From CX pioneer to enterprise imperative
Why governing AI is the next management challenge

From CX pioneer to enterprise imperative, and why governing AI is the next management challenge. I'm Jamie Litherland, and I am the VP of Business Development at VOSS. I was at the Cavell CX Summit in London this week, and I left with a clearer conviction than I arrived with. The customer experience world has spent years quietly proving what the rest of the enterprise is only now waking up to. That artificial intelligence deployed well doesn't replace people. It frees them to focus on the work that needs a human. But it also surfaced an uncomfortable truth. The tooling required to deploy AI responsibly and to prove it was worth the money is largely missing. That gap is exactly where I think the next few years of enterprise technology will be decided. Contact centers and CX teams were among the earliest serious adopters of machine learning in the enterprise. Sentiment analysis, intelligent routing, agent assist, automated quality scoring. These weren't pilots. They were production systems handling millions of interactions. CX teams learned, often the hard way, how to measure whether a model was improving outcomes or just adding cost and complexity. But what struck me at the Cavell CX Summit is how quickly that hard won discipline is bleeding into the rest of the organization. HR is building onboarding and offboarding agents. Finance is automating reconciliation. IT is deploying agents to triage tickets and write code. The techniques that CX team spent a decade refining are now everyone's problem, and most of the organization doesn't have CX's scar tissue to fall back on. Here's where the optimism in the room ran into reality. The single biggest source of uncertainty I heard repeatedly was cost and whether anyone could demonstrate a return. This isn't hand wringing. The numbers coming out over the past few months are genuinely startling. Some enterprises are burning through their entire annual AI budget in three months, and others are watching spend double or triple. The culprit is agentic AI specifically, where a chatbot answers one question, an agent takes multiple steps on its own, and it consumes far more tokens than a simple chatbot. So the budgets companies set in late 2025 are already gone. The headline example is hard to ignore. After rolling out Claude code to roughly 5,000 engineers in December 2025, Uber saw usage nearly double by February 2026. And by April, the company had burned through its entire 2026 AI budget four months in. Microsoft reportedly canceled most of its internal clawed code licenses, partly over cost, six months after rolling them out. These aren't fringe start ups. They're some of the most technically sophisticated companies on earth, and it compounds. One health care enterprise consumed a trillion tokens over six months, generating more than six million dollars in unplanned cost before the finance team even understood what was driving it. The problem, as one cost governance analysis put it, is that these costs aren't hidden on purpose. They're invisible by design, accumulating through normal agent activity that no one had instrumented or governed. The analysts see no ceiling. Goldman Sachs projected in May that token consumption will multiply 24 times between 2026 and 2030. And a 2025 survey found that 85% of companies miss their AI cost forecasts by more than 10%, with nearly a quarter underestimating by 50% or more. Perhaps most telling for anyone weighing up an investment, Gartner predicts that over 40% of Agentic AI projects will be canceled before reaching production by 2027, driven by the real cost and complexity of deploying agents at scale. Layered on top of cost is accountability. The EU AI act (Regulation 2024/1689) is now in force and phasing in with transparency rules taking effect in August 2026 and the major high risk obligations now anchored to December 2027 following the digital omnibus agreement of May 2026. I don't read this as anti business and neither should you. The point of the regulation is to hold the AI supply chain accountable for outcomes, and that's the same thing good automation is trying to achieve. But it does demand something most organizations can't yet produce, evidence. The act expects enterprises to show what each AI system does, which risk tier applies, who is responsible, and what controls support its use. Crucially, most enterprises today are deployers, meaning their obligations relate to how they use AI, including human oversight and continuous monitoring, not necessarily how it was built. The very large organizations have teams forming to handle this. Mid market companies mostly don't, and smaller ones definitely don't. This burden is going to be passed down the solution stack, and the majority of organizations will need help meeting it. This is the problem VOSS AI agent management was built to solve, and it's a problem we've solved before in a different guise. For over 20 years, VOSS has been in the user management business, working with the world's largest telcos and enterprise customers, managing tens of millions of users, and integrating into the business systems that actually run an organization. ITSM, billing, and workflow engines. Governing sprawling, fast changing, expensive estates at carrier scale is precisely what we do. AI agents are the new estate. VOSS AI agent management gives organizations three things they're missing today. The first is visibility, a unified registry that automatically catalogs every digital worker through a telemetry pipeline, ending the era of invisible agents operating outside IT's view. Then there is governance, a control point for access to models and MCP resources with budgets, rate limits, and guardrails enforced at the agent, team, or organizational level so you find out before the invoice, not after. And finally, there's business value insight. Cost tracked not at the raw infrastructure level, but tied to the actual processes an agent runs and the outcomes it delivers, so you can finally answer the return on investment question. That derailed 40% of those projects. That last point matters most. Monitoring spend isn't enough. The organizations that stay in control are the ones that can connect every token to a business outcome and prove it to a regulator, a CFO, or a board. The CX world showed us that AI works when it's measured, governed, and pointed at the genuinely human parts of work. The rest of the enterprise is about to learn the same lesson. The only question is whether you'll have the tools in place before the budget runs out or the auditor calls. So if you're wrestling with AI cost visibility, governance, or proving ROI on your agent estate, I'd welcome the conversation. You can email me directly or connect with me on LinkedIn.