KPMG: Inside the AI agent playbook driving enterprise margin gains

Host A: Welcome back to AI Catchup Weekly, I'm here with my co-host, and we've got a really meaty one for you today — KPMG just dropped their first Global AI Pulse survey, and the numbers are both impressive and, honestly, a little humbling.

Host B: Yeah, I saw the headline figure and did a double take — we're talking about organisations planning to spend a weighted average of $186 million on AI in the next twelve months. That is not pocket change.

Host A: It really isn't. But here's the kicker — despite all that spend, only 11 percent of organisations have actually reached the point where they're deploying and scaling AI agents in ways that produce real, enterprise-wide business outcomes.

Host B: So we've got a lot of money chasing a pretty narrow slice of actual results. What's going wrong for the other 89 percent?

Host A: KPMG puts it pretty clearly — most companies are layering AI tools on top of existing workflows. Think co-pilots, summarisation tools, that kind of thing. You get incremental gains, but you're not fundamentally changing how the work gets done.

Host B: Right, it's like putting a faster engine in a car that still has square wheels. The organisations actually winning — the ones KPMG calls "AI leaders" — they're redesigning the process first, then deploying agents inside that new structure.

Host A: Exactly. And the performance gap is stark. Among those AI leaders, 82 percent say AI is already delivering meaningful business value. Among everyone else? That drops to 62 percent. A 20-point spread that compounds over time.

Host B: That compounding effect is what should keep executives up at night. Three to five years from now, that gap could be the defining competitive variable in entire industries. So what does the $186 million actually buy — or not buy?

Host A: Great question. A lot of it goes to the visible stuff — compute, licensing, professional services. But the survey suggests most organisations are seriously underweighting the operational infrastructure costs. Things like integrating AI outputs with legacy ERP systems, managing retrieval pipelines, maintaining audit trails for regulated industries.

Host B: The boring-but-critical stuff that nobody puts in the initial proposal and then everybody panics about six months into deployment.

Host A: Precisely. And there's a really interesting finding around governance too — which I know sounds dry, but bear with me. Among organisations still in the experimentation phase, only 20 percent feel confident managing AI-related risks. Among AI leaders, that jumps to 49 percent.

Host B: So maturity and confidence go hand in hand. And I imagine that confidence is what lets the leaders move faster, not slower — they're not stopping for a fresh governance review every time they want to try something new.

Host A: That's exactly the point KPMG's Global Head of AI, Steve Chase, makes. He says there's no agentic future without trust, and no trust without governance that keeps pace. The leaders have baked governance into the deployment pipeline itself — automated monitoring, explainability tools, escalation paths — rather than treating it as an afterthought.

Host B: There's also a regional angle here worth flagging for any multinationals in our audience. ASPAC is leading the charge — 49 percent of organisations there are actively scaling AI agents, compared to 46 percent in the Americas and 42 percent in EMEA. And ASPAC is also out front on the really complex stuff, like orchestrating multi-agent systems.

Host A: So the playbook is becoming clearer: redesign the process before you deploy, invest in the unglamorous operational infrastructure, and treat governance as an accelerator rather than a brake. The organisations that crack all three are the ones pulling away from the pack.

Host B: And for everyone else, the clock is ticking. That 11 percent isn't going to stay small forever.

Host A: It really isn't. Alright, that's going to do it for this edition of AI Catchup Weekly. If this got your brain working, share it with a colleague who's wrestling with their own AI strategy — they'll thank you.

Host B: Or maybe they won't, because now they have a lot more to worry about. Either way, we'll be back next week with more. Take care, everyone.