for essential work
Two labs active
AI should make the world's important work move faster.
We build vertical AI products for professional workflows where progress still depends on documents, sources, expert judgment, review and hand-offs. Narrow first. Real adoption. Then compound.
The biggest AI opportunity is not more answers.
Hospitals, courts, firms, builders, accountants and public services do not run on prompts. They run on evidence, responsibility, coordination, review and trust. That is where AI has to earn its place.
- 01
Essential industries carry too much manual coordination.
The cost is not in the thinking; it is in the assembly around it.
- 02
Generic AI helps in moments. Products change systems.
A chat window does not survive a Monday. A workflow does.
- 03
Vertical workflows are where AI becomes real infrastructure.
Narrow enough to work, important enough to matter.
Broad ambition. Deliberately narrow work.
Each lab starts with one concrete workflow. Every lab teaches us to ingest messy documents, preserve sources, route expert review, track state and improve output quality in production — and those parts carry to the next one.
| No. | Lab | Workflow | Status | |
|---|---|---|---|---|
| 01 | Processual |
Case work for legal teams in Portugal, with sources, review and control. Every citation resolves to the original document — the model never writes one. | Active | processual.pt |
| 02 | EduNexo |
Portuguese teachers organise learning evidence and reach evaluation decisions with more clarity, rigour and control. | Active | edunexo.pt |
| 03 | — |
Document-heavy, high-judgment, costly hand-offs, measurable output quality. If that describes your work, we would like to hear about it. | Exploring | Propose one |
Impact only counts when the product survives real use.
- 01
Useful beats impressive.
The product has to earn repeat usage after the novelty of AI is gone.
- 02
Trust is product design.
Sources, review, permissions and accountability are not enterprise extras.
- 03
Narrow is the path to scale.
A specific workflow is how an AI product becomes real enough to expand.
- 04
Experts stay central.
AI should raise the leverage of professionals, not pretend judgment no longer matters.
What we refuse to build
- Thin wrappers that demo well and disappear from daily work
- Automation that removes accountability from high-risk decisions
- Horizontal tools that force every industry into one interface
- AI theatre: impressive output, unchanged workflow
What we want to make possible
- Experts spending more time on judgment and less on assembly
- Documents becoming structured context instead of operational drag
- Review loops that preserve trust while increasing throughput
- Products narrow enough to work and important enough to matter
Have a workflow where better AI would create real leverage?
We are looking for document-heavy professional work where a narrow product can improve throughput, preserve trust and make an important system work better.