AI opportunity map
Sorts candidate uses of AI by the value they create and whether the business could actually run them, so the conversation stops being a list of tools. Most of the value sits in high-volume, low-judgement work nobody finds interesting.
The printable canvas and the handoff into Claude Code are part of a paid plan. See what a plan includes. It needs the Claude desktop app on this machine, and your team and the prompt library already installed in that project — we cannot see your disk, so open it there.
What it is
A structured assessment of where machine capability changes the economics of one specific business, rather than where it is technically impressive. Each candidate is placed on value created and on feasibility today, and feasibility is dominated by data availability and process consistency rather than by anything about the model.
- The task inventory
- Where time and cost actually go, by volume and by judgement required. High volume with low judgement is where value concentrates, and it is rarely the part anyone wants to talk about.
- Value
- What changes if this works — cost removed, revenue enabled, cycle time cut — stated as a number. Candidates without a number lose to candidates with one, regardless of merit.
- Feasibility
- Data availability, process consistency, integration cost, and whether anybody owns the outcome. This is where most candidates die and it has almost nothing to do with the model.
- The judgement line
- What the machine decides and what a person decides. Drawing it explicitly is the difference between a deployed system and a permanent pilot.
- Risk and governance
- Where a wrong output causes real harm, what the regulatory position is, and who signs it off. Cheap to think about now and very expensive to retrofit.
- The adoption question
- Whose job changes, and what they get out of it. Tools that make somebody's day worse do not get used, however good they are.
How you run it
- Inventory the work, not the toolsWhere the time goes, at what volume, needing how much judgement. The opportunity list falls out of this, and it looks nothing like a vendor's.
- Put a number on value before scoring anythingCost removed or revenue enabled, however rough. An unquantified candidate loses every prioritisation to a quantified one, whatever its merits.
- Score feasibility on data and process, not the modelWhether the data exists, is accessible and is consistent enough to use. This decides the outcome, and it is precisely what a pilot skips.
- Draw the judgement line for each candidateWhat the system decides, what a person decides, and what happens when they disagree. Left unwritten, this becomes the reason it never leaves pilot.
- Pick few, and record why the rest are outTwo or three, resourced properly. A portfolio of twelve pilots is a way of not choosing, and it is how organisations spend two years learning nothing.
The prompt
Run this tool in your own Claude
The short prompt starts your partner against the library on your disk. The long one carries everything with it and needs nothing installed.
Your playbook
It lands in the earliest stage this tool suits. Move it on the Playbook page.
You’ll need
- The processes that consume the most time or cost, by volume
- An honest read of data quality and where the data actually lives
- Who would have to change how they work, and whether they would
You’ll end up with
- Candidates scored on value created and feasibility today
- A short list worth doing, and a named reason for everything cut
- What has to be true — data, skills, governance — before the shortlist can run