← Back to Plan
2

Stage 2: Plan

Find the Real Opportunities

Not every process should be automated. Some are better left alone. Some need AI. Some just need better documentation. Knowing which is which is the hard part.

The Prioritisation Problem

Once you start looking, you'll find AI opportunities everywhere. That's the problem. You can't do everything. Resources are finite. Attention spans are shorter.

The temptation is to start with the exciting stuff - the demos that impress, the use cases that sound futuristic. But exciting doesn't mean valuable. And valuable doesn't mean achievable.

The best opportunities are often boring: high-volume, repetitive, well-understood processes where AI can have immediate, measurable impact.

Common Traps

The Hype Problem

Someone saw a demo. Someone read an article. Now there's pressure to "do AI" without clarity on what problem it solves. You end up with solutions looking for problems.

How to handle it:

Start with pain, not technology. "What's costing us time or money?" comes before "Where can we use AI?" If you can't articulate the problem in business terms, it's not ready for AI.

The Pet Project

A senior leader has a favourite idea. It's politically difficult to say no, even when the ROI doesn't stack up. Resources get allocated to the wrong things.

How to handle it:

Create a scoring framework before evaluating projects. Effort vs impact, data readiness, risk. Let the framework say no, not you.

Shiny Object Syndrome

New AI capabilities emerge weekly. GPT-5 just dropped. Claude can do X now. The temptation to chase the latest thing is constant. Focus fragments.

How to handle it:

Pick boring problems. High volume, repetitive, clearly defined. The exciting use cases make good demos but rarely survive contact with reality.

The Budget Conversation

AI projects need investment - tools, training, time. But budgets are set. Business cases require numbers you don't have yet. Chicken and egg.

How to handle it:

Start with experiments that don't need budget approval. Free tiers, existing tools, volunteer time. Prove value small before asking for money big.

A Simple Evaluation Framework

Before greenlighting any AI project, ask these questions:

Criteria
Question
Why It Matters
Volume
How often does this happen?
AI needs repetition to justify investment. A process that happens once a month isn't worth automating.
Consistency
Is the process standardised?
Chaotic processes need to be fixed before they can be automated. AI amplifies mess.
Data
Do we have the inputs AI needs?
No data, no AI. If the information lives in people's heads or scattered emails, start there.
Stakes
What happens if it goes wrong?
Start with low-stakes processes. Getting AI wrong on customer billing is worse than getting it wrong on internal reports.
Measurability
Can we measure improvement?
If you can't measure before and after, you can't prove value. Pick processes with clear metrics.

Good vs Bad Candidates

Good Starting Points

  • +Data entry and transfer between systems
  • +Report generation from existing data
  • +First-pass document review and summarisation
  • +Customer query triage and routing
  • +Meeting notes and action item extraction
  • +Research compilation from multiple sources

Proceed With Caution

  • -Anything requiring nuanced human judgement
  • -High-stakes decisions without human oversight
  • -Processes that change frequently
  • -Tasks with poor or inconsistent data
  • -Politically sensitive areas
  • -Things that need to be fixed, not automated

Not Sure Where to Start?

I can help you evaluate your options and prioritise the opportunities that will actually move the needle.

Book a Discovery Call