Services AI First

AI strategy and delivery

Get AI past the demo and into production.

Most companies are experimenting with AI and getting very little from it. I help leadership, engineering and product teams get from curiosity to something that ships — and keeps working without me.

Diogo Figueiredo
  • 2026

    co-founded Unblock AI, a monthly meetup in Lisbon for people building with AI.

  • 4M

    people using the products I have led.

  • 17+

    years building software, from startup to enterprise.

Is this for you?

  • You have tried AI tools and nothing reached production.
  • Your team is building AI features without a way to judge the output.
  • Leadership wants a plan, not another demo.

What I do

AI strategy

A short list of bets, not a technology tour.

  • Where AI actually helps your workflows
  • Build, buy or fine-tune
  • Sequenced by impact and risk
  • Written for the people who approve the budget

LLM in the product

Features that hold up on real input.

  • Features on OpenAI, Anthropic or open models
  • Prompts that survive production
  • RAG over your own data
  • Evaluation and quality metrics

Workflow automation

Less repetitive work, same accountability.

  • Internal tools for ops and support
  • Multi-step agents with limits
  • Wired into your data and APIs
  • Cost and latency kept in check

Team enablement

A team that keeps shipping after I leave.

  • Workshops on your backlog
  • AI-native development patterns
  • AI literacy for non-engineers
  • Habits and ownership

How it works

  1. Discovery

    I look at the workflows, the data and the team, and find where AI is worth the effort.

  2. Pilot

    One scoped use case, shipped in weeks, measured against something real.

  3. Scale

    The pilot becomes a repeatable capability, with tests and an owner.

  4. Sustain

    Tooling, habits and review, so it keeps improving without me.

Formats and duration

Advisory
A day or two a month. Strategy, hiring, hard calls.
Pilot
Four to eight weeks on one use case, measured.
Workshop
One to three days with the team, on your own workflows.

Questions

Do we need our own model?

Usually not. Most of the work is choosing the right model, getting your data to it, and measuring what comes back. Fine-tuning comes later, if at all.

How long before something ships?

A scoped pilot in weeks, not months. Then we look at the result and decide whether it deserves to scale.

Our team has never built with AI.

That is the normal case. Workshops and pairing on your own backlog, so the knowledge stays when the project ends.

How does it start?

A call about what you are trying to do. If it looks useful, I send a short written plan: the first month, the cost, and what I need from you.

What are you trying to ship?

Tell me where AI is stuck between the demo and production.