The Question We Kept Hearing at AGS26: When Do You Actually Redesign?
At AGS26 on 10 September, 250 finance, technology and GBS leaders met in Petaling Jaya. One question wasn't on the agenda, but it sat under almost every session: when do you actually redesign? It's also the question we hear most from GBS leaders across ASEAN.
Most centres already have AI running somewhere. A pilot here, a bot there, a batch of licences rolled out. But ask whether the centre actually works any differently, and the honest answer is often "not really". So, they hold off until the tools improve, the budget comes through, or someone else proves it first.
Our answer, and the one we heard again and again on stage: don't wait for the tool. Redesign when you know which work is human and fix the process before you automate it.
Waiting for the tool is the wrong instinct
We understand the instinct. Technology looks like the expensive, risky part, so it seems sensible to get it right first.
But the tool is rarely what's holding a centre back. A leader from McKinsey & Company shared that about 64% of AI licences inside large organisations sit unused, and that the gap between an average user and a power user on the same tool can run six to 17 times. "Providing an AI licence is not the same as creating AI adoption."
In other words, many centres are already paying for AI they aren't using. Buying more won't fix that.
Redesign starts with the work, not the technology
The fix is to change the order. Start with the problem and the outcome you want, then choose the technology.
A leader from Shell put it simply: "Never automate a broken process." Her team fixes data quality upstream and rebuilds workflows before any AI tool touches them.
L'Oréal BEST took the same route. Its Kuala Lumpur team streamlined its processes first, then moved beyond "checkbox" automation to build custom AI "companions" tailored to specific roles. Its leader also challenged how success gets measured: "AI adoption is a KPI. Process transformation is the outcome."
We start every redesign with one simple test. For each activity, ask: is this judgement, or is it moving information from one place to another? Does it need something only a person has, like context on an exception or ownership of the outcome? Or is it a rule applied at volume? Once you've drawn that line, the tool question gets smaller, and cheaper.
Why the order matters to the CFO
Skip that test and three problems show up later, usually in the business case review.
The savings never land. AI frees up time, but unless leaders decide where that time goes, it gets absorbed back into the same jobs. The hours are real. They just never reach the P&L.
Nobody owns what happens after go-live. A leader from BASF put it well: "The technology might be live, but it's the operating model that keeps it alive." If no one has decided who owns, checks and improves the AI's work, your controls team will ask. So will your auditors.
The vendor designs your centre for you. If the tool comes first, its workflow becomes your operating model, and that's expensive to undo.
Fixing the work doesn't mean standing still
None of this means waiting until every process is perfect. AstraZeneca shows why.
Its GC GFS Innovation Hub evaluates ideas for business impact, scalability and reuse, and scales the strongest ones globally. The ideas that never scale still count. They remove inefficiencies, standardise and digitise how the team works, and strengthen the digital foundation for AI. As its leader put it: "I often describe AI as a high-performance race car. It can deliver extraordinary results, but only when we have built the digital highway that allows it to perform at its full potential."
Small fixes build the highway. The redesign decides where it goes.
Three questions to ask before you redesign
If your team is asking "when do we redesign?", answer these for one process first:
Is this work mostly judgement, or mostly moving information?
If it were automated tomorrow, who would own the outcome, and where would the saved time go?
Would we design this process the same way if we were starting today?
If you can answer all three with confidence, that process is ready. The tool choice follows from the answers.
Back to the question from AGS26
The leaders on stage came from very different companies. What they shared was the order of work: fix the process, decide what stays human, then automate. In our experience, that order is what separates centres that scale AI from centres stuck in pilots.
It matters beyond any one centre, too. Malaysia's RM28 billion GBS sector now spans around 750 companies and more than 150,000 professionals. As our Head of Digital put it: "The industry has to redesign workflows and upskill talent, not simply adopt new technology, to sustain the shift from cost savings to enterprise value."
So when do you actually redesign? Not when the tool is ready. When you know which work is human. You can start answering that this quarter, with one process.
Want a structured way to start? The AGOS AI GBS Readiness Assessment maps where your centre stands today and where redesign will pay off first.
Comments