FTC Probes AI Agents as HMRC Signs £2.4bn Salesforce Deal

What happened

Two regulatory and commercial developments are shaping the customer-experience and AI agenda this week. In the United States, the Federal Trade Commission has opened scrutiny into autonomous AI agents — software designed to act on behalf of users or businesses with minimal human oversight — amid concern about what happens when such systems behave unpredictably or "go rogue" in live customer or commercial interactions.

Separately, in the UK, HM Revenue & Customs has signed a Salesforce agreement reported to be worth £2.4bn, a significant multi-year commitment that signals continued investment in cloud-based service and data platforms within a major public-sector tax authority.

Why it matters

The FTC's interest in autonomous agents lands at a moment when businesses across sectors are racing to deploy agentic AI in customer service, sales and operations — often faster than governance frameworks can keep pace. Regulatory attention this early in the agentic-AI adoption curve suggests that accountability, auditability and fail-safes will increasingly be treated as product requirements, not afterthoughts, for any organisation deploying AI that can act rather than merely respond.

The HMRC-Salesforce deal, meanwhile, underscores that large-scale platform modernisation remains a live priority for public-sector bodies handling millions of citizen interactions. A commitment of this size points to sustained demand for cloud CRM and data infrastructure as government agencies seek to consolidate services and improve how they manage citizen and business engagement over the long term.

By the numbers

• £2.4bn — the reported value of HMRC's agreement with Salesforce.

The Renascence take

Put together, these two stories describe the same tension from opposite ends: one is about a regulator trying to catch up with AI that acts on its own, the other is about an institution betting big on the platforms meant to manage exactly that kind of complexity at scale.

The real story here isn't the size of a contract or the existence of an inquiry — it's that organisations are now operating AI systems faster than they can explain them. Agentic AI fails or frustrates customers not because the technology is broken, but because nobody designed for the moment it acts outside expectations. Any operator deploying agents — in a call centre, a claims process or a tax authority's back office — should be asking who owns the failure mode, not just who owns the roadmap. Procurement size and regulatory attention are both proxies for the same underlying truth: autonomy without accountable design is a service-experience risk waiting to surface.