AI and automation

Most businesses do not have a staffing problem. They have a process that requires people to retype things.

An engineer building an automation workflow on a large monitor

What we do

Process mapping

Before anything is built, we document how work actually moves, including the undocumented workarounds. Automating a broken process just breaks it faster.

Workflow automation

Invoicing, payment reminders, reconciliation, onboarding, approvals and recurring reports, built to run on schedule or on trigger with a record of every run.

Agentic automation

AI agents that handle multi-step work with judgement inside defined limits, using the tools they are given and escalating to a person at the edge of their authority.

Finance process automation

Transaction capture, bank and mobile money reconciliation and document matching, with an audit trail that exists by design rather than being rebuilt at year end.

Customer conversation systems

WhatsApp and web assistants that answer, qualify, take orders and collect payment, delivered through Kasa-bot, our flagship product.

Systems integration

Connecting accounting software, CRM, payment gateways, spreadsheets and messaging so data moves once instead of being re-entered at every step.

Automation built to survive an audit

This practice sits next to an audit practice on purpose. We are the people who will one day have to examine systems like the ones we build, so we build them to be examined.

Every workflow logs what it did, keeps the source document attached to the entry, and puts a person at the points where money moves. We are deliberately conservative about where an AI agent is allowed to act without a human checkpoint.

How a build runs

Map

Document the current process and decide what should be automated, and what should not.

Design

Write the flows, rules, approval points and failure handling, and agree them with you before building.

Build and test

Develop and test against your real, messy data, not clean samples.

Launch and tune

Go live on a narrow scope, monitor, adjust, then widen. Documentation and training included.

What we build on

Self-hosted where it matters, so your data and your customer records stay under your control.

n8n
Evolution API
Groq
Airtable
Redis
Docker
Mobile money and card gateways
Accounting software integrations
Dedicated VPS hosting

Common questions

Do we need to replace our existing software?

Usually not. Most of our work connects the tools a business already uses so they stop depending on people copying data between them.

Where is our data held?

For sensitive workloads we run the stack on dedicated infrastructure rather than a shared platform, so data stays under your control. We agree data handling in writing before any build.

What happens when an automation fails?

Every workflow is built to log failures and alert a person. A silent failure is the most expensive kind, so we design specifically to avoid it.

How long does a typical build take?

A first process is often live within two to four weeks. Larger programmes are delivered in stages so value arrives early.