---
title: "Recovery Reimagined for the AI Era | SHIFT 2026 On-Demand"
type: "WebPage"
language: "en-US"
url: "https://www.commvault.com/resources/video/recovery-reimagined-for-the-ai-era"
date: "2026-11-13T13:26:51-05:00"
modified: "2026-11-13T13:26:51-05:00"
description: "Discover how AI is transforming cyber recovery, accelerating business restoration, and strengthening cyber resilience. Watch SHIFT 2026 on demand."
image: "https://www.commvault.com/wp-content/uploads/2026/11/Recovery-REimagined-for-the-AI-Era_20261118-1434.png"
breadcrumbs:
  - name: "Commvault Home"
    url: "https://www.commvault.com/"
  - name: "Resources"
    url: "https://www.commvault.com/resources"
  - name: "Video"
    url: "https://www.commvault.com/resources/video"
  - name: "Recovery reimagined for the ai era"
---

# Recovery Reimagined for the AI Era • SHIFT On-Demand

- [Home](/)
- Videos
- Recovery Reimagined for the AI Era • SHIFT On-Demand

ON-DEMAND SESSION

## Recovery Reimagined for the AI Era

See how AI-powered, full-stack recovery can help organizations restore cleanly, limit disruption, and prove readiness before an attack.

[Request a demo](/request-demo)

Play Video      ![Video thumbnail](/wp-content/uploads/2026/11/Recovery-REimagined-for-the-AI-Era_20261118-1434.png)

---

- [Key Takeaways](#Takeaways)
- [About This Session](#About)
- [Related Resources](#Resources)
- [FAQs](#FAQs)

[Back to all on-demand sessions](/shift#watch-now)

---

## Key Takeaways

**Prove Recovery Before It Matters**

Test clean recovery regularly and show evidence that critical systems can return on time.

**Recover at Cloud Scale**

Use elastic, parallel processing to restore massive cloud environments faster.

**Restore Only What Changed**

Recover affected data and agent changes without disrupting healthy operations.

**Protect the Full Stack**

Recover identity, infrastructure, applications, data, services, and AI agents together.

**Limit the AI Blast Radius**

Control what agents hear, see, and can do across connected systems.

**Close the Recovery Loop**

Unify discovery, detection, clean recovery, and continuous improvement at machine speed.

---

## About This Session

**AI is accelerating attacks and operational mistakes, forcing organizations to rethink how they prepare for recovery.**

This session explores what it takes to recover fast, confidently, and cleanly when trusted tools, identities, and infrastructure may be compromised. Leaders explain why documented plans are not enough and why organizations need tested proof of recovery readiness.

#### Recovery at Cloud Scale

See how elastic cloud architecture, service-specific optimization, precise data selection, APIs, and infrastructure as code can accelerate recovery. The session also demonstrates AI-assisted planning that validates configurations and helps teams execute instead of rebuilding recovery plans manually.

**Modern recovery must restore the entire business service, not merely its database, virtual machine, or backup copy.**

Learn how full-stack recovery coordinates identity, cloud configuration, networking, applications, data, SaaS services, and AI agents in the right order.

#### Control Agentic AI Risk

Follow a retail scenario where an AI agent changes thousands of product records. The demonstration shows how teams can audit agent actions and roll back both the agent and affected data while preserving legitimate business activity.

**Cyber recovery must be clean, isolated, repeatable, and proven continuously before an incident occurs.**

Discover how threat scanning, clean recovery points, isolated testing, automated runbooks, and readiness reporting help teams restore trusted operations with confidence.

#### Build Continuous Resilience

Connect security and recovery in one loop that discovers what matters, detects changes, recovers cleanly, and improves after every test.

CONTINUOUS CYBER RECOVERY

## CyberRecover

Restore clean, verified operations after destructive cyber incidents.

[Explore solution](/solutions/cyber-recover)

CONTINUOUS CLOUD RECOVERY

## CloudRecover

Recover cloud data with speed and precision.

[Explore solution](/solutions/cloud-recover)

---

## Related Resources

SOLUTION BRIEF

## Cyber Resilience

Explore fast, clean recovery with validated data and isolated testing.

