Agents can now write and file their own reports
Every run now ends with a short, plain-language report in the tool your team reads. Here is how we built it and what it changed for our pilots.
For a year, our agents did the work and left a log. Logs are great for auditors. They are terrible for the person who just wants to know if the job is done.
Why reports
When we asked pilot teams what they did after a run finished, almost everyone said the same thing: they opened the log, scrolled to the end and wrote a summary for someone else. The agent had done the job, and a person was still doing the paperwork.
So we taught agents to do that part too. At the end of every run, Kestrel now writes a short report: what it did, what it skipped and why, and anything that needs a decision.
“The first week, my team stopped asking me ‘did the vendor thing run?’ They just read the report in Slack.”
Maya Reyes, Northwind
How they work
Reports are built from the run itself, not guessed afterwards. Every step already records its inputs and outputs, so the report can point to the exact record it created or the exact rule that stopped it.
- Plain language first. Three to five lines, written for the person who asked.
- Links to everything. Each claim links to the step and the record behind it.
- Filed where you read. Slack, Teams, email, Notion or a Google Doc.
Add “report to #channel” at the end of any brief and Kestrel will file the report there automatically.
What changed
Across 40 pilot teams over six weeks, the time spent checking finished runs dropped by 71%. More surprising: teams approved twice as many agents for unattended schedules, because they could see clearly what each one had done.
What’s next
Next month, reports learn your format. Show Kestrel two reports you like and it will match the structure, tone and length for every agent on your team.