Sean Regan

How can a multi-national build a strategy around 1 vendor given the continued business risk showing up in AI models? This is the basis of the work we are doing on AI Control Tower that routes work through the right model with full visibility across 1p & 3P AI, machine identities, spend etc. The idea that a LLM vendor sees and stores all queries for 30 days is pretty hard for me to stomach as a technology vendor, or almost any vendor You are essentially handing Anthropic or OpenAI the logic that runs your business. Is there an proxy play an AI platform in the enterprise could play? Could that harness also cut your token spend by routing to the right models and combining deterministic steps and token burning (probabalistic steps) in the right mix? That's what we are building at @ServiceNow and this weekend's shutdown accelerates our mission for enterprises. teams.public.onecdn.static.microsoft/evergreen-asse…

@JayaGup10 Additionally: Why burn a token if you don't need to? Combine deterministic workflows with Intelligence. Not one or the other. Doing so delivers 2 key things. 1.) Lower token cost 2.) Better outcomes by reducing the probabalistic steps to only those necessary.

If they are a large enterprise, ServiceNow AI Control Tower discovers 1st and 3rd party AI, manages it, and maps to business use cases and ROI. The platform also model mixes so that you can run the right workloads through the right model and manage cost. Outside of large enterprises, not sure what people are doing. Inside of large enterprises, this is where we see everyone going. Think about it like a CMDB, if a new device hits your network it is discovered, managed, tracked. Same with new agents, new agentic identities, new AI workflows across any cloud, token spend etc.

Presenting to the C-Suite of a top 30 company globally (200B+ in revenue) on how they can reduce their Anthropic spend. What are some best practices others have seen for some of the largest companies in the world?

@JayaGup10 Working on a white paper that covers this but in the meantime. servicenow.com/products/ai-co…

@lifeof_jer We asked Jer every question this thread left unanswered — starting with: why didn't you stop the agent?" Blog: servicenow.com/workflow/ai/le… YouTube Screenshare of Jer's environment: youtube.com/watch?v=b43PCy… YouTube Interview: youtube.com/watch?v=Z-Lh1N…

Everyone was talking about @pocketosai and how an AI Agent deleted the company but no one got @lifeof_jer to share his screen so that we could all learn form his experience. Until now. I asked Jer to open his laptop and he was brave enough to sit down with me to tackle the toughest questions from X, Hackernews, Reddit, ServiceNow, and Tom's Hardware. Here it is, unpolished and real, the blog, and the recordings. We're still in the Dr. Ian Malcolm/Jurasic Park phase of AI. "So preoccupied with whether or not we could, no one stopped to think if we should" lnkd.in/g94TJEHK As Samual L. Jackson reminded us. "Hang on to your butts!" This is the first of many of these stories we will dive into on @AgentsGoWild give us a follow if you want to see the AI challenges and learn how to avoid them.

AI is the biggest tailwind for cybersecurity therefore it is also the biggest accelerant for cyber risk. In the enterprise, AI needs a harness, it needs to run on a platform with identity controls, automated workflows, and governance from the start. If the CISO at a major bank isn't sleeping, can anyone? The only way to match an adversary moving at machine speed is to defend at machine speed. That's not just a staffing problem. It's an tooling problem. Mythos is going to put this all into motion at warp speed.

This is a compounding crisis. More identities → more permissions → more vulnerabilities → across more apps → exploited faster → by fewer defenders. Unless your teams have tools can sense, decide, and act at machine speed with governance and security built in, they won't be sleeping properly in any version of the future I can see.

