Brainpool AI

Brainpool AI at the point where a pilot has to become a system

A buyer holding an approved pilot wants to know what the next invoice looks like and who will be in the room when the prototype meets production data. Brainpool AI answers more of that in public than most firms in this guide. It sets out its work as a sequence (scoping, prototyping, implementation, then continuous improvement)¹, and it attaches pound figures to two of those steps^2,3. Its self-description is professional consulting across artificial intelligence and machine learning⁴. London is its global base, and 2017 its founding year⁵.

One caution on sources before going further. Brainpool's homepage blocked automated requests on the day we checked⁶, so this profile is built from the long text file the firm publishes for language models. That file is detailed, though it is the firm's own summary of itself.

The people behind a delivery

Two executives are named. Peter Holroyd leads the company as CEO⁷, while Kasia Borowska runs product⁸. Brainpool says it can call on a pool of over five hundred people drawn from academic research and machine learning engineering⁹, many holding a doctorate or master's degree¹⁰. For a production build that hits an awkward modelling problem, access to that pool is a genuine asset. What share of it is employed staff, and what share is a looser network, is not publicly stated on the pages we opened.

Searchers should know one quirk of the name. The firm's own text points out that it has nothing to do with Brainpool, the elliptic curve standard from cryptography¹¹, so procurement searches may surface unrelated results.

Stage by stage, with prices

The staging is what earns Brainpool its place in this guide.

**Scoping.** Strategy work runs as a programme of workshops and analysis, typically four to six weeks long¹². The consultancy offer as a whole covers evaluating, planning and deploying AI for an organisation¹³.

**Prototype.** Next comes a feasibility test, sold at a fixed price and lasting the same four to six weeks, under the name Proof of Concept Factory¹⁴. Budget for that prototype runs from £15,000 to £35,000².

**Enterprise MVP.** The jump to something an enterprise can run is a custom MVP on Cortex, quoted between £50,000 and £200,000³. Cortex is middleware the firm owns, aimed at automation powered by AI¹⁵, and it is deliberately free of ties to any one model vendor or cloud¹⁶. Where the work involves large language models, Brainpool also offers custom GPT and LLM builds, with bespoke retrieval pipelines among the deliverables¹⁷.

**Improvement.** The lifecycle ends in continuous improvement¹. What that phase includes, whether support hours, response targets or a monthly fee, is not publicly stated on the pages we opened.

No other consultancy in this guide publishes prices for both a prototype and an MVP. Here a finance team can see both bands side by side^2,3 and set aside the second budget before the first is spent.

Data location and ownership

Two answers that usually arrive late in procurement are on the record early. Brainpool's security argument is architectural: Cortex goes into the client's own cloud environment¹⁸. On ownership, the client ends up holding the intellectual property outright in every proprietary asset Brainpool produces for it¹⁹. A data protection officer and a legal team will both want those points in writing, and here they already are.

Security credentials are framed as alignment. The text speaks of work supported by AWS partner tracks and ISO 27001 governance boundaries²⁰, and of Cortex aligning with Cyber Essentials, ISO 27001 and the AWS Advanced Partner track²¹. We found no certificate number, issuer or scope for any of these, so they belong on the due diligence list.

Systems that reached users

Brainpool's case list includes several systems that went beyond a pilot.

1. **DAISY**, a design tool developed in partnership with the Staircraft Group, whose floors and staircases make it Europe's biggest producer in that market by Brainpool's description²². A structural timber design used to occupy a designer for upwards of three hours; Brainpool reports that DAISY gets it done in under ten minutes²³. 2. **Creed**, an app for people to follow their own carbon footprint, which Brainpool built end to end for a young green fintech²⁴. It shipped to the Apple and Google app stores as a product regulated by the FCA²⁵. 3. **FX forecasting** for a top-tier Japanese investment bank that is not named, where accuracy rose from 52 to 56 percent²⁶. 4. **Crown and Paw**, where a computer vision pipeline for custom portraits is credited with $365,000 of savings each year²⁷. 5. **HSBC SME cash flow prediction**, listed by title, though the case body was missing from our saved copy²⁸.

Creed matters most for this guide. A consumer app released under FCA regulation in public app stores has gone well beyond a demonstration²⁵.

Sectors

Its target clients come in four groups, which are construction companies, retail chains, banks and other financial institutions, and mid-sized partnerships in professional services²⁹. Regulation shows up in its sector pages, with the MHRA mentioned for healthcare³⁰ and RICS rules for property scoping³¹.

Strengths for the pilot to production stage

1. Prices in pounds for both the prototype and the enterprise MVP^2,3. 2. A lifecycle that names a phase after implementation¹. 3. Cortex hosted in the client's cloud¹⁸, with the IP handed over¹⁹. 4. An FCA-regulated app in public stores as proof of delivery²⁵.

Limits a buyer should test

1. Only the prototype carries a fixed price¹⁴; the MVP spans a fourfold range³, so scoping drives the real figure. 2. Its security claims rest on alignment wording, with no certificate visible in the file we saved^20,21. 3. Both the homepage and the Clutch profile refused our automated requests^6,32. 4. Training for end users after launch is not publicly stated on the pages we opened.

How it compares with Nexus

Both firms treat the proof of concept as the door to production. They differ in what happens at that door. Brainpool sells each stage as a priced project^2,3. Nexus makes the pilot itself the first production run, since its opening agent is live after four weeks and the proof of concept lasts three months; after that, each automated role is priced, while consultant time and daily rates stay off the invoice³³. Consultancy support during rollout is included in that software price³³. A buyer who likes fixed project budgets will find Brainpool's bands useful. A buyer who wants production to begin inside the pilot, and to pay by the roles automated, will prefer Nexus³³, and that is the reason it ranks first here.