A founder in Lagos told us he spent ₦2.3M on AI tools in 2025.
He had a customer service chatbot. A content generation tool. A data analytics platform. A sales forecasting model.
He was using exactly one of them a year later.
This is not unusual. The AI buying frenzy is real. Nigerian businesses are spending millions on tools they don't need, for problems they haven't defined.
The fix is simple. Before you buy any AI tool, answer one question:
What specific problem am I solving?
Not "improving efficiency." Not "staying competitive." A specific, measurable, observable problem. Let's build a framework around this.
The 5-question AI decision framework
Use this before every AI purchase. Write the answers down. If you can't answer all five, don't buy yet.
- What process is causing measurable pain today? Name the exact process. "Customer service" is too broad. "Our team spends 12 hours per week manually answering the same 5 questions about delivery status" is specific. Assign a weekly time cost to it.
- What does success look like numerically? "Faster response time" is not a target. "Reduce response time from 4 hours to under 5 minutes" is. Write the before number and the after number. You'll need the before number later.
- What data does this tool need to work? Most AI tools require clean, structured, historical data. Do you have it? How far back? How clean is it? If you need to prepare data first, add that cost to your budget.
- Who will use this daily? The person who will actually operate the tool — not the person who approved the purchase. Will they use it? Can they use it? Does it fit their actual workflow?
- What happens when this tool fails? AI tools fail. Models drift. APIs go down. What's your fallback? If the answer is "we'll figure it out," your dependency risk is too high for critical processes.
How the framework works in practice
Let's apply it to a real scenario.
Scenario: A logistics company in Abuja wants to buy an AI-powered route optimisation tool.
Question 1 — The pain: Currently, their dispatcher spends 3 hours every morning manually planning routes for 25 delivery vehicles. On bad days (traffic, road closures, last-minute orders), it takes 5 hours. Total weekly cost: 20 hours of dispatcher time at ₦12,500/hour = ₦250,000/week in routing labour, plus ₦180,000/week in excess fuel from suboptimal routes.
Question 2 — The target: Reduce routing time to 30 minutes. Reduce fuel costs by 15%.
Question 3 — The data: The tool needs historical delivery locations, traffic patterns, and vehicle capacity data. They have delivery addresses in a spreadsheet but no structured traffic data. They'd need 2 weeks to clean the address data. Traffic data would come from the tool's native integrations.
Question 4 — The user: The primary user is the dispatcher — a 50-year-old man who has been doing this job for 12 years. He doesn't use computers beyond WhatsApp. The tool needs to be usable on a phone with voice commands, or they need a translator.
Question 5 — The fallback: If the tool goes down, they revert to manual routing. That's acceptable — they have the spreadsheets.
Decision: Buy, but with conditions — 2-week data cleanup phase, 1-week training for the dispatcher, and a 30-day trial to confirm fuel savings.
When to buy vs. when to wait
Buy now
You can answer all 5 questions clearly. You have the data. You have the user. You have a fallback. You know how you'll measure success. The tool solves a specific, painful problem that costs you more per month than the tool costs.
Wait
You can't answer question 1 or 2. You think AI would be useful but you can't name the specific process or the current cost. Put the budget into an operations audit first. Find the real pain, then look for tools.
Build instead
You can answer all 5 questions, but the available tools don't fit your specific workflow, data structure, or local context. Nigerian businesses often find that "off-the-shelf" AI tools are built for American or European operations. A custom solution, while more expensive upfront (₦300K-₦800K), often delivers higher ROI because it fits exactly.
The ₦500K question
Here's a useful heuristic: If you can't describe the problem in one sentence that includes a number, you're not ready to buy AI.
Bad: "We need AI for customer service."
Good: "We spend 25 hours per week answering the same 10 questions, and 15% of inbound leads go unanswered within 24 hours."
The first sentence will cost you ₦500K+ chasing the wrong solution. The second sentence tells you exactly what to buy — a system that routes and answers common questions automatically.
AI is a tool, not a strategy. A hammer doesn't decide which nails to hit. You do.
Meshgryd Systems helps Nigerian businesses evaluate, select, and implement AI tools that match their actual operations — not their feature checklists. We start with your problem, not the technology. Start a conversation →