Agentic Tech Ops Right Here Right Now: Everything in this video is GA today. Not a roadmap, not a prototype. Agentic tech ops with a decade of banking-grade controls underneath. New Cutover Learning Lab: youtu.be/5D06WxnWlE0?is=0IUO…

At AWS re:Invent, we’re showing agentic recovery live: AI agents supporting a great agentic loop in app recovery runbooks, with controls and approvals as needed. This is all available in GA in Cutover right now. Stop by booth #697 See you in Vegas.

Grateful for the partnership between London Stock Exchange Group and Cutover on resilience. In times like these, opening on time, every time, is not a slogan. It is a discipline. cutover.com/success-stories/… @gocutover

Grateful to @SimonPrinceAI for the free PDF of Understanding Deep Learning. Studying it with Grok as a tutor (asking questions while it has the text open) has been a delight.

Moving the Overton window nicely for us. This shows something we used to treat as pure science fiction is actually possible Huge shout-out to the Tesla teams and SpaceX for making it real. youtube.com/watch?v=cvKIR-c1…

Just the Ticket From Marcus our CPO: In the ITSM world, work starts as a ticket. Change, incident, problem — all of it lands in a record, gets a number, and moves through a lifecycle. It's the foundation most enterprises run on, and for a huge amount of work it's exactly right. The question isn't whether tickets work. It's where they stop working — and what happens to the runbook-shaped hole they leave behind when IT orchestration is what the job actually demands. What the ticket does well For routine work, the ticket carries everything you need: - Scope — what's changing, and to what Steps — the handful of actions required to do it - Approvals — CAB sign-off, governance, the audit position Status — high-level progress the service desk and stakeholders can see - Timing — the change window the work has to land inside Add automation behind it — a pipeline, a script, an Ansible job — and an engineer can execute a standard change end to end without ever leaving the record. Where the ticket runs out of road: Then complexity arrives. The change isn't one system, it's fourteen. The steps aren't five, they're two hundred. They have to run in a precise sequence, across six teams, in a four-hour window, with rollback gates and a hard cutoff after which you can't go back. Or it's a Sev-1 at 2 a.m. and the "steps" are being invented live by people on three continents. At that point the ticket is still the right place to record that the work happened. It is no longer capable of running the work. You can see the tell. A description field stuffed with two hundred lines of copy-paste. A linked spreadsheet with a tab per team. A bridge call where someone reads out step numbers. A chat channel doing the coordination the ticket can't. Status is "In Progress" for three hours — technically true, operationally useless. That's not a ticket anymore. That's a complex operational event wearing a ticket as a costume. Complex events need execution, not a record: A ticket is a system of record. It captures intent and outcome. What it can't do is drive dependent steps in real time, tell forty resolvers which one is theirs right now, hold a gate until a check passes, or show a CIO live progress against the window without someone being interrupted to answer. That's what runbooks do, and it's where Cutover comes in. Every step is a task with an owner, a duration, and dependencies. Automation executes where it should. People execute where judgment is required. AI agents handle the toil in between. Progress is live and self-serve, so stakeholders stop calling the incident manager for updates. The audit trail is a byproduct of doing the work, not a reconstruction assembled two weeks later from chat history. The outcomes follow the model. One leading global bank ran over 100 live incidents through Cutover Respond in the first year and reported a 28% MTTR improvement. On the recovery side, customers see up to 50% reduction in execution time and a 60% increase in audit efficiency. The ticket stays. The execution moves. This isn't a rip-and-replace argument. ServiceNow remains the system of record. The ticket still carries the approval, the classification, the CI links, the closure. Cutover becomes the system of execution underneath it — triggered from the ticket, feeding status back to it, closing the loop. One record. One place the work actually runs. The right tool for the size of the job. For most work, the ticket is just the ticket — and that's enough. For the events where minutes cost millions and one missed step means a failed change, it never was.

