ES

About me

I am Head of AI GTM Engineering at Sesame HR. Before that I was Global Lead of automation and technology at Dentsu in London, and I co-founded a conversational AI startup. I see every company as a system: business, people, tools and AI. My work is to make it programmable.

What a programmable organization looks like

My view: a company is a system, and with AI it can be programmed. Its processes are in view and change under control. In every area, AI agents do part of the work, people bring judgement where it matters, and someone decides what gets built and answers for the result. I apply it every day at Sesame HR, across six areas and five markets, and at other companies that work with me.

Each agent or tool specializes in one thing and is part of a chain: it uses what others produce and feeds other flows. That way every decision is made with the best information. They are gears of the same engine.

Portrait of Ibán Ríos in a grey blazer and white shirt

Ibán: decides and is accountable

I decide what gets built, set the rules the agents work under, and answer for the result.

One possible example of what it could look like, area by area:

  • Marketing

    One agent watches the campaigns and another prepares ads and content in several languages. What they learn flows to sales and data.

    People choose the message and the budget, and approve what gets published.

  • Sales and revenue

    An agent supports the sales team: it qualifies every lead using what marketing already knows, hands it to the right person, and can even close simple sales. Every closed deal goes back to marketing to sharpen who to look for.

    Salespeople own the conversations and the deals that matter.

  • Data and market

    One agent tracks competitors and another measures which channel brings revenue. The other areas draw on this, and leaders ask in plain language.

    Leaders decide what the numbers mean and what to do next.

  • Internal tools

    Agents build and look after the portals and tools that connect each piece to the others.

    The teams that use them say what they need and whether it works.

  • Quality and security

    An agent checks every change, whichever area it comes from, before it goes live, and keeps keys and permissions locked down.

    A person sets the rules, not an agent, and only a person can loosen them.

  • Research

    An agent reads documentation, regulation and new tools, and writes down what it learns for the others to use.

    A person decides, not an agent, and answers for the decision.

What stays human

  • Strategy: deciding what matters and what to leave out.
  • Judgement: knowing when an answer is wrong, even if it looks right.
  • Relationships with customers, colleagues and partners.
  • Taste and new ideas.
  • Leading the people who work around the system.
  • Answering for the result.

The difference between a demo and a reliable system

Today anyone can ask an AI for a website, a chatbot or an agent, and it works by the evening. Accepting what it produces without reading or testing it is called vibe coding. For a prototype, that's fine. For something that touches customer data, money or your company's name, it is not.

An agent reads emails, web pages and documents, and if one carries hidden instructions, a badly built agent follows them. This is called prompt injection, and no single trick fixes it. What works is how it is built: each agent with only the permissions it needs, anything from outside treated as information and not as orders, and everything logged.

Why me

I have worked in agencies and in large companies, across marketing, revenue and product. At Dentsu International I ran the automations for the whole network from London. In 2021, before ChatGPT, I co-founded and led Chat Ergo Bot, a conversational AI startup.

Today I lead the internal AI projects at Sesame HR. Each team is a client: I build on its real data and I measure myself by its result. I teach at EDEM business school and build my own products, such as ValidWise and Privacy Guard.

Why this is not vibe coding

A demo built with AI works until it fails. And when it fails, it is usually with your data or your customers'. These are the four differences that matter.

  • Security and data

    Vibe coding

    Keys end up inside the code, the agent obeys instructions hidden in an email, and customer data travels to unreviewed tools.

    How I do it

    Security comes first: keys outside the code, only the permissions each task needs, and nothing from outside is treated as an order. GDPR and the AI Act are part of the design from the start.

  • Reliability

    Vibe coding

    It works in the demo and with ten users. With real customers, it falls over.

    How I do it

    Automatic tests, usage limits and cost measured per task. Every change passes checks before it goes live. I have been building systems for more than ten years and I know where they break.

  • Maintenance and traceability

    Vibe coding

    Each automation runs on its own. When it fails, nobody knows what it did or why, and nobody dares to touch it.

    How I do it

    Every piece is documented, tested and connected to the others. Everything is logged: what each agent does, why, and what it costs. What no longer serves is retired.

  • Business judgement

    Vibe coding

    It solves the task you asked for with the tool it knows, without seeing the rest of the company.

    How I do it

    I have worked in business, marketing, sales and product. I see how each piece affects the others before I build it. And keeping up with AI is part of my job.

How I work

My team: me and my AI agents

I work with AI agents that build, review and research. Each one has a role and limits it cannot skip. I decide and answer to you.

Portrait of Ibán Ríos in a grey blazer and white shirt

Ibán: decides and is accountable

I understand how your company works, decide what gets built and how, and answer to you for the result. I set the rules below, and the agents cannot loosen them.

  • Builder agent

    Writes the code, the tests and the documentation for each change.

    Cannot approve its own work. It is judged by tests it cannot touch.

  • Reviewer agents

    Read the sensitive changes looking for bugs and risks. They did not write them.

    Can only read: they do not touch the code they review.

  • Research agent

    Reads documentation, regulation and tools before anything is decided.

    Installs nothing. What it proposes does not go in without my go-ahead.

  • Automated checks

    Check every change: security, accessibility, speed and that nothing breaks.

    Can only get stricter. Loosening them takes a decision of mine, in writing.

Rules nobody on the team skips

  • No key or password in sight: not in the code, not in messages, not in logs.
  • Agents work in isolation, with only the permissions they need.
  • Nothing goes live without passing the automated checks.
  • If something breaks when it goes live, it gets fixed or rolled back. It is never left broken.
  • Every change is on record: what was done, why and who asked for it.

