Tendercore
Win more tenders, with less admin — an AI tender assistant for Danish construction
- Role
- Product Lead, Fruity AI
- Year
- 2025
- Client
- VAM Entreprenør
- product design
- design engineering
- B2B
- SaaS
- web
As Product Lead at Fruity AI, I designed and built Tendercore, an AI tender assistant that reads every document, automates sub-vendor pricing, and assembles the Bill of Quantities, so bid teams can focus on the offer that wins the job instead of the admin around it.
01
My role
- Product Design
- UX Design
- User Interviews
- Frontend Development
- Monorepo Architecture
02
The problem
Danish construction contractors run tenders by hand across dozens of disconnected documents — PDFs, Word specs, Excel sheets, and DWG drawings — while chasing sub-vendors for prices and stitching together a Bill of Quantities under deadline pressure. Generic AI chatbots don't help: they don't know Bills of Quantity, the sub-vendor pricing loop, or how a tender actually gets assembled, and they trip over Danish entreprenør terminology, units, and trade conventions. They force yet another tool on the team, and sending sensitive tender contents to a shared model means a contractor's data can become someone else's training set.
03
The process
As Product Lead at Fruity AI, I built Tendercore for the real tender workflow rather than as a generic chatbot. It reads the formats teams already work in — PDF, Word, Excel, and DWG drawings, plus email such as Outlook and Gmail, accounting systems, and storage like SharePoint and Google Drive — and surfaces ambiguities, critical requirements, and necessary assumptions early. It automates the sub-vendor pricing loop and assembles the Bill of Quantities end to end, scoring how much real data backs each suggested price, then outputs into Excel, Word, and Power BI so there's nothing new to learn. The whole system is domain-tuned for Danish construction terminology, units, and trades, and each customer runs in an isolated environment with credentials injected at runtime and nothing shared across customers.
04
The outcome
Tendercore is proven in production at VAM Entreprenør. As Head of Digitalization Henrik Bang puts it, it has made the team better at managing risk early: ambiguities, critical requirements, and necessary assumptions are surfaced faster, giving a more reliable foundation for solution, qualifications, and pricing, and a more professional process all the way through to delivery. It also turns each tender into reusable, structured data about requirements, risks, and prices, converting hard-won experience into an active resource for the next bid.
- Proven in production at VAM Entreprenør
- Reads PDF, Word, Excel & DWG, plus email and SharePoint
- Outputs to Excel, Word & Power BI, so nothing new to learn
- Isolated environment per customer, no shared training data