Bool AI Engineering

AI Engineering that works in production.

We build and run AI agents and AI-powered applications, with a person accountable for every change.

Certified Services Partner, Claude Partner Network
13Years
150+Team members
300+Projects delivered
17+Countries
750+Certifications
01Where AI Pays Back

Start where the hours go. Keep a person in charge.

Six processes where we already have AI working for mid-sized companies and enterprises.

Invoices and finance

TodayEvery document retyped into the ERP

With AIData read and matched to the bank statement; a person approves every record

Front desk and service

TodayFive windows open to answer one question

With AIOne screen for every attendance, and an assistant that knows your services

Hiring

TodayRecruiters screening every candidate by hand

With AIAI interviews and classifies candidates; people make the decision

Legacy apps and spreadsheets

TodayCritical logic nobody dares to touch

With AIAudited, covered by tests and rebuilt with agents, with a person reviewing every change

Everyday work

TodayShadow AI, no rules, no view of cost

With AIApproved tools, a register of use cases and an audit trail for the AI Act

Requirements and scoping

TodayThin requirements, surprises found late

With AIAn idea becomes requirements, user stories, tests and wireframes, straight into your backlog

Works with what you already runMicrosoft 365Entra IDGoogle WorkspaceBigQueryOutSystemsMendixAzureInvoiceExpressMoloni
Not sure where to start?

Anyone can generate code with AI. We make sure it works in production.

AI Diagnostic · 2 weeks

Find the row that pays back first.

Short conversations where the hours go, then a written plan: opportunities ordered by impact and effort, and the first project named and sized.

Talk to an expert
02Services

Four outcomes. One accountable team.

Each service line starts small, has delivered work behind it, and ends in something your people use and we can keep running.

Automate a process with AI agents

An agent does the repetitive part of a process. A person approves what matters.

Proof: invoices into the ERP, AI-assisted hiring at a global CX operator, a public services front desk.

First step: AI Diagnostic

Add AI to your applications

AI features inside the applications your people already use, with the cost of every call visible.

Proof: an AI tutor in production for children learning to code, an AI layer serving several education products.

First step: AI Diagnostic

Modernise legacy and low code with agents

We audit what you run, cover it with tests, then rebuild or evolve it with agents.

Proof: 20 gaps closed in an AI-built prototype, a platform MVP in two months, agents working inside low-code platforms.

First step: Code audit

Govern AI across the company

Approved tools, training, a register of use cases and an audit trail for the AI Act.

Proof: a country AI framework for a European food retailer, presented to its head office as a model.

First step: Team training or AI Diagnostic

After go-live

And we keep it running.

Support and evolution in hours packages or a dedicated squad, with the same people who built it. You are never left holding a system nobody understands.

Three ways to start, each with a clear outcome.

Each entry point ends with a decision you can take on your own, and leads to one of the services above.

AI Diagnostic

Know where AI pays back

Two weeks, fixed scope, one written plan.

Leads to: Automate a process · Add AI to your applications

Code Audit

Know if it is safe to run

A production readiness review with a prioritised fix plan.

Leads to: Modernise legacy and low code

Team Training

Make AI a daily habit

Hands-on sessions by area, on your real cases.

Leads to: Govern AI across the company

03Why Bool

Thirteen years of delivery. Now AI Engineering.

Our first AI work dates from 2023, and our AI products have been in client hands since 2024. Every story on this page comes from delivered or ongoing work.

Certified Services Partner, Claude Partner Network
Claude Partner Network

Bool is Certified in the Claude Partner Network.

The Claude Partner Network is how Anthropic works with the firms that put Claude into production for customers. The Services Track recognizes partners for building a certified delivery team, delivering Claude use cases for clients, and proving the work through public customer references.

Our people hold Claude certifications, and we are building from here.

About the program

AI-augmented delivery. What you get.

TeamSmall and senior. Agents do the repetitive work.
SpeedIn production in weeks, not quarters.
ControlA person approves every change.
QualityTests written with the code.
Your dataStays where it must.

We use what we sell. We run Bool the way we propose you run AI.

Agentic engineering is the floor

Every engineer delivers with agents. Our career framework names the human line in every role: what AI may draft and what a person must decide.

Tests written by agents, reviewed by people

About 30 functional tests written by an agent from a spec, then about four hours of human review. The typing goes; the judgement stays.

