On Thursday, July 24, I went to GitLab Connect at Crown Sydney with two of my team leads. A full day of talks, panels, and hallway chats focused on one big question: what does software development look like when AI isn’t just helping, but acting?
The day opened with Yasmin London as MC, setting a good tone: upbeat, sharp, and grounded in digital safety and leadership. From there, the pace didn’t really let up.
Here’s what stuck with me.
AI isn't coming — it's already here
Andrew Haschka kicked things off with a stat-heavy keynote. Agentic AI — not just code suggestions, but full end-to-end task execution — is already on the radar for most execs:
- 9 out of 10 think it’ll make teams more creative and competitive.
- 89% expect it to be standard practice in under 3 years.
We’ve moved past autocomplete. By 2025, they’re betting we’ll trust agents to handle real work — with a human reviewing the final product.
Developers want flow. Most don’t get it.
One stat made the room squirm: most developers only spend 8 hours a week coding. The rest is meetings, coordination, context-switching, cleanup.
“Time saved in coding is lost in scrum meetings.”
Agentic AI aims to plug those gaps. Not just by writing code, but by testing it, deploying it, and maybe one day, planning it too.
Trust will make or break it
With agents doing more, the question becomes: how do we keep things secure, auditable, and sane?
GitLab’s answer is “secure-by-default.” Guardrails. Observability. Human sign-off. And a bit of pragmatism:
“If you don’t give people the tools internally, they’ll use the external ones anyway.” — Scott Owen
From Cochlear to Cloud Acceleration
The customer and partner spotlights added some nice texture. Victor Rodrigues from Cochlear spoke about how they’re building scalable customer software platforms. Adrian De Luca from AWS followed with a talk outlining the evolution of AI tooling — from autocomplete in 2023, to assistants in 2024, to full agentic systems by 2025 — and how cloud-native foundations are enabling faster iteration and DevSecOps maturity across the region.
Government is further ahead than you'd think
The public sector panel was a surprise highlight. Victor Dominello and Suneetha Bodduluri (Service NSW) spoke about what it takes to modernise government services — and how AI fits in.
Dominello broke down the evolution nicely:
Gov 1.0 – Paper Gov 2.0 – Online portals Gov 3.0 – Post-COVID expectations of real digital experiences
Their north star? A single digital front door. People shouldn’t need to know which agency to go to — they should just be able to ask for help. And behind the scenes, AI might be doing a lot of the routing.
He also made the stakes clear:
“If we don’t become more productive, we’ll face one of two outcomes: higher taxes or fewer services.”
The stack is ready (or getting there)
AWS showed off Kiro — their new agentic dev platform. Early days, but it looks promising. I’m keen to try it on a side project.
GitLab’s stack is maturing too — and I’m looking forward to using more of it.
I hadn’t realised how many big companies are using GitLab internally. Feels like GitLab is winning the enterprise market quietly... interesting.
From Prompt to Production
Jim Williams and Samer Akkoub from GitLab gave the closing talk on “From Prompt to Production,” covering how agentic AI could play across the entire SDLC. Testing, security, deployments — all handled by smart agents, with humans focused on high-leverage work.
Real Teams, Real Progress
The customer panel that followed reinforced it: real teams are already figuring this out. Leaders from ASX, Squiz, and The Lottery Corporation shared how they’re evolving workflows and embedding AI where it adds the most value.
A Leadership Interlude
And yes, there were sports stars too — Meg Lanning and Glenn McGrath dropped by to talk leadership and resilience. Bit of a tone shift, but a good one.
Deep Dive: GitLab + AWS in Action
The afternoon wrapped with a technical deep dive led by GitLab's Rob Williams (who happens to be our company's GitLab Solutions Architect — shout out to Rob) and AWS's Tomas Mihalyi. The session gave a hands-on look at how GitLab Duo and Amazon Q are powering agentic software development in practice.
They broke down how the integrated GitLab + AWS platform supports secure, scalable DevSecOps workflows, boosting productivity across planning, coding, security, and operations.
Honest Demo Moment
That said, the AI demos were a little hard to follow. Agentic workflows take time, so the presenters leaned on 'here’s one I prepared earlier' examples — which didn’t always match the setup exactly. If you’ve done demos before, it made sense. But I do wish they’d been more explicit about it. Even just calling it out would’ve helped the room follow along.
Wrapping up
Agentic AI is arriving faster than most teams are ready for. The tools are coming. The platforms are aligning. What’s left is the human part — trust, process, and accountability.
Also: still waiting for a fireside chat that actually includes a fire.