Inigo AnguloContact
Work01How I think02Lab03Writing04
Build: AI native development

From opportunity to production, with AI agent teams and human gates.

For companies and founders who need to turn an idea, an opportunity or an internal problem into working software. The same product judgment I have applied for 9+ years, carried further: I define and scope the product, orchestrate a team of AI coding agents to build it under human review and quality gates, and own testing, security, deployment and the handover. Not a prototype: a system in production that your team can continue. I did it on Acervo first, for a private client.

  • Product definition and scope
  • AI native development workflows
  • Agent orchestration
  • Human review and quality gates
  • Testing, security and deployment
  • Production ownership and handover
01  AI build & deployment

Working software, not a roadmap slot.

I scope what’s worth building, orchestrate a team of AI agents to build it, and own the rollout to production, inside your business.

17 daysfirst commit to first production deploy on Acervo: a thin first version, not the finished product
445pull requests merged in seven weeks, every one through five checks
15gated releases, each with a security report and one human approval
0production incidents caused by an agent change

What this is for

01

An internal tool nobody has time to build

Operations, back office, a customer portal. The business case is obvious and it has been on the backlog for a year. Weeks, not quarters.

02

A product bet you want to test with real software

Not a Figma prototype and not a no code mock. A thin product in production with real users, so the decision is made on evidence.

03

A team that wants to adopt agents safely

You have engineers and you want the method: how to split work, what gates to put in, what to never let an agent touch. I bring the system I run and adapt it to yours.

How a build runs

  1. Define and scopeOne week. What the product must do, what it must never do, the data model and the security boundaries. Written as tickets with acceptance criteria. If you are not sure what to build yet, the product opportunity assessment below is this step on its own.
  2. Build with gatesAgents work in isolated branches. Every change passes lint, types, tests, a secrets scan and a review before it reaches main. No agent has production credentials.
  3. Human approvalReleases are promoted by a person, after a security pass. The approval is one id and one letter, and it is logged.
  4. Ship and measureProduction from week two or three. Analytics and an audit log from day one, so we know what is used.
  5. Hand over or operateDocumentation, decision log and the agent playbook. Your team keeps the system, or I keep operating it for a period.
A first step that fits most situations

Product opportunity assessment. Two weeks to understand the business opportunity, identify what should be built, define an initial scope, name the risks and dependencies, and recommend an execution approach. It stands on its own, and a build that starts from it starts from a written scope instead of a conversation.

Start with the assessment →

Proof, not promises

Acervo

A production platform for car collections, one person plus eight agents

I defined the product, designed it and shipped it with an AI engineering team under two human gates. The case shows the real numbers from the GitHub API: 445 merges, 1,456 CI runs, 15 releases, and the day the security gate said no.

Read the case study →
A production platform for car collections, one person plus eight agents

Who this is for

A good fit

  • Companies with a clear problem and no engineering bandwidth
  • Founders who want a real product in production before raising or hiring
  • Engineering leads who want to introduce agents with guardrails
  • Internal tools, portals, dashboards, workflow products

Probably not

  • Hard real time systems, embedded, or anything where a mistake is physical
  • Regulated workloads that require certified processes I do not hold
  • Projects where nobody on your side can own the product after hand over

Questions I get asked

Is this no code?

No. It is real code in a real repository, with tests and a release process, written mostly by AI agents under my direction and review. You own it and any engineer can continue it.

How do you keep agents from breaking production?

Isolation and gates. Agents never hold production credentials, every change passes automated checks and a review, and promotion to production is a human decision with a security report attached. The Acervo case documents the whole flow.

What is included in a typical engagement?

Scope, design, build, testing, security review, deployment and a documented handover. The first version is a sprint of two to six weeks; after that, either your team continues or I keep shipping for a period. Agent costs are a few hundred francs a month and I pass them through at cost.

What does 17 days to production prove, and what does it not?

That one person with a scoped product and the right gates can have a thin first version in production in under three weeks. It does not mean a finished product: the fifteen releases after it are where most of the product lives.

Which stack?

TypeScript, Next.js, Postgres (Supabase), Vercel for most products, because agents are strongest there and the ecosystem is boring in the good way. I adapt to yours when there is a reason.

Do you work on site?

Hybrid near Zurich, remote across Europe. A build needs one kickoff in the room and weekly demos; the rest does not.

Discuss a product idea.

Tell me the problem and who it is for. Thirty minutes, no deck; I will say whether it should be built, and how.

The same profile, other directions

Build, lead and optimize are three dimensions of one product practice, with hands-on design in all of them. The case studies show the same person doing each.