Skip to content
Epic Software Labs
← All work

UK public-sector procurement

AI procurement startup

Took a prototype eSourcing portal from founder-built demo to a production-ready, two-sided procurement platform — while the pre-seed closed and the founding engineering team came together.

Role
Founding technical partner
Stage
Pre-seed funded
Sector
UK public-sector procurement
Two machined halves of one form interlocking along a toothed seam — two sides of a market meeting and fitting exactly.

Outcomes

Round closed
Pre-seedRound closed
Production launch (buyer + supplier)
0 → 1Production launch (buyer + supplier)
Founding engineers hired
3Founding engineers hired
Enterprise integration path shipped
SAPEnterprise integration path shipped

Context

The client was a founder-led prototype targeting AI-assisted bid writing for suppliers responding to UK public-sector tenders. Promising demos, real design partners, but a single non-technical founder, no engineering hires, and a seed conversation starting in weeks. The product also needed to expand beyond suppliers to serve buyers — framework providers and procurement managers — which is where the bigger market was.

The problem

Three things had to happen in parallel, none of which the founder could do alone: decide whether the prototype’s architecture could carry a two-sided product or needed a rebuild; get a credible technical story and a technical counterparty in front of pre-seed investors; and hire a founding engineering team without burning runway on the wrong first hires.

  1. 01

    Re-architected the prototype into a production monorepo

    Carried the supplier and buyer journeys, shared design system, mock-SAP integration for enterprise pilots, and a tender-scraping pipeline feeding the AI bid workflows.

  2. 02

    Closed pre-seed alongside the founder

    Owned the technical diligence track, built the financial model, ran SeedLegals contracting, and acted as the technical counterparty on investor calls.

  3. 03

    Hired the founding engineering team end-to-end

    Wrote the hiring plan, designed the interview loop, ran the technicals, made the offers. Three founding engineers hired.

  4. 04

    Expanded the product from one-sided to two-sided

    Grew the platform beyond suppliers to serve buyers as well — framework providers and procurement managers — opening the larger market without disrupting the seed conversation or the design-partner pilots.

  5. 05

    Set the 12-month roadmap to seed milestone

    Tied to two design-partner pilots — one supplier-side, one buyer-side framework provider.

  6. 06

    Built client-specific features for sales demos

    Shaped the product surface to what each prospect needed to see to progress the deal, rather than insisting they meet a generic demo halfway.

The Maverick moment: the SAP we couldn’t get

A MENA-based enterprise technology buyer needed an SAP integration to move forward in their evaluation. SAP’s partner-onboarding timeline didn’t fit the sales cycle — the meeting was weeks from being lost. Rather than wait, we built a SAP-lookalike service that mirrored the real SAP API shape and the specific PO flow the demo needed, layered the client’s procurement flow on top, and ran the end-to-end demo as if the integration were live. The deal progressed to a second meeting; without it, it would have stalled in the sales cycle. The mock-SAP service still lives in the dev stack and now doubles as a local-development SAP for new engineers.

Takeaway

When the by-the-book path doesn’t fit the runway, find the version of the technical answer that unblocks the commercial moment without burning the architecture behind it.

Trade-offs

Every architecture is a set of things you chose not to have. These are the ones that mattered here.

  1. Azure over AWS

    The default would have been AWS, but the design partners and target enterprise buyers were Microsoft shops with SAP integration requirements. Optimised for the buyer-side sales motion over engineer familiarity.

  2. Bought design tokens, built the rest

    Used the Untitled UI Figma library as the design-system spine instead of rolling our own — saving roughly six weeks of design-system work to redirect into product surface area.

Stack

  • TypeScript
  • Next.js
  • Python
  • LLM orchestration
  • Azure
  • Postgres
  • SAP (mock + integration path)

Capabilities applied

What’s next

Continuing through seed: scaling the engineering team, hardening the production platform for the first paying buyer-side pilot, and supporting the raise.