In short
Start the map with outcomes about people
- Start with outcomes about people. An AI inventory can show you the chatbot and miss the scorer.
- Four columns are enough: workflow, person affected, software or vendor, and personal information used.
- Ninety minutes produces a first map for legal and the board. It does not audit every system.
- Mark Unsure as Unsure. A blank reads as a no, and an unchecked no can leave a decision out of the policy.
Detail
Map with the people who run the work, in 4 columns
Many organisations start this work in the wrong place. They ask IT for the AI register and get a list of copilots, licences and experiments. None of that shows whether software already decides who gets a shift, a refund, a shortlist or a hold on their account.
Sit down with the people who run the work: hiring, credit or collections, claims, customer service, fraud, rostering and customer pricing. Ask each area where a computer program recommends, ranks, scores, flags, routes or completes an outcome about a person, using information about that person. Include pilots that run on real cases, because they can already be in scope.
Write 4 columns. The first is the workflow, in the words the team uses. The second is the person affected, such as a candidate, customer, claimant, staff member or account holder. The third is the software or vendor, even if the name is embarrassing. The fourth is the personal information used, described in kinds such as identity, financial, behavioural, employment and location data. If nobody in the room can fill a cell, write Unsure. That cell is now a work item.
Then test each workflow against APP 1.7. A workflow stays on the map while any part of the test below is met or Unsure.
APP 1.7 and APP 1.8 apply from 10 December 2026. They apply where an APP entity has arranged for a computer program to do either of 2 things:
- make a decision
- do a thing that is substantially and directly related to making a decision.
An APP entity is an organisation or agency covered by the Privacy Act 1988. Both of these conditions must also be met:
- the decision could reasonably be expected to significantly affect the rights or interests of an individual
- personal information about the individual is used in the operation of the computer program to make the decision or do the thing that is substantially and directly related to making the decision.
The privacy policy must then describe 3 things:
- the kinds of personal information used in the operation of such computer programs
- the kinds of such decisions made solely by the operation of such computer programs
- the kinds of such decisions for which a thing, that is substantially and directly related to making the decision, is done by the operation of such computer programs.
Ninety minutes is enough for a first list across the obvious functions. It is not enough to inspect every configuration. Stop when you can brief privacy counsel with named workflows, in place of a feeling that the organisation probably has some AI. The APP 1.7 Rapid Review is the next pass: it tests likely exposure and turns the list into a decision map. These 90 minutes make sure that pass starts from facts.
Commercial impact
APP 1.7 drafting fails when operations cannot name decisions
Privacy policy drafting for APP 1.7 fails when the legal team must describe kinds of decisions that operations cannot name.
A 90-minute map also prevents 2 expensive mistakes. The first is treating every generative tool as in scope and spending months on the wrong systems. The second is assuming that a human in the loop means there is nothing to disclose, while staff accept ranked lists they have not read.
Boards do not need a 40-page register from this session. They need to see whether the organisation can produce a list at all. If 90 minutes with the people who run the work cannot produce one, the board has an ownership problem to resolve.
In practice
Illustrative cases of decisions found in mapping sessions
Illustrative composite: in a 90-minute session with an operations director and a people and culture lead, the live list had 3 items. They were recruitment ranking, rostering and a performance dashboard that fed bonus conversations. The generative writing assistant did not make the list. The decisions often sit in tools nobody demonstrates.
Illustrative composite: a claims manager said the system suggested a settlement range that staff could override. The useful follow-up asked how often staff override it and whether they must give a reason. The workflow stayed on the map. Staff review does not take a workflow out of scope where the program's output is still a key factor in the decision.
Illustrative composite: a finance team listed a collections scorer it had forgotten was still running. It used account history and contactability flags. Nobody had included it in the AI steering pack, because staff thought of it as the old bureau tool. APP 1.7 applies whatever the tool is called.
Illustrative composite: when a cell stayed Unsure after 90 minutes, it stayed Unsure, and an owner was assigned to confirm it within a week.
What to do
Steps to produce a decision list privacy counsel can use
- Map outcomes about people first, then name the software, so the AI register shows the actual decisions.
- Use 4 columns and allow Unsure. Record a gap in knowledge as Unsure.
- Treat human approval as a fact to test. It does not automatically exclude a workflow.
- Stop at a list that privacy counsel can use, then examine the high-impact workflows in more depth.
- Commission an APP 1.7 Rapid Review as the next pass when the first map shows likely exposure before 10 December 2026.
Last reviewed
Map the decisions before 10 December 2026.