1. Assessment
We map one process as it actually runs, including the manual workarounds people have built around it. This is where most of the value is found, and it usually takes days rather than weeks.
AI Consulting · Lisbon, Portugal
AI Consulting for Companies in Portugal
We start with one process that is costing you time, establish whether AI is the right instrument for it, and build only if it is.
How an engagement runs
2. Data readiness
We check whether the information the process needs exists, where it lives and what state it is in. A model cannot compensate for data that was never recorded.
3. Scope
If there is a case worth building, we agree the outcome, the timeline and the cost before anything is built. If there isn't, you get the assessment and we stop there.
4. Build and hand over
We build against the systems you already run and hand over something your team can operate without us — code, configuration and documentation included.
Most AI projects fail before any model is chosen
The common failure is ordering: a company picks a technology and then looks for a problem it might solve. The project produces a demonstration that works in a meeting and is never used afterwards, because nobody established what it was supposed to replace.
We work in the opposite order. We take one process that costs the company time today, map how it actually runs — including the manual workarounds that never made it into any documentation — and only then decide whether artificial intelligence is the right instrument. Sometimes the answer is a rule, a form, or deleting a step nobody needed.
This is why our first conversation is about the process and not about the data. What happens, who does it, how long it takes, and what goes wrong. Data comes into it only once a specific case looks worth pursuing.
What makes a process a good candidate for AI
Three conditions matter more than any other. The process has to be repetitive enough that automating it pays back. The information it depends on has to exist somewhere usable. And a mistake has to be recoverable, or at least detectable, because no system of this kind is right every time.
A process that runs twice a year fails the first test. One that depends on knowledge living only in someone's head fails the second. One where an undetected error reaches a customer or an accounting entry fails the third. We check all three during the assessment, and we tell you when a case does not clear them.
AI at small-company scale is a different argument
A large organisation automates a process that runs thousands of times a day in order to reduce headcount. A ten-person company automates to stop one person spending a day a week on something that does not need a person. The second case is smaller in absolute terms and often a much larger share of what the business has available.
That difference changes what we build. Systems for smaller companies have to be operable by the people already there, with no dedicated platform team and no monthly licence that quietly becomes the largest line in the budget. We build with as few external dependencies as the job allows, so that what we hand over still runs in three years.
Where this has already been done
Three projects, three sectors, and three different kinds of document — we say which is which on each one.
Event configurator for a hotel group
Eight properties across three regions replaced the email quote request with a seven-step journey that returns a proposal to the client and an already-qualified lead to the sales team.
Delivered project. Catalogue figures, September 2026.
Read the caseAI assistant for life insurance
An insurer replaced its «Call Me» button with an assistant that talks, gathers what the quote needs and returns a result in the same interaction — at 11pm or on a Sunday.
Delivered project. No conversion figures published.
Read the caseAI for complaint handling in local government
An agent that reads the complaint, summarises it, extracts the location, classifies the responsible service, geolocates the incident and drafts a reply — with three levels of human supervision.
Architecture description, not a delivered project.
Read the caseAI in real estate: an email sales assistant
A property development that answers by itself over email, with AWS Lambda and an LLM using the project’s real knowledge — with nothing for the buyer to install.
Delivered project. No conversion figures published.
Read the caseFrequently asked questions
What is AI consulting, in practical terms?
AI consulting is deciding where artificial intelligence is worth applying in a specific business, and then building it. In practice most of the work is not about models: it is mapping how a process runs today, checking whether the data needed actually exists, and establishing what a good outcome would look like. The build follows that. A consultancy that starts by recommending a tool has skipped the part that determines whether the project works.
How do you decide whether AI is the right answer?
We look at three things: whether the process is repetitive enough to be worth automating, whether the data it needs exists in a usable state, and whether a mistake is recoverable. If a process runs twice a year, or if the information lives only in someone's head, or if an error is expensive and hard to detect, AI is usually the wrong instrument. Saying so early costs a project and saves a client.
We are a small company. Is AI worth it at our size?
Often yes, and for a different reason than at large companies. A large organisation automates to cut headcount from a process that runs thousands of times a day. A ten-person company automates to stop one person spending a day a week on something that does not need a person. The second case is smaller in absolute terms and frequently a bigger proportion of what the business has to give.
Do you need our data before you can say anything useful?
No. The first conversation is about the process, not the data — what happens, who does it, how long it takes and what goes wrong. We only ask to see data when a specific case looks worth pursuing, and we can work under a non-disclosure agreement from the start if you prefer.
What happens if the assessment concludes there is no case?
You get the assessment and we stop. That outcome is more common than the industry likes to admit, and it is worth more than a project built on a process nobody had examined. We would rather decline the work than deliver something that quietly fails to be used.
Who owns what you build?
You do, and we hand over something you can operate without us: the code, the configuration and the documentation. We build with as few external dependencies as the job allows, so that the thing still runs in three years without a subscription we chose on your behalf.