Agents and workflows, built around how your team already works.
Teams can start with training and continue into a build.
Workflow map
- A document arrives
- Agent drafts the reply
- A person checks it
- Filed and sent
What we can build
Four kinds of build.
Agents that draft
from your documents
Letters, reports and claims, drafted from your records.
Document intake
from a phone
A receipt photographed on site, read, checked and filed.
A knowledge base an agent can work over
Documents organised so an agent answers from your records.
An agent your
team can message
On WhatsApp or Telegram, answering from your knowledge base.
How a project runs
One workflow at a time, with a named owner on your side. The next step starts on your decision.
Map
We map one workflow with the people who do it.
You keep
The written map of the process.
Build
Built around the process as it is, changed only where broken.
You keep
The skills and agents, with their instructions.
Connect
Documents organised first, then connected to your data and tools.
You keep
Your documents, organised and named.
Hand over
Your team runs it. A named person checks the output.
You keep
The written procedures.
How much does AI automation cost in Kenya?
Every project is quoted after the Map step. There is no public price list.
- Covers
- One workflow,
Map to Hand over. - First step
- Map, short and on its own.
- Form
- A written quote.
Built by a team we trained
Two inspection applications, from an empty folder, in three mornings.
The programme ran as training. After the foundations, participants built two inspection applications with our team in the room. 16 participant accounts merged pull requests into the main branch; the applications were theirs from the first commit.
- Pull requests merged
- 137 in all; 130 in the three mornings, 91 of them opened by participants
- Sector
- A national utility
- Format
- Five mornings on site
- People
- About 30 on day one
- When
- September 2026
Questions about a build
Which AI tools do you build with?
The ones your team already runs. The pattern works in Claude, ChatGPT or another tool.
What happens to our data?
The agent works over your own records, organised first. A person checks its output.
What do we need to provide?
The people who do the workflow, their documents, and one named owner on your side.
Start with one workflow.
Tell us about the work your team repeats most.