Before KPMG began selling its AI delivery model to enterprise customers, it sold it to OpenAI.
Frontier Lab hired KPMG to build an internal supply chain and fulfillment orchestration platform—essentially tasking the consulting firm with designing the kind of AI-native workflow system that KPMG now wants to sell broadly.
This “client zero” deployment, as KPMG calls it, is the basis of a new alliance between the two companies, announced today, with KPMG named an OpenAI Elite Partner – the highest tier in OpenAI’s partner network.
Colleen Kapase, vice president of strategic global partnerships and ecosystems at OpenAI, said Assets This Elite Partner status is reserved for a “limited group of global partners” who have the reach, scale and delivery capabilities to support the adoption of AI in enterprises worldwide.
“We’re beyond experimentation,” Chad Seiler, KPMG’s U.S. industry leader for technology, media and telecommunications, told me in an interview. “This is about use in large companies.”
The core of what KPMG is selling is a bet on how enterprise software will change from now on. Currently, most employees interact with work through applications: logging into systems, navigating screens, and clicking through modules. Seiler’s argument is that this is coming to an end.
“When we talk about headless, what we really mean is decoupling the work experience from the underlying systems, screens and modules, while maintaining those systems as a system of record,” he said.
In the model he describes, employees stop navigating the software and start describing what they want to get done. AI agents interpret intent, coordinate across backend systems, and execute – or escalate to a human if judgment is required.
The end point of his narrative is the voice. “Over time, you’ll be talking more than typing,” Seiler said. “Instead of just interacting with your ERP system or your CRM system in the traditional way through cumbersome interfaces that are limited in what you can do, you’re sort of unchained and able to literally have conversations with your systems and use them to take action and take actions not just within that system, but connecting those to other data sets and other systems through an intelligent agent layer.”
A July 20 KPMG blog post co-authored by Swami Chandrasekaran and Matteo Colombo puts the shift more formally: “The legacy SaaS isn’t going away. Its interface is evolving. More specifically, a new desktop is emerging.”
The underlying databases and enterprise applications are not going away. They become an infrastructure invisible to the user, running beneath a layer of agents that handles the translation between human intent and machine execution.
Table of Contents
A sandwich, squashed
To flesh out the abstraction, the interview touched on a “sandwich” framework from Princeton’s Arvind Narayanan, whose research, “AI as Normal Technology,” divides work into three layers: a decision layer at the top, an execution layer in the middle, and a delivery layer at the bottom—what he calls the “decide, execute, deliver sandwich.”
Narayanan’s argument is that AI only compresses the execution level, which initially was never more than a third of the work, while the decision and delivery levels – judgment and responsibility – resist compression and may expand.
I showed Seiler the framework on a screen during our interview. He immediately agreed that it matched what KPMG was seeing in operations – the image actually looked like a thin hamburger patty. But he added a small wrinkle: Speed itself creates new verification burdens at the deployment layer and a new kind of overhead that wasn’t present in Narayanan’s original thesis – talking more about the work than doing it.
“The employee spends less time learning the geography of the software and can focus more on the results they want to achieve,” Seiler said. In his opinion, the decision bun doesn’t just stay put while the execution bun shrinks. It can expand and absorb the coordination work that used to live in the middle. But one day the mindless work will become so good that it might disappear.
When reached for comment, Narayanan argued that software developers have always spent a staggering fraction of their time writing specifications and product requirements documents, which fits into the “bun” section of his metaphor. However, he argued that judgment and accountability are not just structurally resistant to compression, although they do to some extent. It is also the case that “AI is rapidly increasing the ambition and complexity of projects, so that the *ceiling* of judgment and accountability is rising as AI shifts the floor upward.”
Narayanan pointed this out Assets to another blog post where he addressed the SaaS issue and said it was slightly different from the decision sandwich. On LinkedIn, he warned of a “lock-in” in which the AI agent “becomes the main queryable repository of all… tacit knowledge, creating dependency and stickiness,” meaning it is effectively an employee “who cannot be fired without *every* team losing workflow and know-how.”
On the timeline, Seiler and Narayanan are more aligned with each other than their respective positions suggest. Narayanan describes organizational adaptation to AI as a decade-long process – closer to factory electrification than an overnight disruption. KPMG offers an implicit hedge along the same lines, warning against wholesale reinvention: “The most successful companies will think about where to reinvent themselves – and where not,” and points out that “the same workflows that have been in place for years may continue to be the best fit.”
Meanwhile, Narayanan said he doesn’t believe AI is urgent to the extent that Frontier Labs, for example, portrays it as superintelligence by 2027, “but nonetheless it is a more urgent shock than most organizations are used to dealing with.” In other words, it takes a long time to reinvent the sandwich.
Why KPMG says people still matter
The most interesting point Seiler makes is not about technology. It’s about why a consulting firm is well positioned in a world where technology is becoming a commodity.
His answer: decades of custom institutional knowledge that no frontier model has. “We know their business models, their people, their culture, their systems, their data, their politics, their silos,” he said, “in a way that is intimate and at scale that some of these frontier models don’t.”
Kapase agreed that KPMG has “deep experience in business transformation,” particularly in highly regulated industries, the public sector and cybersecurity, where governance and implementation expertise are critical. She pointed to public sector modernization and a product called Daybreak Cyber as key aspects of the partnership, in addition to KPMG’s client-focused work within OpenAI. OpenAI is committed to broad access to its entire ecosystem, she added, so KPMG does not get exclusive access to unreleased OpenAI features.
For his part, Seiler described the OpenAI alliance as additive rather than exclusive: KPMG has parallel partnerships with other frontier labs, including Anthropic, and does not expect large customers to standardize on a single AI provider.
“We don’t think there will be many cases where we have a single customer using exclusively a Frontier model to run everything,” he said. He acknowledged that some customers are already using cheaper alternatives – including open source models from China – for narrower tasks as a cost and stability hedge.
Regarding the rise of open source models, Kapase said that OpenAI’s focus is on “helping customers get greater value from OpenAI.” She found that GPT-5.6 delivers more intelligence from every token and higher performance per dollar, with Sol being 54% more efficient at agent coding tasks. KPMG employees in its consulting and internal teams have been using OpenAI capabilities daily since the company integrated them into internal AI tool aIQ Chat in 2023, Kapase noted, with KPMG identifying Codex-related use cases while building AI-enabled capabilities for clients.
What sets the OpenAI deal apart, according to Seiler, is the go-to-market dimension. “It’s one thing to work with the labs and another to also work with them and go to market with them.” Client zero deployment is proof of this: if KPMG could build this for OpenAI itself, it would be significantly easier to market to any other enterprise customer.
KPMG concludes with a note that sounds less like a technology announcement and more like a management consulting memo – and that’s probably the point. Adopting agent AI is “a portfolio of business decisions to be made, not a technology migration.” For a company whose value proposition has always been helping large organizations make difficult decisions carefully, this is less a hedge and more a positioning statement.
https://fortune.com/2026/07/21/exclusive-kpmg-openai-elite-partner-headless-software/
