x
Minerva
Redesigning an AI tender-search platform to learn from how clients actually work - and feel trustworthy from day one.
Industry / Business type
B2B AI SaaS
Scope
Product UX redesign
Results
1-week redesign
Onboarding steps cut ~60%
Match relevance, continuously learned
The challenge
What was the problem?
Minerva's problem went deeper than interface polish. At its core is an AI tender search that only works if it learns the business well enough to surface the right opportunities - and that learning was too heavy, too fragmented and too dependent on setup steps users skipped.
Clients completed profiles only partially, skipped configuration and rarely came back to refine it. So the system either missed valuable tenders or returned too many irrelevant ones, which quietly killed trust. There was a perception gap too: users didn't review disqualified tenders, they just assumed the system had failed.
The same trust-through-clarity challenge showed up in Lokero, though there it lived inside a marketplace flow rather than an AI-assisted enterprise product.
The fix
Here's what we've changed
We approached Minerva as a self-improving product experience, not just a search interface. Instead of front-loading the thinking into long onboarding, the system was redesigned to collect better signals continuously - through guided feedback and clearer evaluation moments.
We reworked the experience around explainability. Users can now see why a tender appeared, why another was rejected, and where the system is uncertain - and Minerva asks for lightweight corrections in the moment, so it improves through small interactions instead of heavy setup.
We also rebuilt the tender list for high-volume evaluation, surfacing the key information earlier so users can scan, compare and decide without opening every tender. The product feels less like static search and more like an active decision-support layer - closer to choice architecture we used in modue, applied to enterprise recommendation logic.
The results
And did it actually work?
The result is a product that feels more teachable, more transparent and far easier to trust.
Instead of heavy upfront setup and hoping accuracy improves later, the redesigned loop lets the platform learn through interaction, explain its reasoning, and show users how their feedback shapes future results. That feeds directly into accuracy, engagement and retention - because when users trust the feed, they keep using it.
That's what makes this strategically valuable: it doesn't just improve interface quality, it improves how the product learns and how users understand that learning.
If you're building a SaaS product where clarity, trust and friction decide whether people stick, start with the 48h Audit.
Other projects
