APS leaders need re-engineered processes before more AI tools, according to Commonwealth Bank’s Data and AI Leaders

Government leaders keep naming workforce readiness as the real barrier to AI delivery.

One place to start is redesigning the work itself before anyone touches the tooling.

Desmond Seeley, Data and AI Leader at Commonwealth Bank of Australia, sat down with ADAPT’s Content Lead, Justina Uy, to talk about reshaping work types, keeping humans accountable as AI takes on more of the grunt work, and adopting AI tooling safely inside one of the country’s most regulated environments.

At Government Edge in Canberra on 16 November, Desmond will give delegates a method for reshaping work types and adopting AI tooling safely, proven inside one of Australia’s largest and most heavily regulated organisations.

Key takeaways:

• AI adoption should start with sorting work into what AI should do and what people should keep, rather than with buying tools.
• Accountability stays with humans regardless of how much governance and monitoring work AI takes on.
• A golden path starts as a process re-engineering exercise, with tooling decisions coming after.
• Closed feedback loops, applied properly, turn AI anxiety into managed, measurable value.
• A regulated, well-funded private sector background can shortcut government’s trial and error.

Splitting grunt work from creative work

For Desmond, AI adoption means relooking at the work itself, distinguishing between grunt work and creative work.

Grunt work is what AI should be doing. Creative work is what people should keep doing.

At Government Edge, he’ll walk delegates through a real engineering process he’s re-engineered using this split, and what it changed about how people work and where the value actually showed up.

Accountability stays human, even as AI takes the grunt work

If the workforce isn’t ready to challenge or question a tool, does the guardrail actually hold, or does the risk just move from the tool to the person?

According to Desmond, accountability never shifts off humans.

AI’s role is reducing the cognitive load of governance and monitoring work, the detailed checking that eats people’s time, while humans keep the creativity and the responsibility.

Golden paths are a re-engineering exercise, not a shopping list

ADAPT’s own research has found only 6% of government leaders are seeing measurable productivity gains from the AI tools they’ve bought.

Desmond puts that gap down to mindset and sequencing.

A golden path is an integrated developer platform built by re-engineering a process end to end, then embedding tools inside it, rather than bolting tools onto an unchanged process.

Have no fear, use closed loops to manage what AI can’t

Asked for his one piece of advice for the 160-plus APS tech leaders heading into 2027’s operating model redesign, Desmond kept it simple, have no fear.

He points to a proper cybernetics view, building in closed feedback loops that catch what’s drifting out of control, as the mechanism that turns AI unease into real, managed success.

What a regulated, well-funded bank can teach government

Coming from the private sector to speak at a government conference, Desmond sees his background in a highly regulated, well-funded organisation as an advantage.

Large banks have already spent the money and absorbed the trial and error involved in tightening controls without breaking them.

He believes government agencies can skip a chunk of that same learning curve, and lift their risk appetite slightly, by drawing on lessons already tested at scale inside banking.

Hear more from Desmond at Government Edge where his session will give delegates a method for reshaping work types and adopting AI tooling safely inside a regulated environment.