Estimates built from evidence

Every estimate starts from a written work list, is sized by an engineer and is checked by script before it leaves. No figure is typed by hand.

04Delivery

Small senior teams, amplified by agents.

A Product Engineer owns the what. An Agentic Engineer owns the how. Agents do the repetitive work. You always talk to people.

01

Spec

We agree what to build and how to accept it.

Gate: scope
02

Plan

The agent proposes a plan. A person approves it.

Gate: plan
03

Build

Agents implement, test first.

04

Review

An engineer reviews and owns the merge.

Gate: review
05

Release

Tested, documented, traceable.

AI never prices, commits or signs off. A person does.

Three ways to work with us.

The scope is clear

Fixed price build

Phased deliveries with acceptance gates.

Continuous product work

AI native squad

A dedicated team, monthly and renewable.

Evolution and support

Hours package

Blocks of hours, used as needed.

05Governance

Your data, your control.

Four answers, in place from the first day. Built for the EU AI Act.

Where your data stays

On your infrastructure when it must. Local models when needed.

What reaches the model

No production personal data in prompts.

Who decides

A person approves every change and every release.

What is on record

Every AI call traced: model, cost and time.

Three doors, one model

Every way AI enters the business

  • Everyday AI: approved tools and training.
  • AI in applications: a register of use cases by risk.
  • AI that writes software: human review and an audit log.
Four gated phases

From policy to a working funnel

  • 1. Model
  • 2. Reference architecture
  • 3. Pilot
  • 4. Funnel of use cases
Built withClaudeOpenAI-compatible modelsLocal open modelsLiteLLMLangfuseMCPTypeScript.NETPythonPostgreSQLEntra ID
06Success Stories

AI delivered, not just demoed.

Eight anonymised stories from retail, customer experience, accounting, education and insurance. One per tab.

Case study · AI-Augmented Development · Retail technology · Italy

A certified fiscal integration, built in 11 days

11 daysfrom accepted proposal to working integration
6 weeksBool’s estimate for the same work without AI
0 defectsin production, through the Winter Olympics and Paralympics

The challenge

Cashierless stores in Italy had to register every sale under Italian fiscal law, through a certified provider, by a fixed mid December deadline. The provider asked for payment data the client did not hold.

How we built it

A week of working sessions with the client and the provider mapped the API and the data. One Bool developer and an AI agent then built it: the agent wrote and tested the code, the developer designed and reviewed it.

How the work ran

  1. 1Developer designs the integration
  2. 2Agent writes the code
  3. 3Agent tests it against the fiscal API
  4. 4Developer reviews and owns each change

Anonymised client case · Delivered December 2025 · Java, fiscal API integration

Thirty minutes. One process. A clear first step.

Talk to an expert
07Contact

Start with one conversation.

Thirty minutes to agree which door fits: an AI Diagnostic, a code audit, team training, or a first agent built on your real process. No commitment, and no forty page proposal.

To find us

Office Lisboa
R. do Centro Cultural 27A
1700-106 Lisboa
sales@bool.pt

Your data will be processed to respond to your inquiry under our legitimate interest, in accordance with our Privacy Policy.

08FAQ

Questions we hear before the first call.

How do we start?

A 30-minute call, then an AI Diagnostic or a code audit with a fixed scope and a written outcome.

Who will we work with?

A Product Engineer who runs discovery with your people, writes the spec and owns the backlog and the weekly demo. Agentic Engineers who build with AI agents and own every merge. Specialised agents handle specs, code, tests, reviews and documentation. You always talk to people.

Which models do you use?

The right one per task: Claude, OpenAI-compatible models or local open models. You can choose. An AI gateway lets us change models without touching your applications.

Does our data leave the company?

Not when it must stay. Models can run on your infrastructure, and agents reach your data only through registered tools.

Who owns the code?

You do, on acceptance.

What happens after go-live?

We keep it running, in hours packages or with a dedicated squad.

Do you work with OutSystems and Mendix?

Yes. Our agents work through MCP inside the low-code platforms you already run. See our OutSystems and Mendix practices.

Which sectors have you worked in?

We have AI work delivered or in delivery in retail, customer experience, education, accounting, public sector, insurance, construction, postal and logistics.

Do you work in our time zone?

Yes. Nearshore from Portugal: your time zone, English by default, and teams that start small and grow with you. Offices in Lisboa, Porto, Açores and Castelo Branco.

When will we hear back?

Within one working day of your message.