Proku is building an AI-native bid management platform to help Swedish B2B companies selling to the public sector win more tenders, by combining each company's own bid history with outcome data and human corrections into a system that improves with every bid written.
Public procurement in Sweden represents roughly SEK 900 billion per year across approximately 18,000 published tenders. On the supplier side, the process is brutally inefficient. A mid-sized consulting or services company routinely spends 40 to 120 hours on a single bid: reading hundreds of pages of requirement documents, hunting down CVs and reference assignments, rewriting the same quality descriptions for the fifth time, and then reformatting everything to match a procurement officer's template. Win rates typically sit around 20 to 30 percent, which means most of that work gets written off.
The problem hits hardest for companies competing on qualitative criteria, where evaluation is based on descriptions of methodology, competence, complexion of the tender and delivery rather than price alone. That is exactly where human writing determines the outcome, and exactly the part that is least systematized today. The knowledge lives in the heads of one or two people, rarely in a system. When they leave, it disappears.
The consequence is that companies either walk away from business they could have won, or burn their margin on bid work that never converts.
Proku is the workspace and data plattform where the bid is actually written, not another monitoring tool that stops at sending notifications.
The platform ingests the tender documentation and breaks it down into a structured requirement matrix covering mandatory criteria, weighted criteria, and the evaluation model. It then generates drafts per requirement, grounded in the company's own previous bids, CV database, and reference assignments, with integrations to the systems where that material already lives: Cinode, CV Partner, SharePoint. Every draft is traceable to its source, which matters in a context where an inaccurate claim can disqualify the entire bid.
What makes this hard to copy is not the model but the data. Because we own the writing surface, we capture three signals no monitoring service has access to: what was written, what the human changed before submission, and whether the bid won or lost. That combination means quality compounds with usage, both per customer and across the customer base.
My co-founder Sadegh and I previously built and sold a digital agency. We spent an unreasonable amount of our own time on public sector bids, and kept having the same experience: we knew we were the right supplier, but lost because we did not have the energy to describe it as carefully as we had in the last bid. The work was repetitive, the knowledge already existed in the company, and we still started from a blank document every single time.
When we started talking to other companies, we realized this was not our incompetence. It is a structural problem across the entire supplier side. The public sector has digitized its half of the process. Suppliers are still in Word.
This is a market with enormous volume, clearly measurable outcomes, and almost no modern software. We believe whoever owns the writing surface toward the public sector in the Nordics builds something very large, and we intend to be the ones who do it.