At a glance
42% of Australian organisations already have a chief AI officer or equivalent, rising to a projected two in three by 2027.
47% name a clear mandate from the CEO or executive as the top condition for the role delivering.
Only 9% see managing vendors and platforms as the job, against 60% for strategy and roadmap.
88% expect the role to overlap with the CTO, but only 5% think it should absorb the CTO's remit.
Australian governments have produced four different AI structures in twelve months, from a federal office to a state ministry.
Chief AI officers need authority, not just a title
This is part two of our interview with Datacom's director of artificial intelligence, Lou Compagnone. Part one ran as a Q&A with Lou Compagnone, published as she submitted it.
The chief AI officer barely existed in Australia two years ago. Today 42 per cent of organisations have one or an equivalent, the Australian Public Service expects every agency to appoint one, and two in three organisations expect to have one by 2027. What the person does, who they report to and what they are allowed to decide are all still open.
Datacom has put the first real numbers on it. The company commissioned the research house Omdia to survey 507 Australian IT and business leaders in April and May 2026, two thirds of them private sector, and published the findings as a white paper, The emerging role of the chief AI officer in Australia. We tested what it says against what is actually being built, in companies and in government.
Asked what decides whether a chief AI officer succeeds, respondents named a clear mandate from the CEO or executive above everything else, at 47 per cent. Then 43 per cent said the role should sit below the executive table. The job they describe is strategic: 60 per cent put an organisation-wide AI strategy and roadmap at the top, 50 per cent responsible AI governance, and only 9 per cent managing vendors and platforms.
Datacom's director of artificial intelligence, Lou Compagnone, wrote the research. She told Certified Strategic the title "almost doesn't matter", so long as the person holds "a seat at the executive table and they can create influence across the organisation." Without clear authority, organisations may struggle to realise the intended value of the role.
73% say the public service requirement shaped their approach
The Australian Public Service requirement asked agencies to appoint a chief AI officer from their existing senior leadership by 30 June 2026. 73 per cent of Datacom's respondents say it is shaping how their own organisation thinks about AI leadership, 23 per cent significantly. Compagnone credits the public sector with leading on this one.
Governments have since produced four structural answers to the same question in twelve months. Prime Minister Anthony Albanese announced an Office of AI inside the Department of the Prime Minister and Cabinet on 15 July 2026, to coordinate a national AI framework and set Australian standards. Victoria created a Minister for Artificial Intelligence on 6 August 2026, Premier Ben Carroll appointing Anthony Carbines to keep AI development "ethical and responsible". NSW has run an Office for Artificial Intelligence since September 2025 under Minister Jihad Dib, with an AI Review Committee chaired independently by Edward Santow.
Five components turn an AI strategy into delivery
Asked what she wants inside a government AI office, Compagnone starts with what precedes the org chart. "I really want them to be clear about their blueprint for how they're going to do their transformation, so not just to have a chief AI officer and then the organisation operates in the way that it always has." Strategy alone does not survive contact either. "They really need to think about setting up a proper operating model for how they'll execute on that strategy. Otherwise it just becomes a document that's printed and put in a drawer."
The model she describes, which Datacom runs internally and says government would be smart to adopt, has five parts.
Component | What it does | Why it matters downstream |
|---|---|---|
Centre of enablement | A PMO for AI transformation, running intake and prioritisation of use cases | A single queue where infrastructure requirements can be assessed before build |
AI academy | Capability uplift and change management, owned by people and culture | Determines whether deployed capacity is actually used |
AI council | Accountable for investment allocation and risk | Where a use case's cost, including token spend, is weighed against its value |
AI lab | Incubator for larger opportunities, linked back to the centre for smaller ones | Where non-standard compute requirements first appear |
AI foundations owner | Token management and optimisation, data readiness, working with the chief AI officer | The closest thing to an infrastructure counterpart in the model |
Source: Certified Strategic Editorial, interview with Lou Compagnone, Datacom, August 2026.
