# Old St Labs

> Old St Labs is a reviewed AI agency in London, United Kingdom, listed in the AIAgencies.eu register and last assessed 2026-09-23.

Canonical page: https://www.aiagencies.eu/agency/old-st-labs
Generated: 2026-09-28 · Curated by Gabor Kiss · No paid placements — inclusion, score, and rank are never sold.

## Facts
- Location: London, United Kingdom
- Website: https://www.old.st
- Hourly rate: €45-85/hr (directory-listed)
- Team size: 50-249
- Founded: 2020
- Services: AI Development, AI Consulting
- Industries: Healthcare, SaaS & B2B, E-commerce
- LinkedIn: https://www.linkedin.com/company/old-st-labs

## About
Old St Labs is a London product studio founded in 2020, building web and mobile applications with an AI software development line. It works largely with early-stage and scaling companies.

## Documented clients
Documented engagements: Brainie (Brainster GmbH), Livis Analytics, Mariana UFP, Mantra Systems, Warden Consulting, Estima, Flux HQ, Association of British Climbing Walls, Impiricus, and Autistica. All 10 carry a source URL on file.

## Curator assessment
Reviewed 2026-09-23 by Gabor Kiss against published evidence.

One of the 11 published case studies is AI, and it explains the build: Brainie, a math tutor that routes GPT and PaLM through a custom session engine that adds or drops lesson modules by student performance, beside a LaTeX workboard.

The second visible AI engagement, a transcript-scoring platform for Livis Analytics, is documented in a Clutch-verified founder interview describing prompt iteration that brought scoring variance down to a standard deviation of 0.1–0.2.

Neither shows a system under real load: Brainie's 78% "indistinguishable from human tutoring" figure comes from test sessions and its testimonial slot still reads "xxxx", and Livis was "just about to go to market" when its founder reviewed the work in February 2026; four further AI builds appear only as one-line homepage entries.

The weaker fit is buyers who need evidence of an AI system running in production, or machine-learning work beyond LLM features inside a web or mobile product.

### Key strengths
- The Brainie study describes architecture, not just outcomes: a connector layer handling prompts, context and response evaluation across two model providers, and a map of concept prerequisites
- Repeat work is visible: Mariana UFP runs two compliance platforms the studio built (trade surveillance and client-communications monitoring), and Military App is a long-term embedded engagement
- Named people on both sides: co-founders Rob Woodhead and Rob Lynch, a named Director of Engineering, and client executives quoted by name, such as the CEO of the Association of British Climbing Walls

### Good to know
- Neither AI engagement read reports numbers after launch—ask for an AI product that is live with paying users and how its output quality has held up
- Delivery runs through teams in the Philippines and South Africa under UK leadership—ask who on your team has shipped LLM work before and where they sit
- The process page says tooling ranges from no-code platforms to AI engines—ask which parts of your build would be no-code before comparing quotes

---
Source: AIAgencies.eu — the curated register of European AI agencies. Rate provenance: "agency-confirmed" = disclosed via submission or claimed profile; "directory-listed" = Clutch or two agreeing B2B directories (lone unverified sources are never written). Full method: https://www.aiagencies.eu/methodology · Rate index: https://www.aiagencies.eu/rates