How I think about product and AI.

A few things I have learned to insist on.

1. 01

   Start with the problem, not the technology.

   An LLM is not a strategy.

2. 02

   Discovery before delivery.

   The expensive mistake is usually building the wrong thing well.

3. 03

   AI changes workflows, not just interfaces.

   The better question is rarely “where can we add AI?” It is “how should this process work if intelligence becomes cheap?”

4. 04

   Human expertise is part of the architecture.

   In many enterprise systems, human-in-the-loop is a feature, not technical debt.

5. 05

   Adoption is part of the product.

   A proof of concept that nobody adopts creates no value.

6. 06

   Architecture should preserve options.

   Avoid unnecessary model, vendor and platform lock-in.

7. 07

   Measure outcomes, not output.

   Features shipped are not business value.

Research and publications

Work on agile frameworks and IT sourcing, BizDevOps, and narrative in story-driven video games.

1. Journal article · 2021

   Reconciling agile frameworks with IT sourcing through an IT sourcing dimensions map and structured decision-making (opens in a new tab)

   Fouad Amiri, Sietse Overbeek, Gerard Wagenaar, Christoph Johann Stettina. Information Systems and e-Business Management, 19 (4), 1113–1142. Springer

2. MSc thesis · 2020

   Defining IT Sourcing Strategies in Large-Scale Agile Organisations: A Configuration-based Approach (PDF, opens in a new tab) PDF

   Fouad Amiri. Utrecht University

3. Paper · 2019

   Narrative in story-driven video games: A comparative study of emergent, embedded and mixed narrative techniques

   Fouad Amiri. Utrecht University

4. Paper

   Human-centred BizDevOps

   Fouad Amiri

All publications on Google Scholar (opens in a new tab)

Questions I am developing.

Working titles from current product and AI work. These are not published essays yet.

1. AI transformation · Workflows

   Why enterprise AI transformation should start with workflows

   Use cases are a symptom. The unit of value is the workflow: who does what, with which information, and where judgement is needed.

2. AI products · RAG

   RAG is not a product strategy

   Retrieval-augmented generation is an architecture pattern. The product questions still need answers: who it is for, which job it does, and how success is measured.

3. Search · Commerce

   Why hybrid search still matters in the age of LLMs

   Semantic understanding is a breakthrough. Exact attributes, business rules and evaluation still decide whether search sells.