AI isn’t enough: How Frontier Firms redesign how work gets done

Across boardrooms today, the pressure to act on AI is unmistakeable. Investors expect it. Customers assume it. Competitors are quick to talk about progress. At the same time, AI’s potential is still taking shape, leaving leaders in an uncomfortable middle ground — confident they need to move, but not always certain where to focus or whether they’re making the right bets.

As they navigate in low visibility, one question is clear: how to turn AI momentum into sustainable value. Where to start? How to scale responsibly? And how to ensure today’s investments compound rather than stall? Very few organizations have AI fully figured out yet. That’s not a failure. It’s a signal that it’s still early in the frontier of AI.

The organizations leading on AI are what Microsoft describes as Frontier Firms: human-led, AI agent-operated organizations. Employees at these firms don’t simply use AI tools — they orchestrate AI agents. People set goals, apply judgement and own accountability, while AI agents analyze information, propose actions and carry-out multi-step processes across systems. For these teams, AI is embedded into how work happens every day.

It can be easy to mistake Frontier Firms as the organizations that implement breakthrough AI tools. Instead, these organizations undergo full-scale transformations that use AI to unlock productivity and build habits at scale before delivering tailored experiences across end-to-end business processes.

This evolution isn’t a destination reserved for digital natives or early adopters. We see it as a practical operating model for organizations ready to pursue AI with intention. Through the Microsoft and KPMG alliance, and our work alongside leaders navigating this shift, we’ve seen that progress doesn’t come from chasing the next breakthrough. It comes from rethinking how value is created, governed and sustained over time — and designing for it from the start.

Right now, AI activity is high for most organizations. The impact, however, is uneven at best. Recent research out of MIT suggests that the vast majority — 95 percent — of corporate generative AI initiatives stall or fail¹. The outcome, we believe, has less to do with the technology, and more to do with business value and how organizations scale the benefits from small pilots to implementations across business areas while building trust.

Frontier Firms treat value as the primary design constraint. Before any technology decision is made, leaders get clear on the business problem they’re trying to solve and who benefits if it’s solved. In practice, this looks like prioritizing a smaller number of use cases tied directly to specific value markers be they cost, growth or risk. From here, they explicitly define how they will measure value. This could be adoption, utilization or time-to-decision rates, or margin improvements, loss avoidance or working capital impact. 

Vodafone demonstrates this point clearly. After piloting Microsoft Copilot with 300 users, Vodafone needed a clear, data-backed case to scale. Working with KPMG professionals , it identified where AI would deliver the most value across roles and functions. With a clear view on value, the company targeted a Copilot rollout to 68,000 employees and confidently moved from experimentation to enterprise-wide impact.

Once value is defined, the next question becomes clear: how will work need to change to deliver the value projected? When AI is introduced without adjusting roles or aligning to day-to-day workflows, people are left to make sense of it on their own. That’s often where momentum fades and well-intended investments stall. 

Frontier Firms approach this differently. They treat AI as a work design challenge, not a technology deployment exercise. Instead of automating tasks in isolation, they redesign processes end to end, assigning AI agents to analyze information and coordinate actions across systems. This frees employees to focus on decisions and outcomes, rather than execution. 

We’ve seen this shift play out across KPMG firms. When KPMG invested early in Microsoft AI technologies, the firm enabled its professionals to access Microsoft Copilot, Azure, and related tools — supporting a shift from experimentation to enterprise‑wide impact. We redesigned how work flowed across functions , clarifying where AI could act autonomously and where human judgement mattered most. Adoption accelerated once our teams understood not just how to use AI, but when to use it and how success would be measured. It also enabled our teams to develop AI-powered platforms like KPMG Velocity that integrates our Global Advisory solutions to work collaboratively across service areas.

The same pattern appears in client work. At Telstra, the KPMG Australia team built an AI-driven compliance accelerator on KPMG Workbench, an open multi-agent AI platform, to help its teams manage more than 20,000 controls by risk level². The team also introduced processes improvements that focused the team on higher-risk decisions rather than manual reviews, increasing productivity and creativity across teams by 80%. Together, this gave the company’s leadership greater confidence in the quality of their controls.

As we prepared to introduce AI into our internal workflows across KPMG firms, we first established the KPMG Trusted AI framework³ as the foundation of how our principles would be put into practice. We brought together our teams in risk, cyber, legal and technology early to align on how AI would be used, where accountability would sit and what “good” looked like across the lifecycle. Those early decisions, and refined as we’ve continued to test and learn, ensured the framework guided our approach. With these guardrails in place, our teams were able to develop new use cases, which leaders could approve faster.

For leaders ready to build their frontier AI capabilities, early choices can set the direction for everything that follows. Based on what we see working among Frontier Firms, these are the moves that can create momentum without overreaching.

For leaders navigating this moment, we want to make one message clear: Lasting advantage won’t come from adopting AI faster than your peers. Step back from the tools and rethink how work gets done. Design for value, trust and scale from the start.

KPMG can help. We combine our lived experience as early adopters of AI across our member firms, with our deep industry, risk, technology and transformation expertise allowing us to help clients move from intent to impact with confidence. Our global alliance with Microsoft further strengthens our position. Together, we bring well established platforms, responsible AI principles and real-world implementation experience to support organizations as they evolve toward a human-led, agent-operated model — securely, responsibly and at scale.