Gravitee

Your AI agent saved the company money. Great! Compared to what? That’s a question more organizations need to be asking. On a recent episode of Trusted2Act, my co-host Cole Howard asked Anita Ganti how companies should measure the real value of an AI pilot. Her answer was refreshingly practical. Before celebrating what your shiny new AI agent accomplished, compare it against a realistic alternative. Maybe that’s traditional automation, a better process, or good old-fashioned regex! And don’t forget the human cost. If your agent writes code faster than your people can review it, have you really improved productivity? Or just moved the bottleneck? I think this is an important part of agent accountability. It’s not just about whether we can trust an agent to do something. It’s whether we should have given it the job in the first place. Great conversation with Anita. Catch it on Trusted2Act! https://lnkd.in/gHXyaqK6

Just because I have permission to do something, should an AI agent acting on my behalf have that same authority? That's a question I explored with John Gren on the latest episode of Trusted2Act. John has spent his career in identity and access management (IAM), and he pointed out something interesting. We've gotten pretty good at identity and authentication. But authorization? That's where we've struggled, even before AI agents came along. Think about it. If I have permission to delete a GitHub repository, does that mean my AI agent should too? Of course not. But how do we enforce that distinction? John shared some great examples of evaluating both the human's identity and the agent's identity before allowing an action.I challenged him on one thing, though. We cant possibly write a policy for every decision an autonomous agent might make. That's where the conversation got really interesting. John introduced the idea of guardian agents that monitor actions and bring humans into decisions when needed. This is exactly the kind of thinking we need as we figure out how to manage and build trust in our digital workforce through accountability. Give the episode a listen, and I'd love to hear your thoughts. Trusted2Act, available wherever you listen to podcasts.

Only 7% of organisations have a named person responsible for the AI agents running in their business. I joined Pascal Bornet on his podcast, to talk about the dangers of responsibility without accountability and how we close that gap. Watch the conversation: https://lnkd.in/eBsfs22E

Stop paying for LLM tokens you don't need. 🛑 If you're dealing with unpredictable AI costs from agent loops, duplicate calls, or oversized context windows, managing it app-by-app isn't the answer. You need centralized control. In our upcoming live session, John Gren and I will show you how to use Gravitee to set token budgets on a live API, enforce token-based rate limits, and route requests smartly across LLM providers so your costs actually match your workload. 📅 Oct 27, 2026 | 9:00 AM EST / 3:00 PM CET 👉 Register now to get your token spend under control: https://lnkd.in/e65QUHHZ #Gravitee #AI #CostControl #APIGateway #LLMs #TechWebinar

For decades we celebrated the builder. Productivity was the proxy for value. Agents will out-build any human on raw output, around the clock. So if your value was that you were a very good builder, that is a problem. If your value is that you are the architect, deciding what gets built and why, a very capable builder is not a threat. It is what finally lets you build at scale. One question for every leader: are your people still builders, or have you given them the tools, the permission and the accountability to become architects?

You ask an agent: “Find me a hotel in London under £300, re-schedule my meetings, and book it.” You may never visit the hotel website. The agent discovers, chooses, authenticates, and acts for you. That is the Delegation Economy, and it puts new pressure on companies to create the right Agent Experience, not by blocking agentic traffic, but by making it secure and governed. My four takeaways on Delegation Economy control: 1. Human habits become delegated tasks. 2. Your most active customer may soon be an agent. 3. Enterprises need to control agents, not block them. 4. Delegation creates an authority problem. The interesting question for companies is what happens next: how do you let agents act without giving up control? That means governing access, delegated authority, observability, and where autonomy should stop. That is the problem space we’re focused on with Gravitee Agent Management. It will be fascinating to see how products like OpenAI Dots and Meta Muse accelerate this shift and make the Delegation Economy part of the normal way software is consumed. See link to blogpost in the comments. #gravitee #agentmanagement #agenticiam