Delivery standards
How we use AI agents on your work: disclosed, approved, reviewed and accountable.
AI agents are only useful if you can trust the delivery around them. These are the standing terms we work to. Ask us for evidence of anything on this page.
Four commitments on every engagement
Disclosed
Every deliverable states what the agents did, what people checked and who signed it off.
Approved first
We get your written approval before we use AI on your work.
Accountable
One named lead per engagement answers for the result.
Fixed scope
Scope, acceptance criteria and a hard spend cap are agreed before work starts.
How we work to Australia's Guidance for AI Adoption
The Australian Government's Guidance for AI Adoption sets out six practices. This is what each one means on an xShopper engagement.
| Practice | What we do |
|---|---|
| Accountable person | Every engagement has one named lead who owns the result, signs every deliverable and is named in the contract. |
| Impact assessment | Scoping documents the purpose, the data involved, the risks, and what agents may and may not do, before any work starts. |
| Risk management | Scoped access for every agent, a hard spend cap, review gates and a way to pause the work at any time. |
| Information sharing | Every deliverable carries an AI-use statement. On request we give you a register of where AI was used on your work. |
| Testing and monitoring | Acceptance criteria are written first. Tests run before handover, sources are cited and agent actions are logged. |
| Human control | People approve every consequential step and every release. Agents propose; people decide. |
What a deliverable looks like
Every deliverable ends with an AI-use statement, so a reviewer, auditor or procurement team can see at a glance what happened. This is an illustrative template, not a real client record.
AI-use statement (template)
- Agents did: drafted the analysis from the supplied documents; generated and ran the test suite.
- People did: defined scope and acceptance criteria; reviewed every finding against source; approved release.
- Sources: listed per finding, linked to the supplied documents.
- Data handling: as agreed in the engagement scope.
- Signed off by: [Name], [Role], [Date].
What we will not hand to an agent
Agents are fast, not wise. We keep these with people.
- Decisions about people Hiring, performance and employment decisions.
- Final legal, financial or regulatory sign-off Agents can prepare it. People sign it.
- Unreviewed changes to production systems Every change is reviewed and approved first.
- Sensitive or personal information Not unless the engagement scope names it and you approve it.
- Unsupervised communication with your customers Agents draft. People send.
For government and regulated buyers
Australian Government buyers are directed to require suppliers to notify them, and get written approval, before using AI to deliver services. We do that by default, and we can work to your contract's AI clauses. State and territory agencies and regulated financial institutions have their own frameworks, and we will work to yours.
Claims we do not make
We do not claim IRAP assessment, Essential Eight maturity levels, ISO certification or panel membership on this site. If an engagement needs one, we will tell you whether we hold it and show you the evidence.
Name the outcome. We'll bring the fleet.
Tell us the work you can't staff or finish. You get a fixed-scope proposal, written acceptance criteria, a hard spend cap and a named lead who answers for the result.
Request a scoping call