Group
MAPFORTH Use-Case Scorer
Score the work, not the story. Rank the backlog. Read each case onto the operating scorecard.
Step 1 · Describe the use case
Who is the user, in what workflow, seeking what outcome, working with what inputs, and what role does AI play? Write in operational language. Avoid "AI to help finance."
Step 2 · Score MAPFORTH
Pick the description that fits. Higher always means a better AI candidate. The guidance and the fix update as you choose.
Export writes the current backlog here (copy it and keep it anywhere). Import reads what you paste, or choose a file.
Portfolio map · impact against lift
Impact (vertical): rule tightness, task variability, output burden, frequency & repetition. Lift (horizontal): messiness, accountability risk, power required, human-in-the-loop; 5 means light lift. Dot size follows the combined score. Hover for detail; click to reopen.
How to read the map
Every MAPFORTH score keeps 5 as good and 1 as poor, so both axes read the same way: a 5 on Messiness means clean data and light lift. Both axes run 1–5; the quadrant lines sit at 3. The best use cases land top right.
Ranked backlog · scores and initiative descriptions
| Rank | Use case & initiative | Want | Can | Should | Combined | Impact | Lift 5 = light | Decision | Bottleneck & fix | First mover |
|---|
The operating scorecard · four questions
Each saved use case is placed in the quadrant where its MAPFORTH profile predicts the first measurable movement. A balanced portfolio lights all four. A use case reports in all four columns, or it does not scale.
What MAPFORTH is
A prioritization discipline for AI use cases. Eight dimensions describe the work itself. Six (MAPFOR) say how well AI suits the task. Two (TH) say whether the organization can and should apply it. It converts enthusiasm into a defensible, force-ranked backlog.
The eight dimensions
| M | Messiness | How structured and clean are the data and process? |
| A | Accountability risk | What happens if the AI is wrong? |
| P | Power required | How resource-intensive is the work? |
| F | Frequency & repetition | How often does it repeat? |
| O | Output burden | How emotionally or socially loaded is the output? |
| R | Rule tightness | How explicit and stable are the rules? |
| T | Task variability | How predictable is the task? (CAN) |
| H | Human-in-the-loop | How much human judgment is required? (SHOULD) |
Scoring
- Want (desirability) = sum of the six MAPFOR scores ÷ 30.
- Can (feasibility) = task-variability score ÷ 5.
- Should (AI viability) = human-in-the-loop score ÷ 5.
- Combined = Want + Can + Should, maximum 3.00. The backlog ranks on it.
- Bottleneck = the lowest MAPFOR score. Its "fix needed" is the first improvement to attempt.
- Impact = mean of rule tightness, task variability, output burden and frequency & repetition (1–5).
- Lift = mean of messiness, accountability risk, power required and human-in-the-loop (1–5). Scores keep the worksheet's direction: 5 is light lift, 1 is heavy. Strong use cases land top right on the portfolio map.
Fixes for the CAN and SHOULD gates
The worksheet gives fixes for the six MAPFOR dimensions and none for task variability or human-in-the-loop. The scorer carries a fix for every level of both; they appear under each option as you choose it, and in the backlog's bottleneck column when they bind. Clarity Group's additions.
Decision category
- Automate — AI suitability (mean of six MAPFOR) ≥ 4.0, human stake low (score ≥ 4) and variability low (score ≥ 4). AI takes over end-to-end.
- Augment — AI suitability ≥ 3.5, but human stake or variability is high. AI does the heavy lifting; humans decide.
- Assist — AI suitability below 3.5. AI supports only; human-led.
The worksheet gives the categories as "High AIS 4–5" and "Low AIS 1–3." This tool uses 3.5 as the dividing line between them.
How it reads onto the scorecard
Each dimension feeds a primary scorecard quadrant (weight 1.0) and a secondary one (weight 0.5). The quadrant with the highest weighted score, scaled to its maximum, is where management should expect the first measurable movement, and which board lens will read it.
| Primary → secondary | KPI it predicts | What the board reads |
|---|
How to improve a use case
If a use case scores poorly, ask: Can we clean the data? Can we simplify the rules? Can we reduce variability? Should AI assist instead of automate? Do we redesign the workflow before applying AI? Sometimes the right answer is to pivot, constrain, defer, or redesign the work itself.
What it prevents
- Starting with technology instead of problems
- Political prioritization
- Thin ROI measurement
- Pilot enthusiasm without readiness
- Scaling messiness
Provenance
MAPFORTH dimension scales after Tech Mahindra, as taught in the Chicago Booth CAIO program (Week 5). Want–Can–Should and force ranking follow the MAPFORTH worksheet. Balanced scorecard after Kaplan & Norton. Scorecard mapping and decision thresholds are Clarity Group's.
Your backlog is kept in this browser only. Export JSON to keep or share it.