"Our security team hasn't slept properly in 3+ months." That's a CISO at a major U.S. bank. "Take whatever number of people you thought might be in jobs related to AI deployment in the enterprise and multiply it by 10. Then probably 10 again." That's @levie with a clear eyed take on AI Here's why nobody in security is sleeping. A thread. 🧵

@BoringBiz_ @mathewlodge

Just spoke with a couple friends who work on the operational side at very large private equity firms Every PE firm is now scrambling to find and recruit AI talent who can implement the newest tools into their PortCos This includes former and current technology executives who are familiar with the industry, as well as engineers who can drop directly into the business to build custom made tools Will keep the comp numbers confidential but they are absolutely insane, especially at the senior levels For the PortCos that have already piloted this, the efficiency gains are huge. Some are well within the range of 20%+ headcount reduction potential for back office functions

"Lock the agent in the room and let it party" ....... 8 a.m. at the @ServiceNow + @nvidia Session. OpenClaw for the Enterprise.......Agents Gone Wild.........Kill Switches........ What is happening at Knowledge26? Enterprise AI is moving faster than people might typically expect. Yesterday ServiceNow released "Australia" and put 4 superpowers in the hands of Enterprise IT to enable AI safely and quickly. They put a kill switch in AI Control Tower and connected it to all of your agents, data, identities, devices, and enterprise context. Now a room of the largest companies in the word is exploring what only startups could even imagine building 12 months ago..... "the enterprise harness for long running, self-evolving, autonomous agents named after crustaceans". A LLM did not create this post, nor did it name this. I'm fairly certain those words have never appeared together in a sentence before. Day 1 Wrap up Here. lnkd.in/gikGH_v8

@Taakkeshi @theaiportfolios @grok No, it is not. I know the humans who filmed it and where they filmed it and on which day they filmed it. @grok You are welcome.

@theaiportfolios @grok is this video AI generated?

Boris Cherny, Head of Claude Code, on the importance of ServiceNow $NOW

My favorite slide from ServiceNow’s 2026 Financial Analyst Day. The build-it-yourself trap remains for enterprise software. AI is certainly making it easier to put together a basic working demo for simpler software for smaller firms. However, for more advanced software in larger enterprises, there is much more to consider and implement, making the eventual total cost of ownership much higher.

A "Toll" can be seen as a gate. Some tolls feel like a tax and some feel like an express lane you happily pay. It depends on the value delivered. I wouldn't think of Action Fabric as a gate. It's a runtime. Before Action Fabric, outside AI agents could read ServiceNow data, but they couldn't actually execute work: trigger workflows, route approvals, enforce business rules, maintain audit trails. As we saw with PocketOS and others in April, that runtime for action is pretty valuable. (Jer will be with us on Thursday here at Knowledge 26 BTW) Action Fabric makes it possible for any agent, whether it's built on ServiceNow, Claude, Copilot, or something homegrown to execute work in a governed runtime. The consumption model is the same "Assist" currency that already applies to human-initiated AI actions on the platform. Extending it to agent-initiated actions is how you get a single, predictable model across every workload, human or AI. The alternative is agents executing ungoverned work with no visibility, no metering, and no accountability. That's what I'm going to start looking at via @AgentsGoWild as a little side project. Every headless action that runs through Action Fabric gets identity verification, permission scoping, and a full audit trail through AI Control Tower. Some might see this as paying for access. Others will see this as paying for governed execution on the platform that already runs their business. That's closer to how AWS charges for compute or how OpenAI charges for API calls. So if a governed runtime for agents of any stripe are valuable, customers will be happy to open the gates to ServiceNow's AI platform.

@LauraBratton5 Typo on my part. Action Fabric. Yes I watched the keynote today and the investor day yesterday. It’s not a tollgate. It’s a way to monetize the set of entitlements known as a “seat” which gives you software plus usage of AI.

This is poor framing. ServiceNow is not putting up a tollgate. They currently sell seats, yes? Is that a tollgate? A “seat” is an entitlement that lets a user interact with the software. Agent Fabric is just a layer that allows agents to do so in a way that can be monitored. Otherwise you have no way to track the agents. No way to track usage. No way to govern what the agents can and can’t do. Given that the underlying software runs on AI and expends tokens, it’s important to measure it, isn’t it? The same way AWS measures usage? The same way OpenAI measures usage of its API? The same way your electricity company measures your usage? This is no different. @LauraBratton5