Frameworks · Cornerstone

Before you invest in AI, find the workflow that creates the most business value first.

This is the framework behind everything on this site: ten criteria for judging an AI idea, and a simple matrix that turns a pile of ideas into a sequenced roadmap.

Why this exists

The order you build in matters more than what you buy.

Most AI regret starts the same way: a tool got chosen before a workflow did. The fix is unglamorous — evaluate candidate workflows against consistent criteria, rank them, and let the ranking pick your first build. Thirty minutes of scoring routinely saves a quarter of wasted implementation.

The criteria

Ten questions to ask of every AI idea.

01
Business impact
If this workflow improved, would revenue, cost, or customer experience visibly move?
Impact anchors everything. An impressive AI demo on a workflow nobody feels is a science project, not an initiative.
02
Workflow frequency
How often does this workflow run — daily, weekly, quarterly?
Savings compound with repetition. A daily 20-minute task beats a quarterly 3-hour one by an order of magnitude.
03
Process clarity
Could someone write down the steps as they actually happen today?
AI amplifies the process it is given. If the process differs by person, document it first — that is free and often fixes half the pain.
04
Data availability
Is the information the workflow needs in systems AI can reach?
Inboxes and memories don't count. Reachable, structured-enough data is what separates a working system from a demo.
05
Implementation feasibility
Could a competent builder ship a first version in weeks, not quarters?
Prefer initiatives with boring integration paths. Every exotic dependency multiplies the timeline.
06
Risk level
What happens if the AI gets it wrong — an internal edit, or a customer-facing incident?
Start where errors are cheap and reviewable. High-stakes outputs need mature review loops you don't have yet.
07
Speed to validate
How quickly would you know whether it is working?
Weekly-visible results keep initiatives alive. If proof takes two quarters, momentum dies before the payoff arrives.
08
Strategic value
Does this build an asset — cleaner data, a reusable integration, a pattern you will repeat?
Some initiatives are worth slightly more friction because they make every later initiative cheaper.
09
Human review requirement
How much of the output can ship after a quick approval, versus needing rework?
The sweet spot is AI drafts that a person approves in seconds. If review takes as long as doing the work, the design is wrong.
10
Integration complexity
How many systems does this touch, and how well-documented are they?
Each additional system multiplies edge cases. First initiatives should touch one or two systems, not five.
The matrix

Value against feasibility — four honest buckets.

Average your value criteria (impact, frequency, strategic value, speed to validate) and your feasibility criteria (clarity, data, feasibility, risk, review, integration), then place each candidate:

business value →
High value · High feasibility
Quick Wins

Start here. High-impact workflows you could ship in weeks. One of these becomes your first initiative.

High value · Low feasibility
Strategic Bets

Worth doing, not first. Sequence them after a quick win has built skills, trust, and cleaner data.

Low value · High feasibility
Low-Value Experiments

Easy but pointless. Fine as learning exercises; dangerous as your definition of "doing AI".

Low value · Low feasibility
Avoid for Now

Expensive and unimportant. Revisit only if the business changes shape.

feasibility →
How to run it

Four steps, one afternoon.

Step 01
List candidates
Write down every workflow that could plausibly use AI. Aim for ten or more — the long list is the point.
Step 02
Score with the criteria
Rate each candidate 1–5 on the ten criteria (use the AI Use Case Scorecard). Half an hour, honestly spent.
Step 03
Place on the matrix
Value on one axis, feasibility on the other. The quadrants do the arguing for you.
Step 04
Commit to one
Take the top Quick Win and give it a quarter. Park everything else on a written roadmap.
FAQ

Common questions

Why prioritize before building anything with AI?

Because the most expensive AI mistake is building the wrong thing well. Prioritization is how you find the workflow where AI creates real business value first — before you spend money on tools or implementation.

How many AI initiatives should a business run at once?

One, for most SMEs. A single initiative shipped into real use builds skills, data, and trust that make every later initiative cheaper. Parallel experiments usually starve each other.

What if every candidate scores badly?

That is a useful result: it usually means process clarity or data availability is the real bottleneck. Fix that first — the AI Readiness Check will tell you if that is where you are.