Nikki Hao
There's a moment in a conference session when something lands differently — not politely-interesting differently, but the kind where you look at the person next to you and you're both thinking the same thing.
That happened at AWS Summit Sydney, watching a Deputy executive give one of the more honest technology talks we've seen at an event like this. Then a single line appeared across three screens:
AI doesn't fail. Systems do.
The room went quiet.
What Deputy actually said
Deputy — the workforce management platform used by hundreds of thousands of businesses globally — didn't come to tell a success story. They came to tell the story behind the success story.
Their AI deployment stalled. Not because the models weren't good enough or the infrastructure couldn't support it. It stalled because of the structures around the work — how decisions got made, how teams were organised, how work moved from idea to production.
The number that crystallised it: 680 errors in one week. That's not a technology number. It's a structural one — the accumulated result of something misaligned underneath. The skills their teams needed weren't in place. The systems for managing AI workflows weren't designed to carry the load. The way decisions got made and ownership got assigned wasn't built for what they were trying to do.
The technology was fine. Everything around it wasn't.
How they fixed it
They didn't go back to the models. They went back to the organisation.
The approach they landed on (moving from request to result) was built around three stages: Understand, Decide, Act. Simple language for something that took real discipline to build. Real-world problems aren't single prompts. They're workflows. And workflows need systems that can carry them.
The shift was from individual AI tools to something that could operate at scale. Their framing: make it real, make it social, make it habitual. Embed it into the structures of how work gets done, not just into the code.
Why this matters beyond Deputy
Walking out of that session, we had the same conversation we've been having with technology leaders for years — just with new material to point at.
The problems Deputy described aren't AI problems. They're the problems underneath every technology initiative that misses its mark. We've seen them in ERP rollouts, software builds, data platforms, Salesforce implementations. The technology changes. The failure mode doesn't.
When a technology initiative fails, the diagnosis usually lands on the wrong thing. The model wasn't right. The vendor let us down. The timeline was too aggressive. Sometimes those things are true. But more often they're symptoms. The root cause is structural — in the skills brought to the work, in the systems that governed how it ran, in the way decisions got made and accountability got assigned.
Deputy said that publicly, at one of the largest technology events in the country. They had 680 errors in a week. They went back to the structure. They fixed it.
That's the pattern. It's what we see inside technology projects every day.
What the Summit told us about the market
The sessions skewed heavily toward the gap between AI ambition and AI delivery — which is the gap defining most organisations' technology roadmaps right now.
What we didn't hear enough of, and what Deputy were willing to say directly, is that the gap isn't technical. More compute, better models, newer infrastructure won't close a structural gap. The organisations getting AI into production are the ones that treated the structural work with the same seriousness as the technical work.
Skills. Systems. Structures. The same three things that have always determined whether a technology initiative delivers what it set out to.
The technology is the easy part. It always has been.
