Live, my venture
Flowstack AI
Helping small businesses hand repetitive admin to AI they can check and trust, so their people can focus on customers.
- My role
- Founder and CEO
- Timeline
- 2024 to now
- Team
- 9 people
- Links
- —
Tasks done by AI
142
Waiting for your check
6
Hours saved this week
18.5
Needs your approval
Invoice reminder to Ola Foods, ₦240,000 overdueApprove
Reply to customer about delayed deliveryApprove
Weekly sales summary for the teamSent
- Businesses using it
- 300
- Hours saved per team, weekly
- 18
- Customer rating
- 4.8/5
- Tasks checked by a human
- 100%
The problem
Small business owners told us the same thing: they spend their evenings chasing invoices, answering the same customer emails and pulling together reports. They had heard AI could help, but they didn’t trust a robot to talk to their customers.
So the real problem wasn’t automation. It was trust.
How we made it trustworthy
You approve firstNothing reaches a customer until a person on the team approves it.
A clear recordEvery action shows what the AI did, why, and who approved it.
Private by defaultBusiness data is never used to train outside models.
“I get my evenings back. Flowstack drafts the boring stuff, I check it over coffee, and my customers get faster answers than ever.”
What I did
- Interviewed 60 business owners before writing any code.
- Designed the approve-first experience and led the build of the product.
- Hired and lead a team of nine across product, engineering and support.
- Raised a pre-seed round from Orbit Ventures.
Under the hood
For technical readers: language models paired with rule-based checks, running on a Kubernetes platform on AWS.
PythonLLM APIsRetrievalAWSKubernetesPostgreSQL