Field notes on replacing manual work.
Plain language, real constraints, written for the people who have to make it happen.
A WhatsApp assistant is a legitimate strategy for a West African business
For many businesses here the customer is on WhatsApp. A focused assistant that takes orders and answers questions can beat a flashy platform. What to build.
StrategyBuild, buy or co-build: what “we already have the pattern” actually saves you
Building from scratch is slow and risky. Buying rarely fits. A third road, starting from a working pattern and finishing it together, saves the costly part.
StrategyDigitised is not disrupted: the trap of software that only moved paper onto a screen
Moving a process from paper to software is progress, but the process stays the same. The organisations pulling ahead are redesigning the work itself.
PlaybookDocument extraction: the unglamorous use case with the fastest payback
Invoices, claims and forms arrive as PDFs, scans and photos, and someone retypes them. Reading them automatically is often the quickest first project to pay off.
PlaybookThe one-use-case rule: why “do AI” fails and “automate this one form” works
A vague mandate produces a demo. A specific, measurable job produces a result. How to choose your first use case and write it down so it can succeed.
ComplianceWhat Ghana’s Data Protection Act asks of you before you deploy AI
Register as a data controller, protect health data, plan for breaches, and let people challenge automated decisions. A plain guide to Ghana's Act 843.
StrategyWhy most AI pilots never ship, and what the ones that do have in common
A widely quoted study says most enterprise AI pilots return nothing. The number is debated, the causes are ordinary and fixable. Here is what to do differently.
Rather talk it through?
Tell us the job your team does by hand. We will tell you whether it is worth automating, and what a first step looks like.