Copilot and Copilot Studio agents
Staff can build and share agents that read mail, files and Teams chats, and act through connectors.
Ask this weekWho is allowed to build and publish agents, and which connectors can they use?
We help boards and executives in government, financial services and education count, govern, cost and run their AI agents. Every engagement starts with a fixed-fee step of two to four weeks that ends in a decision.
| Agent | Manager | Authority | Status |
|---|---|---|---|
| Claims triageReads claims, ranks urgency | Ops lead | Recommend only | Green |
| Supplier onboardingCreates vendor records | Procurement | Act ≤ A$10k | Amber |
| CRM copilotSwitched on by vendor | None named | Whatever access allows | Red |
| Student wellbeing flagShapes decisions about people | Head of Pastoral | Recommend only | ADM |
of agentic AI projects forecast to be cancelled by the end of 2027, for cost, unclear value or weak risk controls.
Gartner, June 2025of day-to-day work decisions forecast to be made autonomously by agentic AI by 2028, up from 0% in 2024.
Gartnerof leaders expect to use digital labour to expand workforce capacity within 12 to 18 months.
Microsoft Work Trend Index 2025Many agents arrive inside platforms you already pay for. Tick the ones you use to see what they can do, how exposed they are to the new ADM rules, and the question to ask this week.
Staff can build and share agents that read mail, files and Teams chats, and act through connectors.
Ask this weekWho is allowed to build and publish agents, and which connectors can they use?
Autonomous agents for service and sales that answer customers, update records and trigger flows.
Ask this weekDo any agents route cases, score leads or decide who gets an offer?
Agents that summarise, route and resolve IT and HR cases and requests.
Ask this weekDo agents decide outcomes of HR cases or access requests?
Agents that support recruiting, talent and pay processes.
Ask this weekIs any candidate screened, ranked or shortlisted by a program?
An assistant and agents across S/4HANA and SuccessFactors for finance and HR tasks.
Ask this weekDo agents touch payroll, leave or performance decisions?
Agents across finance, HR and supply chain inside Fusion.
Ask this weekWhich agents are switched on, and who in finance owns each one?
Drafting, summarising and custom Gems working over Drive, Gmail and Docs.
Ask this weekIs Gemini used to assess applications, complaints or performance?
Agents that answer customers, issue refunds or decide escalation.
Ask this weekCan an agent refuse, approve or prioritise a customer without a person?
General capabilities as described by each vendor. What's enabled in your tenancy depends on your licences and settings.
Organisations covered by the Act must explain in their privacy policy where computer programs use personal information to make, or substantially help make, decisions that significantly affect people. To write that honestly you need the inventory most organisations have never done. Done well, the compliance work becomes your agent register.
Tap a question to compare. Very few organisations can answer it for their agents.
Closing that gap is an operating-model change the COO, CIO, Chief Risk Officer and People leaders own together. We call the result the hybrid workforce: people and agents planned, governed and reported together.
You get the inventory the Privacy Act now requires, and it becomes the first draft of your digital workforce register.
A permitted-action register, a production gate and a two-page board paper rating each agent red, amber or green.
A kill, fix or scale verdict on up to eight pilots, with a 90-day plan for the top two.
An independent build, wait or substitute verdict for each workload before you sign the renewal.
Graded roles, delegations agents earn through evidence, and a playbook for the managers who supervise them.
A costed private-AI design aligned to the ISM and Essential Eight, with governed knowledge bases.
A shortlist ranked by teacher time returned, with child safety designed in from the first workshop.
A two-week, evidence-based view of where AI pays back and what must be true first.
Each step answers one question and ends in a decision. Skipping Govern is the most common reason pilots stall.
We don't resell licences or take referral fees. Build, buy, wait or substitute verdicts rest on your numbers alone.
Every engagement ends with a decision, an owner and a 90-day plan, written for executives and directors.
First steps run two to four weeks at a fee agreed up front, so you can stop or continue on evidence.
Thirty years at PwC, Deloitte, Accenture, IBM and Capgemini, including P&Ls over A$500 million. You work with the director, not a pyramid.
Having led SI practices, we know how they price, staff and deliver, and how to hold them to outcomes.
Privacy Act, APRA CPS 230 and CPS 234, the Information Security Manual and Essential Eight, and the Commonwealth policy for responsible use of AI in government.
We'll agree which first fixed-fee step fits, or tell you if you don't need one yet.