[Read solution brief about Cyber Resilience](/resources/solution-brief/cyber-resilience)

VIDEO

## Complete Cyber Resilience

See discovery, identity recovery, and cleanroom recovery in action.

[Watch video about Complete Cyber Resilience](/resources/video/from-discovery-to-recovery-the-complete-cyber-resilience-solution-in-minutes)

blog

## Protecting AI Workloads

Learn how organizations can achieve resilience in the AI era.

[Read blog about Protecting AI Workloads](/blogs/how-can-organizations-achieve-resilience-in-the-ai-era)

---

## Frequently Asked Questions

Why must recovery change for the AI era?

AI accelerates attacks, mistakes, and autonomous actions. Recovery must operate continuously and restore trusted systems, data, identities, configurations, and agents together.

What makes cyber recovery different from disaster recovery?

Cyber recovery assumes connected systems may be compromised. It validates clean recovery points, uses isolated environments, and proves the restored application works before returning to production.

How can organizations prove recovery readiness?

Teams can regularly test runbooks in an isolated environment, validate business outcomes, measure recovery time, correct failures, and generate readiness reports before an incident occurs.

What is full-stack recovery?

Full-stack recovery restores the dependencies a business service needs, including identity, cloud configuration, networking, applications, data, SaaS services, permissions, and AI agents.

How can teams recover from harmful agent actions?

Teams need visibility into what an agent changed and the ability to roll back the agent and affected data together without undoing healthy business activity.

---

## Transcript

View full transcript

Good morning everyone, and thank you for taking the effort to come down here to meet with us. So I was in India recently talking to customers about the impact of AI on mission critical industries. And there is one number that stuck with me, 45 minutes. That’s all the time one of India’s largest financial organization has to recover from a cyberattack.

You heard it right, 45 minutes. This organization plays a critical role in India’s stock market. It keeps track of who owns what. Millions of investors depend on it. If its systems go down, the impact goes far beyond IT. Now this goes beyond just one organization. What typically starts as a rule in one region can become the standard for regulated businesses across the world.

Now, even for conventional DR, 45 minutes is a very demanding target. Now, think about what AI does to that clock. As Sanjay just discussed, AI is changing the way attacks happen. Attackers can move faster with a level of automation never seen before.

But it’s not just the bad actors that put us at risk. We are giving autonomous agents identities in our business. These agents have passwords, access keys, certificates, access to our data, and most importantly, the ability to take action on their own. That creates a whole new set of challenges. And we are dealing with these challenges right now.

Now let’s talk about another large global financial organization. Millions of people trust this company with their retirement savings, investments and financial future. They have enormous resources, great technology, and world class technical teams. But they too face the same set of challenges. Now, as they’ve worked through their cyber recovery plans, they realized some of the assumptions on services post attack were not true.

They assumed that after an attack, they could still serve their customers using their phone systems. So as we talked, we realized that those phone services ran on the same infrastructure that is part of the recovery plan. Now imagine that moment. You are under an attack. You can’t contact anyone in your team as the phones are down.

You try to log in. But the PAM systems that secure your access is down, and you can’t trust anything connected to your network. The very tools you need to recover need recovering themselves. You had a recovery plan, but the recovery plan didn’t survive the attack. Can you trust the identity system? Which backup is clean? Where do you recover to? What do you bring back first? And how do you communicate with each other?

These are not questions you want to answer for the first time when an attack happens, especially when the clock is ticking. Recovery is not just about speed. It’s also about bringing back clean data. Now this brings me to the next conversation that I had with yet another customer. This is one of our US federal government customers.

This organization supports some of the most important functions of our government. Now they’ve been talking with some of the leading AI companies and the future risks of AI. And they were preparing for a pretty dire situation. What if AI makes cyberattacks so fast and so widespread that your security tools can no longer defend you, that you can no longer trust anything in your connected environment?

What if all your network and all connected devices are compromised? What do you do then? Well, their answer surprised me. Backups to tape. Yes, you heard it right. Backups to tape. Okay, here we are talking about the future of AI, and we are going back to tape drives. And they are looking at storing this tape drive in a network isolated environment that is not connected to anything.