A sixth element sits outside the centre, and Compagnone recommends government copy it from Datacom: "we have AI domain experts in every single service and function." Those people hold deep expertise in their own area and think about what is coming for their function. She describes them as the connective tissue between the centre and the rest of the organisation.
Government AI offices should also bring the private sector inside the tent rather than consult it at arm's length, she says, "to inform where they're going with policy and how they're going to operationalise it, and also what they're going to do in terms of infrastructure for Australia as a country as well." On governance she rejects the assumption that oversight slows delivery: "governance done well isn't a brake. It actually makes people go faster."
Compagnone argues the role needs more than technical leadership
"A lot of the people debating with me were actually CIOs and CTOs, saying the chief AI officer has to be deeply technical, they need to own the architecture, they need to own the infrastructure," Compagnone told Certified Strategic.
But Compagnone is clear that the chief AI officer role "has to be [about] transformation and not just sit within tech." She describes the role as "a very horizontal role" spanning strategy, governance, technology and people. "If it's really stuck in just group IT, they're going to be thinking about it just from a technology perspective, and not actually all the adoption that needs to wrap around it." She is not arguing it can be a non-technical role. "They need to be technical enough to be able to almost be a translator between the technical teams and the business. I always call myself bilingual in that sense."
Datacom's own numbers say the overlap with technology leadership is real but bounded: 88 per cent of respondents expect significant or some overlap with the CTO, while just 5 per cent think the chief AI officer should own the technology remit outright.
She argues the strongest candidates are likely to bring experience in transformation, service design, change management or operational leadership, alongside sufficient technical fluency, not "someone who's just from an architecture background or a CTO." Australian employers are hiring the opposite profile. One consultant working across government and education observed that organisations "are hiring senior tech talent for these roles, not people who are experts on systems, process and change management". Compagnone endorsed the assessment.
Firms of five to 100 people adapt faster than 10,000-person ones
Compagnone raised organisation size without being asked, and her view cuts against the assumption that scale helps here. "I think it's almost easier to be starting from scratch, because you can design a frontier AI organisation from the beginning." Organisations of roughly five to 100 people she considers well placed, because they have less to reimagine. "I think it's a lot harder for larger organisations, say 6,000 to 10,000 people, because they're very set in their ways in terms of existing processes." The sharpest pressure, she expects, comes from firms built AI-native from day one: "what if we only had five to 10 people, and then we had a whole multitude of agents in support of those people?"
Her objection to treating AI as a headcount exercise is operational. "What really alarms me is when I see things in the news about how organisations are ripping 2,000 roles out of the contact centre," she said. The reason she gives is practical: "you can't just simply replace a role with a piece of technology. These people have a lot of tacit knowledge. There's processes that are in their heads. There's huge interrelationships between the whole organisation." Ford rehired more than 300 engineers in June 2026 after AI-led quality control fell short.
What to watch
Shared service patterns. Compagnone wants something Australia has not attempted. "A lot of them are going to have quite common use cases that are the same. What I'd really love to see is common service patterns or building blocks across government." She points to Estonia, which runs public services over X-Road, an open-source data exchange layer, and names the commercial resistance: "the vendors probably hate that because it's less work for them, but from a consistency perspective for citizens, and from a cost saving perspective for government, it's a definite advantage." Asked whether Australia could do it: "I think it could be possible. It would just take a huge amount of collaboration." The appetite shows in her own data: 58 per cent of respondents want shared AI standards, and only 2 per cent say no collaboration is needed. Shared patterns would consolidate demand into fewer, larger and more specifiable workloads, read against the 9GW of data centre load now in AEMO's connection queue. That window narrows as agencies commit to individual stacks.
Who operators are actually selling to. The person setting enterprise AI direction is increasingly not an infrastructure buyer by background. Technical specifications written for an infrastructure-literate reader will not land with them. Operators are formalising the counterpart function internally, as NEXTDC has done with a sovereignty seat on its executive team.
Whether authority follows the title. A chief AI officer without a clear mandate will struggle to align strategy, governance, technology and workforce change across the organisation. Compagnone's answers on sovereignty, workload placement and the wall between pilot and production run in the full Q&A.