Intervention families
Where AI earns its place
Organised by the four performance layers the diagnostic scores — not by product pillars. These are the families the interventions come from; which one is yours is a question the scan answers, not this page.
How to read this page
This is not a menu. Browsing intervention families and picking the appealing one is precisely the “where can we use AI?” mistake. The page exists so you can see the shape of the work — the selection happens at the AI Priority Alignment Session, against your own diagnostic result.
What you choose to sell, to whom, at what price
The layer where the largest value is won or lost before anyone executes anything. AI is useful here because the decisions are data-rich, repeated, and currently made on partial evidence.
5 levers · 21 initiatives · Are we pointing the company at the right profit, and charging for it properly?
Classification · anomaly detection · decision support
Price and margin leakage detection
Headline pricing looks disciplined, but realised margin does not match it. The leakage is spread across thousands of individual discounts, rebates, freight terms and credit notes that nobody can inspect at line level.
- L03 Proposition & Pricing
- L04 Price Realisation & Terms
- L02 Customer & Segment Strategy
Judged on
- Realised price vs list
- Margin per order
- Value of exceptions recovered
- Time from leak to correction
Input · transaction and contract records
- Invoice and credit-note lines
- Contract and rebate terms
- Discount approvals
- Cost-to-serve data
Output · price waterfall with exceptions ranked
Realised margin gap
Ranked by recoverable value, with the accountable owner attached
- Off-contract discount
- Freight absorbed
- Rebate over-accrual
Every exception traceable to its source document
Quick wins this family produces
- 01
Automated price-waterfall rebuild from invoice lines — the leakage becomes visible without a data-warehouse programme
- 02
Exception scoring that routes only the recoverable cases to a named owner, instead of a 4,000-row report nobody opens
- 03
Contract-term extraction so non-standard terms are found at renewal rather than discovered at audit
Quick wins like these reach you at the AI Priority Alignment Session, already matched to your scan result and tested against the three gates.
Worked example
A worked example built from the initiative catalogue to show the shape of the work. It is not a client case study — APCG does not publish anonymised client stories.
Distributor with strong list pricing and weak realisation
The challenge
Gross margin had drifted two points over three years with no change in list price or input cost. Finance could see the drift in aggregate but could not attribute it. Commercial teams disputed every explanation because no one could point at line-level evidence.
What would be delivered
- Line-level price waterfall rebuilt from source invoices and credit notes
- Exception classifier that separates deliberate commercial concessions from process errors
- Weekly ranked queue routed to the account owner with the supporting document attached
What changes
- The margin gap becomes attributable rather than disputed
- Effort concentrates on the recoverable minority of exceptions
- Renewal conversations start from evidence instead of assertion
Which of these is yours?
That is what the diagnostic is for. Free, about fifteen minutes, and pointed at the one thing this choice depends on: where your business is losing the most value.