Why are vendor claims hard to check?
- Hours: they show effort, not output. With AI, a provider can deliver more in fewer hours, or bill the same hours for less work.
- Story points: each team sizes them differently, so you cannot compare providers.
- Vendor reports: the provider chooses what to show.
4 steps to verify throughput
- Get read-only access to the repositories and the ticket tool where the provider works.
- Measure finished work with one unit for all providers, after the work ships.
- Compare the cost per delivery point across providers and your in-house teams.
- Check quality: rework and review results next to the output.
What do you put in the contract?
| Clause | Why |
|---|---|
| Read-only access to code and tickets | You measure from the source, not from a report |
| One unit of delivered work | You can compare providers |
| Share of work done by AI agents | You see where the AI gain goes |
| Quality floor | Output does not go up at the cost of quality |
How Pensero helps
- Read how to compare several vendors on one scale, which KPIs to put in an outsourcing contract and why to trace work to pull requests and tickets.
- Book a scoping session to see delivery points on your own work.
Frequently asked questions
Can procurement do this without engineers?
Yes, if the tool reads the code and tickets and gives one number per provider, with the evidence behind it.
Is lines of code a good measure?
No. AI makes code volume cheap. Measure finished work by magnitude × complexity.
Which tool measures providers on one scale?
Pensero measures providers, in-house teams and AI agents in Pensero delivery points.
What do I do after the check?
Decide what to change, for example the scope, the price or the provider. Then measure delivery, cost per delivery point and quality again.
