Business operating systems
We build the systems a business runs on.
Eight products for retail, hospitality, food and beverage and healthcare, with eTIMS, M-Pesa and WhatsApp built in. Every one of them speaks to the others, and to the AI agents that run on top of them, without an integration project.
The products
One for each part of the business. All of them connected.
Built to connect
Most business software cannot talk to the rest of the business.
A property system that cannot see the CRM. Stock control that cannot see the till. Each was bought separately, each integrates through nobody, and the AI project stalls because no system exposes anything an agent could use.
Ours are built the other way round. Every write in every system is published as a permissioned capability, and everything else — the web app, the till, the reporting layer, the agents — goes through that same contract. So the CRM can read the property system, and stock control can read the ledger. A new system joins that seam instead of starting an integration project.
- Reporting is not a separate purchase
- Every system reports on itself out of the box — operational reports, Excel exports and a finance view, with no BI tool bolted on afterwards.
- One question, many systems
- A graph layer reads across the whole estate, so a question that spans stock, sales and customers has somewhere to be asked.
- The interface an agent already needs
- Because the contract exists from the start, adding conversational AI is configuration rather than a rebuild.
Every product, one seam
Each product publishes what it can do as permissioned capabilities through one shared contract. The CRM, the graph layer and the AI agents all use that same contract, so each of them can reach every product without a separate integration.
AI-native by design
Most software has AI added later. Ours is built for it.
AI-native is a claim about architecture, so here is what it means in practice. The data model, the event stream and the permission system are designed so an agent can read the operation and act on it safely — with a record of what it did and the authority to do only that.
Then we build the agents. That is the difference between a vendor who sells you an ERP and leaves the AI question to somebody else, and one team that ships the operational system and the intelligence running on it.
- Business systems
- Agents watch stock and cash movement continuously and raise the exception the day it appears, instead of waiting for someone to run a report.
- Sector platforms
- Agents handle intake and triage — reading what arrives, classifying it, and escalating the cases that need a person.
- Analytics engineering
- Agents query the graph in plain language, so asking how two entities connect does not require someone who writes traversals.
- Conversational AI
- This is the customer-facing half of the agent layer: the same tools an internal agent uses, exposed to the person asking.
Where these run
Sectors we are already in production in.
Not a list of markets we would like to enter. These are operations running on systems we built. Anywhere a business tracks stock, staff, patients, claims or customers the same foundations apply — so if your sector is not named here, the answer is still probably yes.
- Imports, wholesale and retail
- Landed cost, multi-branch stock, and margin that still holds when the currency moves.
- Business systemsAnalytics engineeringConversational AI
- Manufacturing
- Production, materials and finished-goods stock tied to the same ledger as finance.
- Business systemsAnalytics engineeringConversational AI
- Hospitality
- Property management, recipe-level stock control, and the cost of a cover, nightly.
- Business systemsAnalytics engineeringConversational AI
- Healthcare
- One patient record across intake, wards, pharmacy and billing, with the audit trail intact.
- Sector platformsAnalytics engineeringConversational AI
- Insurance
- Claims intake, adjudication and settlement, with fraud surfaced by the network it hides in.
- Sector platformsAnalytics engineeringConversational AI
- NGOs and development
- Disbursement traceable to purpose, and beneficiary data checked against duplication.
- Sector platformsAnalytics engineeringConversational AI
How we engage
Three stages, in this order.
Understand
We start with the problem you actually have, not the technology we would enjoy building. That usually means sitting with the people doing the work and reading the data before proposing anything.
Build
Small releases you can see and correct, rather than a long silence followed by a reveal. You review working software throughout, not slide decks about working software.
Deploy and hand over
Into live operations, with documentation, a team that knows how the system behaves, and a clear account of its limits. The goal is a system that runs without us in the room.
Tell us what is not working.
Most useful first conversations start with a problem rather than a brief. Describe yours and we will tell you plainly whether this is work we should take on.