Journey

Oxford to London

2014–2020

Turned a side interest into my job in Oxford: from webmaster to manager at OISE Languages in under a year.

Then London. Paid search at PHD Media (Omnicom) for Unilever and Upfield, and independent performance marketing for MOBS Shoes. Then Dentsu International, one of the big media agency networks: Global Lead, Automation & Tech across Dentsu X and iProspect, for Apple, Square, HTC and Le Creuset.

Employee of the Month at PHD Media. Star of 2019 in the Dentsu International PPC community.

Built GeoAdsFinder there, a tool to check search results that was adopted across the Dentsu network.

Founder

2021–2023

Co-founded Chat Ergo Bot and led it: a platform for building chatbots with conversational AI. Before ChatGPT: we arrived early.

It started with #YoMeTesteoEnCasa (2020), the first COVID-19 self-assessment tool for the whole of Spain, built in 24 hours to take pressure off overwhelmed emergency phone lines. Chat Ergo Bot grew out of it.

Selected by Lanzadera in 2023. Top 75 startups of Spain and Portugal, El Referente, 2022. Planeta Chatbot finalist.

Chat Ergo Bot closed in 2023 because of a co-founder disagreement and the upheaval large language models brought to the market, not because the idea was wrong. It lives on in AIsWise.

Sesame HR

2023–now

Joined as Senior Paid Media & Marketing Operations Specialist in August 2023. Head of Paid Media & Marketing Operations thirteen months later.

MVP of 2024, company-wide across 370 employees. PPC King, Q1 2024.

Managed over €600K a month in paid media across five markets. Rebuilt how campaigns are tagged and tracked, led the move to Google's Consent Mode v2 and built attribution across ad platforms.

Building at scale

2025–now

Interim Head of Demand Generation & Growth: a twelve-person team across five countries.

From April 2026, the first Head of AI GTM Engineering at Sesame HR: a mandate over internal AI projects across the company, reporting to the CRO, COO and CEO.

Alongside that, three products of my own are live (GeoAdsFinder, AIsWise and ValidWise), plus SOVEREIGN, the method this site is built with.

Visiting lecturer at EDEM business school, teaching some of the AI subjects of its Master's in UX/UI Design, AI and Digital Strategy.

Leadership and scope

What I answer for today, and what I answered for before.

  • Scope

    Internal AI projects across the whole of Sesame HR, as its first Head of AI GTM Engineering, since April 2026.

  • People

    Interim head of a twelve-person Demand Generation team across five countries, from December 2025 to April 2026. Before that, in London, I trained the search teams across the Dentsu network in automation.

  • Markets

    Five markets at Sesame HR. From London, Apple worldwide, Square in the UK and Unilever in Europe.

  • Stakeholders

    I report to the CRO, the COO and the CEO, and work with sales, finance, customer success and marketing.

  • Decisions

    What gets built and in what order in the company's internal AI projects, and the rules every agent works under.

  • Impact

    Ten-plus internal tools in production and the company's MVP of 2024.

What I've built

Independent products and projects I have launched

GeoAdsFinder

For SEO and paid search teams. It shows what someone searching Google from any city or country sees, ads included, and lets you see competitors' ads on Meta, Google, TikTok and LinkedIn. Built at Dentsu International for a problem the whole network had, and adopted across it. Today it is a standalone product, free in its basic version, adding visibility in AI search.

AIsWise

For companies that want their website to serve everyone by voice and text, people with disabilities included. Today: AI assistants with WCAG 2.2 accessibility built in. In development: fixing a website's accessibility barriers faster, with a before-and-after audit as proof. It grew out of Chat Ergo Bot.

ValidWise

For companies with a website in Europe that want to know what they are missing to comply. It checks a website with 70 automated tests in five groups: accessibility (EAA, 30), GDPR (15), security headers (14), the AI Act (5) and legal requirements (6). It finds issues; it does not certify compliance.

SOVEREIGN

My method for building software with AI. Every step passes automated checks, and the agent that builds never signs off its own work. This site is built with it.

Privacy Guard

For anyone who wants to know whether someone is watching their Mac. Free menu-bar app for macOS: it detects what monitoring or management software is installed, company or personal, with no admin rights and no network calls.

#YoMeTesteoEnCasa

In 2020, the first COVID-19 self-assessment tool for the whole of Spain, built in 24 hours with two friends. A chatbot guided each person and gave them the emergency contacts for their area. Local authorities and hospitals backed it, and Expansión and Xataka, among others, covered it.

What I build inside Sesame HR and for my clients belongs to them and is not sold elsewhere. There are examples above, in “What a programmable organization looks like” and in “My team”.

Recognition

  • MVP of 2024 — Sesame HR, company-wide across 370 employees
  • PPC King, Q1 2024 — Sesame HR
  • Star of 2019 — Dentsu International PPC community
  • New Employee of the Year, nomination — Dentsu X, 2019
  • Constructive Rebellion, nomination — Dentsu X
  • Employee of the Month — PHD Media / Omnicom, March 2018
  • Top Paid Media Specialist — Mayple
  • Lanzadera Selection 2023 — for Chat Ergo Bot
  • Top 75 Startups of Spain and Portugal 2022 — El Referente, for Chat Ergo Bot
  • Planeta Chatbot Finalist — for Chat Ergo Bot

Got a task you want to hand to AI?

Tell me what the task is and how it is done today. I will tell you where I would start.

Tell me which task you want to hand over