Enterprise GenAI Programme LeadLondon, UK

Azmain
Hossain

I lead the Gen AI programme within Moody's Insurance, from product strategy to user adoption.

My focus areas at the moment are:

  • What an AI product costs to run, and how to calculate that realistically during ideation & design.
  • Choosing what an AI product depends on, and keeping the expensive dependencies replaceable.
  • Institutional memory built to outlive its interface: durable knowledge, a disposable access layer.
  • Capability delivered into the tools people already use, as skills rather than as more applications.
  • What the AI portfolio takes on, and what it turns down.
Read more on each

Assistant Director, Senior Project Manager (Insurance), Moody's

My flagship work Say hello

Flagship case studies

Flagship

the platform.

Prototype to production, live

Adoption-analytics platform for a flagship insurance risk product

Problem

A flagship insurance risk portfolio reported through manual decks and hand-scraped spreadsheets.

Solution

A production analytics platform senior leaders now run their portfolio reviews on.

Weekly
used by senior leadership to run portfolio reviews
Thousands
of colleagues across the division it was made available to
End to end
from the first version through to the live platform
Read the case study
One central emerald-seamed control spindle connected by precision rails to product, release, adoption and team modules.

Flagship

the factory.

Live platform, multiple teams

Application platform and governance model for internal AI tools

Problem

Every team building an internal AI tool picked its own stack, and had nowhere to run the result: working tools stayed on laptops.

Solution

One template settles the stack, and one platform carries an application from a laptop to production.

Every app
goes through one intake and approval gate
No keys
applications never hold their own model credentials
Hundreds
of people using one of the applications built on it
Read the case study
Loose grey gravel travelling in a rail channel into a softly rounded stone arch, crossing an emerald plane where it resolves into identical honed marble blocks.

Writing

Measure of AI

Essays on how to build, think, and stay valuable in the AI economy

Read every essay

WTF is Sovereign AI?

Should I care?

Short Your Own Product

How to fix per-seat pricing in the age of AI

You Can't Sell The Best Thing AI Does

So what can you sell?

Systems I’ve Built

01

End-to-end meeting-intelligence platform

Audio in, programme intelligence out. The hard part was accuracy at acceptable cost: the pipeline replaced a commercial transcription service with something cheaper and more accurate.

In productionPythonFastAPINext.jsPostgreSQL
Chaotic threads entering a graphite ring from the right and resolving leftward into ordered concentric emerald and bone rails.
02

AI-native project-governance tool

Generates a complete project structure from one uploaded document, on top of a full project-management app. Drew interest from a second business unit's PMO after one demo.

Working prototypeFastAPINext.jsClaudeSSO
A document slab entering a graphite cradle and unfolding into roadmap arches, project bays, a decision drawer and secure key.
03

Document-intelligence pipeline

Metadata extraction, LLM classification and human-in-the-loop validation over hundreds of thousands of unstructured documents.

In buildLLM classificationHuman-in-the-loop
A compressed document archive being sampled, classified, manually validated and assembled into a structured lattice.

The path here

Developer / Support / BA / PM & delivery / AI Products Strategy

A path from software developer through delivery into programme leadership, without stopping building.

More about me