companies starts with a diagnosis that maps how decisions are made in each process, measures the baseline, says what to automate first and what not to, and flags the legal risks. If it does not put all of that in writing, in a document the company keeps, it is not yet a serious diagnosis.

This article is part of AI-First hybridisation consultancy.

In short

The diagnosis comes before any tool and before any quote.
A serious diagnosis writes down the decision map, the baseline, the order of the processes, the legal risks and the owners.
It also says what not to automate, and why.
The scope sets the price: first the diagnosis, then a written proposal.
With Legathum, the first result in production arrives within 2 days of implementation starting.

What does AI consulting for companies actually do?

AI consulting for companies helps a business decide where artificial intelligence makes sense in its processes, build it with the company's own rules and check afterwards that it has worked. What sets it apart from selling a tool is the order: first you understand how the work is done and measure it, and only then do you automate.

The underlying goal is company hybridisation with AI: sharing out each process between AI agents and people, with written rules and a person at the critical points. When that split reaches every area, the result is an AI-first company.

The work usually has three phases: diagnosis, implementation and ongoing support. All three are set out in how to implement AI in a company in phases. This article focuses on the first, because everything else depends on it, including the budget.

Why does the diagnosis come first?

Because without a diagnosis nobody knows what to automate, how long the process takes today or who is answerable for it. A proposal that arrives before the diagnosis can only offer what it brought with it.

The diagnosis also prevents two expensive mistakes. The first is automating a process nobody can explain: the agent ends up working from whatever sounds plausible rather than from the company's criteria. The second is automating a process that the EU AI Act treats as high-risk without first designing the human oversight.

And it is the basis of the budget. Until you know which processes and which areas are in the project, any figure is a guess.

What should a serious diagnosis include?

A serious diagnosis puts these pieces in writing. If one is missing, ask why:

What an AI diagnosis should include, and what happens if it is missing
Piece	What it is	What happens if it is missing
Decision map	In each process, who decides, with what information and by which rules	The agent works from what sounds plausible, not from the company's criteria
Baseline	How long the process takes today, how much manual work it needs and which errors it produces	Nobody can tell afterwards whether the implementation has worked
Order of the processes	Which goes first, which next, and why	The most eye-catching process gets automated, not the one that makes most sense
What is not automated	The processes that stay with a person in charge, with the reason	It comes to light too late, when an agent is already involved in decisions about people
Legal risks	Which processes touch personal data, decisions about people or conversations with clients	Human oversight is improvised at the end
Owners	A named person for each process that goes in	Nobody reviews what the agent does or corrects its rules
Next step	A written proposal with the scope, the phases and the price	The company does not know what it is buying or what it will receive

A diagnosis that only brings a list of tools, or a report without a baseline, is a well-presented opinion. The test is simple: with the document in your hand, could you decide what to automate first even if you hired nobody else?

The piece most often skipped is the fourth. Saying what not to automate seems to take work away from whoever is selling the consulting, which is why it is a good sign when it appears.

What does the Legathum diagnosis look like?

In the Legathum AI-First hybridisation consultancy, the diagnosis is two working sessions with Deibson Silva and a written report on what to automate first and what not to. The report is yours, whether or not you hire the next phase.

It follows the order of the Legathum AI-First Hybridisation Method:

Map the decisions in each candidate process: who decides, with what information and by which rules.
Measure the baseline of the processes that pass the first filter.
Score each process and choose the first one, with its owner. How to score them is in which processes to automate with AI first.
Flag the legal risks of each process: personal data, decisions about people and conversations with clients.
Present the report, section by section, with the proposal for the implementation.

Then come the implementation, with six named deliverables, and the ongoing support, which measures against the baseline and adds one new process per cycle, over 3, 6 or 12 months. During implementation, the first result in production arrives within 2 days, measured against the baseline from the diagnosis.

No phase starts until the company has approved the previous one. If the diagnosis finds that a process is not ready to be automated, the report says so, and explains what is missing.

How much does AI consulting cost, and how is it quoted?

Legathum does not publish the price of its consulting, because it depends on the processes and areas in the project. It is discussed in a 15-minute meeting with Deibson Silva, and the written proposal, with the scope, the phases and what each one costs, comes with the results of the diagnosis.

These are the variables that drive the budget for AI consulting, here and anywhere else:

How many processes and how many areas are in the project.
Whether the data the agent needs sits in systems that can be accessed, or has to be put in order first.
How many rules have to be written, reviewed and approved in the Book of Criteria.
Whether any process is involved in decisions about people, which call for more oversight and more documentation.
How much support the team needs to work with the agents day to day.

So be wary of a fixed figure before anyone has looked at your processes, and equally wary of a diagnosis that does not end in a proposal with a scope. The two go together: the diagnosis says what needs doing, and the proposal says what doing it costs.

What should you ask before hiring an AI consultancy?

These seven questions separate a serious diagnosis from a sales presentation:

Does the diagnosis measure a baseline before proposing anything?
Will it also tell me what not to automate, and why?
Where are the rules the agents work by written down, and who approves them?
Who reviews what each agent does, at what point and with what authority to correct it?
How is the EU AI Act taken into account from day one?
Which documents stay with my company if I do not hire the next phase?
How will the result be measured, and against what?

A consultancy that answers these questions with documents, not promises, is showing you how it works. If an answer is "we'll look at that later", make a note of it: that is usually where the problems appear.

Which rules does the diagnosis have to check?

A diagnosis carried out in Spain has to review at least these rules:

GDPR, Article 35: if processing that uses new technologies is likely to result in a high risk to people's rights, the impact assessment is carried out before starting. The Spanish Data Protection Agency (AEPD) has a guide on processing that incorporates AI.
EU AI Act, Annex III, point 4: systems involved in hiring and selecting staff or in decisions about the employment relationship are high-risk. Following the Digital Omnibus on AI (Regulation (EU) 2026/1744), their obligations apply from 2 December 2027.
EU AI Act, Article 26: in those uses, human oversight is assigned to people with the necessary competence, training and authority.
EU AI Act, Article 4: anyone using AI systems takes measures to support the AI literacy of their staff. It has applied since 2 February 2025.
EU AI Act, Article 50: if an agent talks to clients, they have to know they are talking to an AI. It applies from 2 August 2026.
Spain's Workers' Statute, Article 64.4.d: the works council has the right to know the parameters, rules and instructions of the algorithms that affect working conditions.

None of these rules stops you starting. They ask to be considered from the diagnosis, while changing the design is still easy. Short answers on the EU AI Act and the GDPR are in the FAQ.

Frequently asked questions

What is AI consulting for companies?

It is the work of deciding where AI makes sense in a company's processes, building it with the company's own rules and measuring afterwards whether it has worked. It always starts with a diagnosis.

What should an AI diagnosis include?

A decision map, the baseline of each process, the order in which they are automated, what is not automated, the legal risks, an owner for each process and a written proposal with the scope.

How much does AI consulting cost for an SME?

Legathum does not publish the price of its consulting, because it depends on the processes and areas involved. It is discussed in a 15-minute meeting with Deibson Silva, and the written proposal comes with the diagnosis.

When is the first result visible?

With the Legathum method, the first result in production arrives within 2 days of implementation starting, measured against the baseline. The diagnosis, which chooses the process, comes first.

What do I keep if I do not hire the implementation?

The diagnosis report is yours either way: the decision map, the baseline and the order of the processes. With it you can decide the next step on your own.

AI consulting for companies in Spain: what a serious diagnosis should include