Well, I don’t think the important part of this story is tape. It is the depth of planning that some organizations are going through to counter the threat of AI. They are asking a very simple question: if everything is under attack, what can we still trust? How do we know our data is clean? How do we make sure we can recover, no matter what happens now?

Tape is one answer, but it is not the only answer. The real goal is to have a clean, safe place to recover, even when everything else around you has been compromised. So three different customers, three different challenges, but one common goal. Being ready to recover no matter what happens. And that is what cyber resilience is all about. For me it comes down to three simple things.

Recover fast. Recover with confidence. Recover clean. But here’s the challenge. You can’t wait until an attack happens to figure all this out. You have to plan. You have to practice. And you have to practice daily as threats change daily. You have to test recovery in a safe, isolated environment and make sure systems come back up.

Make sure your data is clean. Measure how long it takes. Find out what’s broken and fix it and repeat. That is the foundation of ResOps. We all believe AI makes attacks more likely and more complicated. And if attackers are using AI to move faster, we can’t expect our recovery teams to do recovery manually. We need AI to make those decisions, and we need AI to orchestrate the recovery process.

That’s how we close the gap between an attack and recovery. We use AI to close the recovery loop. At Commvault our mission is clear. Help our customers build recovery they can practice. Recovery that operates at AI speed. and most importantly, recovery they can trust even when the systems around them can’t be trusted. So let me come back to where I started.

45 minutes. Imagine your CEO walks into your office tomorrow and ask, can we recover in 45 minutes? What would you tell them? What would you show them? Not a document that says that you have a plan. You need proof that you have tested clean recovery, and you are ready. And proof that you can recover when it matters most.

That’s the challenge our customers face today. So how do we do this at scale? How do we solve this at scale? That’s exactly what we’re going to talk about next. Woon, our CTO for Clumio, is going to walk us through the key ideas behind recovering at scale in the cloud. And then Pranay will show us how we are bringing data security, identity, cyber resiliency and AI recovery together. All to help our customers close the recovery loop.

So let’s keep moving. The stage is over to Woon. Thank you. Good morning.

Rajiv asked can we recover in 45 minutes? Let me take that question into the cloud. Imagine an application that holds tens of billions of objects in a single S3 bucket. Hundreds of billions of items in a DynamoDB table. Dozens of managed services spread across accounts and regions. And changes with every deployment. What does 45 minutes mean for that application?

At Commvault, my Clumio team focuses on making those applications resilient. Let me explain what that requires. First, recovery has to keep up with the scale of the cloud. Customers take full advantage of the cloud’s elasticity, and they build truly massive applications. And when something goes wrong, can recovery handle that? How quickly can you get started? If you spend the first 30 minutes provisioning recovery infrastructure, guess what? You’ve already used most of Rajiv’s clock.

That’s why we built around the cloud’s elasticity. The moment recovery starts,Clumio fires up. Hundreds of lambda functions or cloud functions, divides the work into smaller tasks and schedules them in parallel. And them work when the work finishes, that compute scales back down to zero. But elasticity alone isn’t enough. Amazon S3, DynamoDB, and Google Cloud Storage each have their own APIs, limits, and performance characteristics.

So we built optimizations specific to each service, from how we distribute the work to how we move the data. That is a substantial amount of engineering effort. But that is how we turn the cloud capacity into recovery throughput. In one customer recovery, Clumio restored 700TB of data in just two hours. That is what this architecture makes possible. Second, executing quickly matters, but executing only the necessary work matters just as much.

Your application might hold petabytes. But the incident might affect only a few gigabytes. A script overwrites one prefix in an S3 bucket. Or an AI agent corrupts a subset of records in a DynamoDB table. Why recover the entire data set to repair the damage? Clumio gives you precise control over what you recover, when you recover from, and where you recover to.

You can select the affected objects, a specific prefix, or a DynamoDB partition. Choose the point in time before the damage happened. And recover to the destination you need, including another account or region. And with Clumio Backtrack, you can recover the affected data directly in place. That means you recover the affected data while preserving the good data around it.

Combine that precision with recovery at scale and you have two ways to save time. Execute faster, and recover only what’s needed. Third, recovering data is only part of recovering the application. Your application also depends on configurations, permissions, networking and services that connect it all together. If the data comes back quickly but your team spends hours rebuilding those dependencies, the business is still down and Rajiv’s clock is still running.

