Whether ‘secret’ or ‘AI’, agents need trust to do their job

For all the fanfare, AI agents are still only conspicuous by their absence.

The likes of James Bond and Ethan Hunt have, ironically, made a name for themselves through secrecy. Operating strictly undercover and able to disappear at a moment’s notice, secret agents depend on staying undetected.

AI agents share a title with these operatives, but you’d hardly expect them to work undercover; so why are tech vendors being so shy about celebrating big adoption wins?

It might be because there aren’t any to shout about.

As John Leonard observed this week, “I’ve seen plenty of tech company demos where agents whizz around 10X-ing this and hyper-personalising that…but on further digging these have all turned out to be proofs-of-concept or rebadgings of RPA to fit with the zeitgeist.”

Even Google’s Cloud London event, which 10 months ago was agents, agents everywhere, was this month notable mostly by their absence. Instead, IT leaders I talked to were looking to more practical (and proven) AI use cases focusing on analytics and customer experience.

What’s going on? How have we moved from “AI agents will book my holiday, manage my finances and buy my groceries” to, in a few short months, a deafening silence?

Image

Description

AI agents, at the Peak of Inflated Expectations, aren’t far behind generative AI, both setting a blistering pace through Gartner’s 2025 Hype Cycle. Source: Gartner

It’s all down to understanding applications and limitations. AI agents today can absolutely autonomously retrieve data, but where’s the value there? To really achieve what’s being hyped, they need wide scale access to internal, possibly privileged, data – and what CIO (or CISO) is going to trust a third-party tool to wander around their infrastructure and browse whatever it likes?

If you want to implement AI agents at scale, and especially if you want some sort of multiagent system, you need to get a handle on your data, as I’ve written before. But it’s also about identity, trust and security.

Bond might operate in the shadows, but if Bob the Bot wants to be useful, he needs to take a very different approach.

Recommended Reads:

As well as our coverage of AI agents above, we’ve published an exclusive analysis of Google’s partnership with the UK government. Both stakeholders have made big promises, but what’s the reality? We dig past the fluff to expose the cold hard truth.

Security is key in modern IT, but it never stands still. The industry is still rushing to handle AI disruption, and now quantum computing is threatening the tentative equilibrium. Samara Lynn, editor of our sister site MES Computing, has looked at what IT leaders are doing now to prepare for ‘Q Day’.

In our latest episode of Ctrl Alt Lead, Amnesty International CIO Paul Smith explains how his role – with workers in warzones, activists facing political pressure and local teams who closely guard their autonomy – is a challenge like no other.

Finally, European cloud provider OVHcloud topped rankings this year in our annual cloud sustainability research; Penny Horwood talked to Grégory Lebourg, global environmental director, about why OVHcloud is frugal by design.