Customer story · Accounting
How Wally tests its accounting agent on every production release
Wally builds an AI agent for accountants that connects to the software they use every day. Each production release triggers an Enoki run, so every new version is attacked before customers rely on it.
- The agent
- Works across accountants' software and client data
- The risk
- Client data leaking
- How they test
- On every production release, through the API
Tijs MartensCTOCurrently we have built a flow where every time we do a production release, we trigger an Enoki run. This means that Enoki will test all use cases of our AI agent.
Tijs MartensCTO, WallyAn agent that works with real numbers
Wally builds an AI agent for accountants. It connects to the software an accountant uses in their day-to-day work, which means it handles important data: real clients, real numbers.
Tijs Martens is Wally's CTO, and his biggest concern is data leakage: one of those data sources reaching someone who should not see it.
The worst case scenario is if one of these data sources would have some form of leak. This can't happen.
A run on every production release
Wally built Enoki into its release flow. Each production release triggers an Enoki run through the API, Enoki attacks every use case of the agent, and the findings come back to the team in their own channels. Small or large, they fix them and ship again.
What changed
The findings changed how Wally builds. Third-party tools and external services now get a second look for the restrictions they enforce, and code the agent executes on the fly runs properly sandboxed.
Thanks to Enoki, we always double-check the third-party tools we're using and the external services we're using, making sure these all have the correct restrictions so nobody can access data that they shouldn't own.
Why an outside team
Wally's team wants to spend its time on accounting, not on tracking every new attack technique. Enoki keeps up with the risks and runs them against each release.
It's not because you're up to date today that you're up to date tomorrow.
More customer stories
How Tally red teams the agent that builds its forms
Enoki attacks our AI agent the way a real adversary would, then hands us a list of what broke with the exact steps to reproduce it.
Your first assessment
Attack your own agent before an adversary does.
Point Enoki at your agent's endpoint. It attacks it the way Wally's was attacked and shows you the exact steps behind every finding.