For a Kenyan SME considering AI, the first project should make a specific business task easier to complete. Start with the work already happening in your organization: finding information, handling incoming requests, preparing documents, or transferring data between tools.
I am Paul Waweru, a Nairobi-based AI consultant and software engineer working with local and global teams through Artebello. My practice connects readiness assessment, implementation, and ongoing improvement. Geography informs the delivery conversation; the workflow determines the architecture.
Bring a workflow rather than a technology shopping list
Show how the task starts, what staff do, which systems they consult, and what counts as completion. Describe the exception that causes the most disruption. If you do not yet have reliable measurements, say so and establish a baseline during discovery.
Artebello's readiness approach examines business, people, process, data, technology, and governance together. The aim is to identify a useful next decision, rather than produce a score without an implementation path.
Check the actual tools the business uses
An organization may work across email, spreadsheets, mobile applications, messaging, accounting software, and paper records. Discovery should establish the real information flow rather than assuming that all important work lives in one database.
Ask which systems provide supported APIs, who owns the accounts, whether test access is available, and which actions are permitted. Mention a required payment or messaging integration early so its access and provider requirements can be evaluated.
The presence of a familiar tool name in a proposal does not establish implementation experience or authorized access. This guide does not claim that every Kenyan business uses the same systems or that my published projects implement every local provider integration.
A fictional first-project brief
Imagine a Nairobi service business preparing quotations from incoming customer requests. An initial pilot could organize the request, identify missing details, and prepare a draft using approved service information. A staff member remains responsible for the final quotation.
The useful discovery questions are practical: who approves the service information, which fields affect the quotation, where the draft is stored, and how a duplicate request is recognized. An automated customer message is a separate action requiring its own boundary.
This example is illustrative. It is not a client case study or a claim about a measured saving.
Plan the operating conditions explicitly
Discuss devices, connection interruptions, working languages, and what staff do when a system is unavailable. Test those conditions for the intended users. Do not assume that an office demonstration establishes reliability for every location or device.
A mobile workflow may need a clear retry and saved-task state. A knowledge assistant may need a named document owner. A tool-using agent may need an approval queue that staff can actually monitor.
Civicore contributes experience with a live operational platform across web, backend, and mobile. Its portfolio record describes software scope; it does not support claims about election outcomes or adoption metrics.
Agree how delivery will be reviewed
Scroll sideways to view the full table.
| Before implementation | At the pilot review |
|---|---|
| A named workflow owner | Whether the task was completed usefully |
| Required sources and integration access | Which source or tool failures occurred |
| Approval and escalation rules | Whether staff could review and recover |
| Baseline with its limitations | Observed results and remaining uncertainty |
| Defined first-release scope | A decision to improve, expand, or stop |
For remote collaboration, also agree decision owners, review availability, and how demonstrations and issues will be shared. The same discipline helps a Nairobi client and a global startup.
Choose the first engagement
If the opportunity is unclear, begin with AI strategy and readiness. If a workflow and its access boundaries are known, AI automation consulting can establish an implementation scope. If the surrounding application needs work, software product engineering covers that foundation.
You can discuss a project with a short description of the task, current systems, and where you want help. That is enough to begin a useful consulting conversation.
Prepared with AI assistance using Paul’s documented project work and the linked sources. Examples are illustrative unless identified as project records.