← All work
Accenture / Internal tool
Teams chatbot
Say hi, pick a check, get a report.
A Python chatbot inside Microsoft Teams that joins three internal APIs, so 25+ engineers stop doing the same checks by hand.
25+engineers use it
75%less review
10 hsaved per cycle
35→5min per setup
Role
I built the tool, the bot flow and the API integration.
Stack
Python, SaltStack, Microsoft Teams API, SUSE Manager, AIOps APIs
The problem
Checks were compared by hand after every change. A batch of 30 servers took 40 to 50 minutes to verify, and setting a monitoring blackout took about 35 minutes.
What I built
- A Teams bot: say hi, pick a check, paste a list or select all, and get a results file back.
- Python that joins the AIOps, SUSE Manager and Teams APIs, authenticated with service accounts.
- Pre-check and post-check runs that compare automatically, cutting review 75%, about 10 hours per 550-server cycle.
- After tracing repeat alerts to a bad blackout setting, a Teams workflow that reads the time window from plain English and cuts setup from 35 minutes to 5.
What went wrong first
My first check engine had no limit on parallelism and caused a P2 overload. Batching based on CPU load and automatic retries fixed it. The 75% is an estimate from observed batch timings, not an instrumented measurement.
Accenture. Film is a reconstruction with sample data.