I Asked Muse to Help Me Book My Holiday Travel. It Didn’t Fly Far.
THE SO WHAT
If a flagship AI assistant still struggles with basic travel booking, the near-term ceiling for consumer agents is orchestration, not full autonomy. Product teams should design for “AI drafts, human confirms” flows rather than promising end-to-end task completion.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIUK says it will again test OpenAI and Anthropic models before release
Pre-release access for the UK AI Security Institute means frontier model risk assessments are drifting toward a government–lab co-regulation pattern. If you ship on top of OpenAI or Anthropic, budget for UK-originated safety requirements to show up in your enterprise contracts and compliance checklists.
Applied AIBritain steps up its defences against AI risks
Finding 407 vulnerabilities across nine agencies for £13,000 in tokens is a proof point that AI-augmented red teaming is cost-effective at scale. CISOs should assume regulators will adopt similar tools—and preemptively use AI to probe their own estates before someone else does.
Applied AIHoneyjar offers AI tools to improve PR and comms at every company
AI-native PR tooling means narrative testing, pitch drafting, and media monitoring are about to get cheaper and faster—volume will spike. Comms leaders should focus less on first-draft generation and more on guardrails, approval flows, and measurement so they’re not just flooding inboxes with slightly-better spam.
Applied AIAI Supply Chain Strains May Be the Least of Our Worries
If supply chain bottlenecks are no longer the scariest part of the AI story, the concern is shifting to systemic impacts—energy, labor, financial markets. Boards should be asking not just "can we get GPUs" but how concentrated AI demand could stress their own upstream dependencies.