Work / Case study

Iropeak

From AI idea to a working product

AI EngineeringAI EvaluationManaged Services

Overview

From AI idea to a working product.

Iropeak wanted to use AI in its core workflow, but had no clear first use case and no team to build it. tMinus1 scoped one narrow use case, built it, tested it against real data and handed it over.

Client
Iropeak
Services
AI Engineering, AI Evaluation, Managed Services
Technology
AWS

The challenge

Many AI ideas, no clear first use case and no team to build one.

The team knew AI could help, but every idea was either too broad to ship or too small to matter. Manual work was taking up the team’s time, and no one owned an AI roadmap.

What we built

A ranked use-case discovery, then one assistant built on real data, with guardrails and an evaluation set.

We ran a short discovery to rank the candidate use cases by value and risk, then built the top one as a working AI product, with role-based access, prompt and output guardrails, and an evaluation set built from real past cases.

How we delivered it

Discovery and use-case ranking first, then a working prototype on real data. Two rounds of evaluation and red-team testing with the people who do the work. Production release with monitoring, a runbook and a named owner on the Iropeak side.

The result

A working AI product, handed over to and owned by the Iropeak team.

Iropeak owns the code, the evaluation set and the roadmap for the next use case.

Next

Building something like this?

Request a technical call

A technical conversation. A written scope before paid work begins.

Iropeak A tMinus1 case study

Request a technical call