Bruno Marques Horta — Founder and AI Architect
Architecture of the AI solutions, design of the agents and flows, and the call on when AI is not the right instrument.
Lisbon · since 2017 · six people
About BigLearn
A consultancy founded in Lisbon in 2017. Almost a decade on, we are a Portuguese applied-AI boutique, working mostly with Portuguese SMEs of 20 to 200 people, with direct senior involvement from diagnosis to implementation.
What it means to be a boutique AI consultancy
It means there are few of us and all of us are senior, with long track records behind us, and no layers between the person who sells and the person who builds. Whoever discusses the problem with you in the first meeting is the one who designs the solution, and the one you go to for a fix, an upgrade to a function, or a change the business has since come to need.
How do you keep continuity with a small team?
The decisions, the architecture, the code and the operating instructions for each project are documented and ready to be handed over. We limit how many projects run at the same time so the capacity is real, and when an additional skill is needed we bring in specialists from our network while responsibility and day-to-day contact stay with BigLearn.
Can a small team support an SME?
Yes. In an SME the work starts from one concrete process, with a defined scope and named owners, rather than from mobilising a heavy structure. The senior team is directly involved in the diagnosis, the pilot and the implementation. When the project calls for additional skills, we bring in specialists while keeping a single point of contact and responsibility for delivery.
And what does that limit?
We take few projects at a time. That is a real constraint and we would rather say it: if the calendar, our areas of work or the industry the project lives in do not fit, we say no instead of accepting and delivering late.
Why 20 to 200 people
It is the size where processes are already written down and have an owner — there is someone to ask how things actually run today — but there is no in-house AI team to automate them. It is where one well-chosen project visibly changes the working day.
The number is the size of the unit we work inside, not a ceiling on the company. Several of our clients are groups with many hundreds of employees — what falls inside that range is the department or the operation we come into: a group's innovation and development arm, the MICE operation of a hotel group, the back office of a division, a municipality's citizen service desk, or the newsroom of a media group. That unit is what needs an owner, a process, and enough resources for the change to show.
Below that, an off-the-shelf tool usually solves it — and that is what we say.
A senior team for AI consulting and implementation in SMEs
Miguel Abreu Peixoto — Lawyer and Legal Counsel
The legal and contractual framework of the projects, AI governance, GDPR and the EU AI Act — involved from the solution design stage, not only in the final review.
Senior software engineering
Development in Python, JavaScript and C. Integration with the systems the company already runs — PMS, CRM, ERP, mail and databases.
DevOps, AWS and cybersecurity
Infrastructure, hosting, going live, and data security: minimisation, access control and an architecture matched to how sensitive the information is.
Design and copy
Interfaces, content and the language the solution reaches its users in. An automation nobody understands does not get used.
Operations and administration
Contracts, invoicing, deadlines and keeping to what was agreed — so the technical work is not interrupted by paperwork.
Six, plus the network
Six people are not enough for everything, and we do not pretend otherwise. We have an established network of specialists we have worked with for years, and we call on them when a project needs a skill that is not our day-to-day.
Security audits and penetration testing. LoRa networks and sensors for smart cities and utility metering. Computer vision on the factory floor. Integration with industrial systems — SCADA, PLC, telemetry.
They join the team, not a subcontract
The difference matters. You keep talking to the same people, delivery stays our responsibility, and the specialist works inside the project rather than receiving it over the wall.
It is also what lets us say yes to projects a team of six, on its own, would have to turn down — without growing into a size that would remove the reason we exist.
Where we say no
There are four kinds of work we do not take. We say so upfront, so nobody wastes their time.
Spam, cold calling and mass contact
We do not build spam, cold-calling or unsolicited mass-contact systems. AI makes it cheap to bother a great many people at once, and that is not work we want to sign.
Projects without an owner on the client side
We do not take projects without someone on the client side who has time allocated. This is not red tape: without someone who knows the process and can decide, the project drags and never reaches production.
Licence resale and vendor commissions
We do not resell licences or take vendor commissions. When we recommend an off-the-shelf tool it is because it is the right answer, not because we earn from it. That is what keeps the technical recommendation independent.
AI where a simple rule works better
When a simple rule solves the problem better than AI, we say so. The goal is not to use more AI — it is to improve the process. We have said this to clients who arrived set on buying a larger project.
Where we have most depth
Hospitality
Guest journey, check-in, concierge, reputation and upsell. Its own section, with published cases and guides by process.
Real estate
Lead qualification, a commercial assistant working over email, and integration with the CRM.
Local government
Citizen service with human oversight, complaint and incident handling, and the EU AI Act framing.
We also have published cases in healthcare, construction, media, food service and insurance. They are all in case studies, each with the problem, the approach and the result.
Questions about BigLearn
How many people work at BigLearn?
Six: AI architecture, software engineering, DevOps and security, design and copy, operations, and legal counsel and compliance. The team is small on purpose — we take few projects at a time, and the person who sells is the person who builds, with no layers in between.
What size of company does BigLearn work with?
Mostly Portuguese SMEs of 20 to 200 people — the size where processes are already written down and someone owns them, but there is no in-house AI team yet. But the number is the size of the unit we work inside, not a ceiling on the company: we also work with groups of many hundreds of employees, on projects scoped to a department or an operation of that size — the MICE operation of a hotel group or the newsroom of a media group, for example.
What does BigLearn not do?
Four things, and we say so upfront. We do not build spam, cold-calling or unsolicited mass-contact systems. We do not take projects without an owner on the client side who has time allocated, because those do not go well. We do not resell licences or take vendor commissions, so the technical recommendation stays independent. And when a simple rule solves the problem better than AI, we say so — the goal is not to use more AI, it is to improve the process.
What if the project needs a skill the team does not have?
We call on an established network of specialists we have worked with for years: security audits and penetration testing, LoRa networks and sensors for smart cities and utility metering, computer vision on the factory floor, and integration with industrial systems such as SCADA and PLC. They join as part of the project team and under our responsibility — your point of contact and the party accountable for delivery stay the same.
How long has BigLearn been operating?
Since 2017, based in Lisbon. We work in Portuguese and in English.
Which sectors does BigLearn have most published cases in?
Hospitality, real estate and local government are the three with most depth, each with its own section and published cases. We also have cases in healthcare, construction, media, food service and insurance.
How does a project with BigLearn start?
With a conversation about one concrete process that is costing the company time. In selected cases the proof of concept is free: it shows real results before any commitment. If there is a fit, we put a proposal with defined scope and cost in front of you before work starts.