Overview
A wedge product aimed at ops leaders at 50–500 person companies who feel the spend but can't prove it.
Problem Statement
Finance sees the invoices; IT sees the usage; nobody joins the two. The result is quiet, compounding waste.
Market Need
The enterprise SaaS management category (Vendr, Zylo) works well for large companies. For teams too small for those platforms but too big for spreadsheets, there was nothing purpose-built. That gap had always existed - the new thing was that AI could now read and reconcile invoices without manual extraction.
Target Users
Ops leads, fractional CFOs, and finance managers at 50–500 person companies - people who know they're overspending but don't have the bandwidth to prove it line by line.
Research
Ran 5 in-depth discovery calls with ops leads and fractional CFOs at 50–500 person companies. Every single one had a version of a SaaS audit spreadsheet that hadn't been updated in months. The blocker wasn't lack of interest - it was that line-by-line reconciliation required hours they didn't have.
Insights
- ·The blocker isn't insight - it's the effort to produce it. Everyone knows they're overpaying. Nobody has three hours to reconcile it.
- ·A ranked list of specific dollar savings beats a dashboard. The report is what gets forwarded to the CFO.
Opportunity
Automate the audit that everyone knows they should do and nobody does.
Scope
- ·Invoice ingestion + normalisation
- ·Usage reconciliation for top-20 SaaS tools
- ·Ranked savings report with dollar amounts
Out of Scope
- ·Contract negotiation
- ·Procurement workflow
Assumptions
The central bet: that an ops lead would connect read-only integrations if the payoff was shown as a specific dollar number. People protect their data. I wasn't sure 'estimated savings' would be compelling enough until a pilot user forwarded their report to their CFO on the same day they ran it.
Success Criteria
- ·Target: each audit surfaces a meaningful savings opportunity - enough that forwarding the report to a CFO feels worth doing.
- ·Target: most pilots who run a full audit convert to monthly monitoring without prompting.
- ·Kill condition: if the report doesn't get forwarded in the first week, the artifact isn't valuable enough to drive organic growth.
Expected Behaviour
Connect integrations → run audit → get a ranked list of savings within 24 hours.
North Star Metric
Dollars of Verified Savings Surfaced - because this is the number that earns a second conversation with the CFO.
Supporting Metrics
- ·Audit Completion Rate
- ·Time to First Report
- ·Conversion to Monitoring
Prioritization
Every feature was scored against 'does this add another verified savings line?' Everything else waited.
Wireframes
Report-first design from the start. I built what the report would look like before I built the pipeline that would generate it. The layout was designed to be forwarded - scannable in under two minutes, with a clear headline number.
PRD
One-page PRD per pipeline stage plus a shared savings-rubric doc. The rubric was the most debated artifact: what counts as a verified saving versus an estimate.
Prototype
Manual audit for two people before writing any code. I ran the reconciliation by hand, produced the report in a spreadsheet, and asked whether they'd pay for it. They said yes. Then I automated it.
Final Solution
A read-only integration layer, an LLM-assisted reconciler, and a report designed to be forwarded to a CFO.
Reflection
The decision I'm most confident about: not building a dashboard for v1. The report is what gets forwarded to a CFO. A dashboard doesn't get forwarded. The artifact is the growth loop, not the app.
Lessons Learned
- ·For B2B AI, the artifact matters more than the app. If the output is worth forwarding, you have distribution.
- ·Running the process manually before automating it was the most valuable week I spent on this project. The manual version taught me which edge cases mattered and which didn't.