agent foundry labs
Book a call

AGENT FOUNDRY LABS · AGENT FACTORY · LONDON

AI agents your business actually deploys.

Agent Factory is a framework for domain-specific AI agents that run real workflows in production — evaluated, monitored, and deployed. Not demos. Whatever the domain.

Live in production today — a tender-intelligence platform for Magellan Circle, and a compliance-first outreach engine. Two domains, one standard.

Book a 30-minute call How Agent Factory works

THE PROBLEM

Most agents die in the demo.

An agent that works in a notebook is not a system your business can run. The distance between a convincing demo and a deployed agent is evaluation, tool-use fidelity, retrieval that holds up on real data, and the operational discipline to run it every day. Closing that distance is the whole job. It's what Agent Factory is built to do.

THE FRAMEWORK

One framework, configured for any domain.

Agent Factory is a product, not a fresh build per client. Five composable layers, configured to a workflow rather than rebuilt for it — so every agent runs on the same production line: measured, traced, deployed. We've pointed it at procurement and at outreach research; the layers don't change, only the domain does.

01RETRIEVAL

Hybrid and vector retrieval, indexed against any document corpus. Real data, not toy data.

02TOOL USE

A typed tool layer that calls real systems and APIs with validated inputs and outputs.

03EVALUATION

A built-in eval harness. Every agent ships measured, not asserted.

04OBSERVABILITY

Tracing, logging, and monitoring, built in from day one.

05ORCHESTRATION

Multi-step workflows that fail safely and recover.

IN PRODUCTION

Two agents running today. Two different domains.

The same production line, configured twice. Proof that the discipline travels — not just the domain.

CASE A

CLIENT
Magellan Circle (EU public-affairs & funding advisory)
DOMAIN
Procurement / tender intelligence
DURATION
12-month engagement
SCOPE
Multi-source EU tender ingestion · AI enrichment + semantic retrieval · structured-criteria extraction · retrieval-grounded, fully-cited first-pass bid drafting
OUTCOME
Weeks of manual tender research compressed to hours
STATUS
Live in production
STACK
Edge-native (Cloudflare) services · durable agent orchestration · hybrid + vector retrieval · eval + health monitoring · every claim cited (never-infer)
Read the Magellan Circle case

CASE B

BUILT BY
Agent Foundry Labs (in-house, self-built)
DOMAIN
B2B outreach / sales research
SCOPE
Declarative ICP → web research → fit-scoring & lead-typing → governed, on-brand draft generation; read-only, daily-capped by design
OUTCOME
One profile definition → 18 researched, 15 fully-drafted leads; zero account risk; agent running-cost engineered down ~60–70%
STATUS
Live, in daily use
STACK
Provider-neutral agent runtime · MCP tool layer (web search + read-only browser) · policy-as-data governance · hard compliance rules enforced on every message
Read the outreach-engine case

The first reads thousands of public tenders and drafts cited bid responses; the second researches real prospects and drafts governed outreach. Different domains, one standard — governed, measured, and deployed, not a demo.

Want one on your workflow? Book a call

WHO BUILDS IT

Built by a CTO who's shipped at production scale.

Agent Foundry Labs is led by Haroon Latif — 15 years building enterprise SaaS at production scale. Most recently Chief Technology Officer of Airportr Technologies, the travel-tech platform used by British Airways, Swiss, Lufthansa, Virgin Atlantic, and American Airlines across 6 airports in 4 countries. Earlier, he architected real-time recommendation and data platforms at dunnhumby — the data-science company behind Tesco Clubcard — building personalised experiences for 20M+ shoppers across global retailers including Tesco and Kroger.

HOW WE WORK

We agree the number first — then prove we moved it.

  1. FIND

    the workflow costing your team the most.

  2. BASELINE

    measure where it is today: time, cost, or output — the KPI we agree up front.

  3. IMPROVE

    an agent, automation, or sometimes a simpler process — whatever actually moves it.

  4. MEASURE

    the same KPI again. Success is a number we both agreed, not a feeling.

We start with a scoped pilot on one of your workflows, prove the agent against real data and the measure we agreed, then take it to production with you. One senior engineer leads the work end to end.

QUESTIONS

Answers, before you book.

What are production AI agents?
Production AI agents are AI systems that run a real business workflow every day — evaluated, monitored, and deployed — rather than a one-off demo. They hold up on real data, call real tools with validated inputs and outputs, and are measured against an outcome agreed up front.
What is Agent Factory?
Agent Factory is Agent Foundry Labs' framework for shipping domain-specific AI agents into production. It is five composable layers — retrieval, tool use, evaluation, observability, and orchestration — configured to a workflow rather than rebuilt for each client.
How is a production AI agent different from a demo?
A demo proves an agent can work once; a production agent works every day. The distance between them is evaluation, tool-use fidelity, retrieval that holds up on real data, and the operational discipline to run it — and closing that distance is the whole job.
What does an AI agent development company do?
An AI agent development company designs, evaluates, and deploys AI agents that automate real workflows. Agent Foundry Labs works product-first: it starts with a scoped pilot on one of your workflows, agrees a measurable outcome, then takes the agent to production with you.

GET STARTED

See Agent Factory on your workflow.

Book a 30-minute call