For 2 years now, we've met dozens of teams that want to 'do AI'. In 70% of cases, the project they want to launch first is the wrong one. Here's the matrix we use to help them prioritise.
The shiny object trap
Demos of autonomous agents that do everything are impressive. That's why teams want to build one. But the first question should never be 'which AI agent to build'. It should be: 'which business process is costing me the most in repetitive time and human errors?'
We saw one team wanting to build an autonomous sales agent before they even had a properly maintained CRM. That's putting the cart before the horse.
Our 4-quadrant matrix
We assess each potential use case on two axes: process frequency (how many times per day/week/month), and complexity of required human judgment.
- High frequency + low complexity = simple automation (rules, scripts). AI is overkill.
- High frequency + medium complexity = SWEET SPOT for AI. This is where we deploy first.
- Low frequency + high complexity = human training. AI doesn't bring enough ROI.
- High frequency + high complexity = augmented AI (human validates). High ROI but needs more guardrails.
The hidden criterion: measurability
A use case you can't measure shouldn't be automated first. If you can't say 'before AI we handled X tickets per day with a Y CSAT', you won't be able to say if AI improves or degrades the situation.
Start with a process where you have 3 months of measured baseline. The ROI will be provable, and you'll avoid subjective debates.
AI is a tool, not a strategy. Before launching a project, ask the right questions about frequency, complexity, and measurability. If you want us to do this work with you, we have a 2-day workshop designed exactly for that.