Not sure if others have made real progress on this yet: I want the experience of reading a book with an informed LLM as a constant guide: one that handles math notation cleanly, and that maintains an expanding set of notes and context so the conversation deepens as the book does. Basically a more detailed spec for the Young Lady’s Illustrated Primer. 😜 @nealstephenson

30% of leaders would let AI trigger failovers independently— both sides of that are about smart governance in action. Full report here: my.sociabble.com/AfbAM1GGFvd…

Strong opinions, weakly held. There is one camp right now still arguing that all software companies go to zero. Separately, I like a lot of what @dhh has to say about software you build for yourself — it’s a real niche with real uses, it’s super attractive. But I don’t think either is the main path for the enterprise. Take this with a pinch of salt as I sell into the space. What I think tends to work at scale is collaboration. Across teams, across firms. Horizontal software with teams that have experience across the industry, that can actually be supported, and that carries some accumulated trust and reputation. Building everything yourself also means maintaining everything yourself. Most enterprises aren’t structured to absorb that cost indefinitely, and they usually don’t want to. Winning will be about focus — scarce excellence in building the things that differentiate the enterprise, and buying the commodity. It still makes limited sense to build the commodity. On the technical side, harness and model separability feels directionally right to me. Keep the controls for intelligence separate from the supply of intelligence. We’ve had to solve versions of this problem with human colleagues for years — banks figured out a lot of the hard parts long ago. Agents should be treated more like colleagues than black boxes. Consumer tools can keep playing in the consumer space. Headless operation is interesting too — something closer to a client/server model of interaction through a harness, where the parts that matter actually run on a different substrate with the controls you need. What I’m watching is how quickly some people get locked into the context a particular tool provides. Instinct and Muse are early examples. That pattern could get ugly in the enterprise if it’s not thought through. Enterprise is deliberately not winner-take-all. Satya has been clear on this for a while. You want the context and the substrate of action kept separate, not folded into one big dark agent that owns both. A16z has a reasonable point that vendors to the enterprise remain one of the better places to go faster on AI when you can keep a known blast radius and real controls. We were at our AWS EBC last week with Cutover and the consistent feedback from senior people was that ISVs are currently making the fastest progress on adopting new tech. If you understand the blast radius and you’re careful about vendors — Pat Opet’s advice on this is useful — then collaboration can make sense. It stops making sense the moment it turns into large dark agents that aggregate your enterprise insights, your operational learnings, and your training tokens. Curious where other people are landing.

In 2026, vulnerabilities will be exploited in just one day—down from nearly a year! ⏳ Patching is a coordination challenge, not just knowledge. Cutover unites humans, AI, and automation with shared control and audit trails for faster, safer remediation. my.sociabble.com/UTGPJfNGuSZ…

“The harness that sequences work and the action space the agent is allowed to touch must be distinct from the model itself.” - SN That’s the practical control layer. The model supplies capability; the orchestration layer decides sequence, scope, and what is actually permitted. Permissions, boundaries, and the evidence of what happened sit outside the model, not inside it. It’s the same pattern we already apply to humans and privileged automation: least privilege, segmented steps, reconstructable actions, and the ability to interrupt. When the sequencing and the allowed actions are enforced independently of the agent, a single failure or unexpected behavior stays contained and the overall process remains operable. Full note from Satya: xitter.cf/satyanadella/status/21…

