This page is a list of all the tools on the site, like a box of tools with labels. Each label says what the tool is for, so you can pick the right one.
The site has grown into a set of small methods: templates, scales, rules and checks. This page lists them in the order I'd use them on a real decision, from mapping the work to reviewing the result, with a one-line summary and a link to each.
Index of methods by stage: work mapping, decision framing, quantification, risk, cost, capture and review. Each links to a page with purpose, inputs, steps, a worked example and known limits. Use as a checklist when designing AI support for a decision.
How to use this page
Each row is one method. They're grouped by stage, in the order I'd use them on a real decision. You don't need all of them. For a small decision, the framing questions and a decision record may be enough. For an agent that will make thousands of decisions a month, most of the list applies.
How each method page is laid out is explained in How the method library works.
Map the work
| Method | What it does |
|---|---|
| Map the work, then the decisions | Five steps from a public job profile to a list of decisions |
| Reading an O*NET occupation | How to use O*NET as a starting task list |
| ESCO occupations and skills in practice | The European equivalent, and how to use it |
| The decision inventory template | One row per decision with volume, time, stakes and reversibility |
| Frequency, stakes, reversibility: a scoring guide | Shared 1 to 5 scales so people score alike |
| What your approval logs already know | Volume, wait time and consistency from approval records |
| Ticket logs as a map of decisions | Change rates as clues to where decisions go wrong |
Frame the decision
| Method | What it does |
|---|---|
| Five questions before any decision | What, why, when, how and who, before anything else |
| Save money, make money, look after people | Names the goal so success can be measured |
| Four jobs AI tools do for a decision | Matches the part of a decision to the right kind of tool |
Put numbers on it
| Method | What it does |
|---|---|
| Expected value on one page | Compares options by probability times value |
| When is more information worth paying for? | Decides whether to gather more before choosing |
| Setting a threshold you can defend | Cost of a wrong yes ÷ (wrong yes + wrong no) |
| Act, ask a person, or stop | Three routes for every case a model scores |
| Calibration: are your 80% calls right 80% of the time? | Checks whether stated confidence matches results |
Check the risk
| Method | What it does |
|---|---|
| Start a risk register from the decision inventory | Five questions per decision row |
| Five risk types | Data, decision, people, supplier, cost |
| Mapping risk to work activities | A grid of activities against risk types |
| Likelihood, impact and reversibility on one sheet | Priority score that lifts hard-to-undo, slow-to-spot errors |
| Controls people actually keep | Four tests for a lasting control |
| Review dates that happen | Calendar dates plus event triggers |
| A nudge is a small decision too | Sets alert limits with the threshold method |
Check the cost
| Method | What it does |
|---|---|
| Cost per decision | The unit cost to track first |
| Forecasting Copilot Credits from decision volumes | Volume × credits per decision |
| Forecasting AI spend as a range | Low, likely and high cases |
| Seat licence or Copilot Credits: the break-even | When a licence beats pay-as-you-go |
| The 20-minute weekly review agenda | A short, regular look at spend and value |
Record and review
| Method | What it does |
|---|---|
| The decision record schema | Ten fields written at the time of deciding |
| A 30-second capture in Teams | The five fields that can't be added later |
| Outcome bias | Judges the decision separately from the luck |
A worked example
This example is illustrative. A team wants an agent to suggest whether supplier invoices can be paid without a query. Using the list: the inventory shows 900 invoices a month at 4 minutes each; the scoring guide rates stakes 3 and reversibility 2 with a three-day payment window; the threshold method gives a cut-off from the cost of a wrong payment and a wrong query; the risk sheet puts it mid-table; the credit forecast puts it at a few hundred pounds a month; and every change to the threshold gets a 30-second capture.
What I'm still checking
Some methods overlap, especially the risk register and the risk map. I'm keeping both for now because they answer different questions, but I'd like feedback on whether the list is easier to use shorter.
Sources
This page indexes methods on this site and uses no external facts or figures.