Agent Controls Are Operational Resilience Controls Satya Nadella’s weekend note on treating frontier models as insider risks is one of the clearest articulations I’ve seen of the control need for agentic systems. The core principles are familiar from decades of enterprise security: - Separate the supply of intelligence from the authority over it. - Controls, permissions, and the evidence of what happened must sit outside the model. - The harness that sequences work and the action space the agent is allowed to touch must be distinct from the model itself. - An authorized person must always be able to pause or stop execution mid-task. - Every meaningful action must leave tamper-proof, human-readable evidence that does not depend on the model’s own account. - Validation and audit must be independent of the intelligence being assessed. Whilst at the same time we need dynamic orchestration versus static workflow for some of the complexity of tech ops processes in this age. These are the same controls financial services co’s already place around human colleagues and privileged automation: least privilege, segmented responsibilities, full observability, containment, and reconstructable paths. CISOs have been on this mission for years. Applied to agents, they do more than limit downside. They make the overall system more resilient. A single agent failure, hallucination, or compromise cannot silently escalate or erase its own trail. The orchestration layer retains the ability to contain, replay, and recover. In operational environments (releases, incidents, cyber events, complex change), this is the difference between an agent that can usefully act and an agent that can be trusted to act inside a larger process. The principles Nadella outlines—model diversity, continuous verifiability, independent controls and auditability, containment, and incident disclosure—are practical engineering requirements, not abstract safety aspirations. The most trustworthy system will be the one designed so we need to trust the model the least. Full note from Satya: xitter.cf/satyanadella/status/21…

Why not hand the major incident to an autonomous SRE agent? I get asked this a lot. It's a fair question, and the honest answer is more interesting than yes or no. Coding agents got good for two reasons. Vast public corpora of code to learn from, and tests that verify the answer. Major incident response has neither. Nobody ever captured how enterprises actually recover from a complex outage. Logs record system state. They don't record the decisions people made at 2am, the order they made them in, or which ones worked. That knowledge lived in heads, bridge calls and Slack threads, and then it evaporated. There is no natural home outside of Cutover for this operational event. So the training data for "what good looks like" barely exists. The steelman: agents already handle a real class of incidents well. Known failure, known fix, scripted remediation. That share will grow. But the P1 that crosses twelve teams and three vendors is a different problem, and that's the one that costs money. I am positive on the roadmap of these agents but not without a substrate of control, transparency, execution graph and verification. Until the gap between "what the system did" and "what the humans/agents did" closes, autonomous agents will struggle on real enterprise major incidents. One more thing. Even once that record exists, you may want it to sit outside any single vendor's agent. It's your operational memory. Portability, auditability and regulator conversations all point the same way. Question for the SREs and ops leads here: where does the record of what your team actually did during an incident live today? And would you trust it to train anything?

There's no question agentic AI and human agentic orchestration is already making its way into DR plans. Cutover already provides that in multiple governed and controlled ways. Most of the 300 IT leaders we surveyed agree it improves recovery.

Cutover 2026.19 is now GA! This is one of our most impactful releases of the year, introducing AI-powered collaboration within runbooks, stronger governance controls, and enhancements across the platform, Developer Portal, and Connect. 🤖 Agentic Tasks The headline feature is Agentic Tasks, a new task type that lets customers collaborate with AI assistants directly within a runbook. Teams can investigate issues, gather information, analyse findings, generate recommendations, and perform approved actions — while keeping conversations, decisions, and approvals visible and auditable. ✅ Multi-Stage Template Approvals Customers can now configure multiple reviewers and sequential approval stages, helping organisations align template governance with their internal review and compliance processes. 📊 CSV exports from tables Users can export table data directly to CSV, making it easier to analyse and report on operational data. 🛠️ Developer Portal updates New API capabilities support restoring archived users and managing SAML configurations, alongside improvements to user administration, permissions, documentation, and API error handling. 🔌 Connect updates Customers can now configure Event Webhooks for new Connect versions, providing automated notifications when a new version is available and making it easier to keep integrations up to date. 🚀 MCP v1.0.0 We've also officially released Cutover MCP v1.0.0, giving developers and enterprise teams a stable, versioned foundation for integrating Cutover with AI tools and LLM platforms. Links: Release notes are here: help.cutover.com/en/articles… Developer Portal Agentic tasks: developer.cutover.com/agenti… Developer Portal Updates: developer.cutover.com/update… Cutover Connect: developer.cutover.com/cutove… Cutover's MCP Server: developer.cutover.com/cutove…

59 days until AWS re:Invent 2026! At Cutover, AI supports runbooks by automating routine steps while people handle critical decisions, speeding disaster recovery by 53%. Join us in Vegas at booth #697 to learn more!