This is particularly challenging in the cloud. Every cloud offers several flavors of key value stores, tens of relational databases, and multiple kinds of compute and object storage. The list goes on and on. Developers assemble applications from all of it, and they deploy and change those applications through infrastructure as code and automated pipelines. Recovery has to fit into the same way of working.

That’s why we designed Clumio CloudRecover from the beginning, to be headless, driven entirely through APIs and Terraform. The same pipeline that rolls out that new environment can configure protection.

No one works through a console. Fewer manual steps in between the incident and then the fix. Fast, precise, and automated.

But the problem is much harder. And there’s a difficult last mile problem. APIs let us automate the recovery step, but most of the time someone still has to manually drive the recovery process because of configuration drifts, application specific requirements and context. And this causes the automated process to slow back down to human speed.

We have the right foundation today. Recovery at scale. Precise recovery and control through APIs and code. And now we’re building AI-assisted recovery on that foundation. Available today in early access. An AI agent will validate the configuration, will align the requirements, and present the plan for your team to confirm so that the team can execute the plan instead of reconstructing one. Because Rajiv’s clock stops when the application is working again.

Let me show you what that looks like in action. This is an expenses app. 100 million receipts, all tracked in DynamoDB, with their images stored in an S3 bucket. Clumio backs up both daily. Then one morning, over 50 million receipts are gone. DynamoDB is missing the records, and the images are missing from S3 too.

Sales should have 87 million receipts, but it’s down to just 37 million now. So we ask our agentic harness in plain English to bring them back. With the Clumio remote MCP server it checks the DynamoDB table and S3 bucket against our backups, and finds the missing records and images in last night’s backup. Before it changes anything, it shows us the plan.

We say go ahead, and Clumio restores everything straight back into the original DynamoDB table and S3 bucket. And because Clumio restores in place, the app picks them up right away. No new DynamoDB table, no new S3 bucket, and nothing to reconfigure. And the MCP confirms it. Every missing receipt is back and the totals match the backup. Your agent asked, Clumio restored, and the app never had to change.

All without ever opening the Clumio UI. That’s recovery with the Clumio remote MCP.

Next up on stage, Pranay will show how we’re bringing data, identity, security, and recovery together in the AI world to close the recovery loop. With that, please welcome Pranay.

Thank you, Woon.

Now, Rajiv told you about three real customers. I want to start with one that isn’t. Meet Acme, a retailer. And every challenge it’s about to face could be yours. Its storefront runs in the cloud, its system of record in its own data center, and it takes tens of thousands of orders a day. And like most of you, it’s putting AI agents to work across all of it.

Now, picture an ordinary Tuesday. Acme’s merchandizing team uses an AI agent to keep prices and promotions current across the online store. That morning, someone added a new clearance rule for winter items to the agent’s system prompt. It’s routine stuff, but the agent reads that rule far more broadly than anyone intended. And within minutes, 2400 products are priced to zero or gone from the store entirely.

Now, nobody attacked Acme. There was no ransomware, no breach. One well-meant change from someone just trying to do their job and it turned into a serious business problem. So think about what your team does next. How do you stop the agent from doing it again? How do you undo what it changed without rolling back everything else around it?

Now, Sanjay said, recovery gives you the confidence to say yes to AI. Acme’s Tuesday shows how hard it is to get this confidence. Now, from my perspective, three things have dramatically shifted. First, where decisions get made. They’re moving out of systems of record and into systems of intelligence, where agents decide and act on their own. Acme’s agent didn’t check with anyone, it just acted.

Second, the unit of recovery has changed. Recovery isn’t about data anymore. It’s the whole application, the infrastructure and configurations, the identities, the data and application services, and now the agents. You can’t rebuild all of that by hand. And third, the economics of security are changing. Now we are spending $250 billion on cybersecurity every year. Yet AI enabled attacks are up 89%, and attackers can move through an environment in just 29 minutes.

And AI is making it cheaper every day to find another way in. Just think about the economics. Cyberattacks are a function of two things. The likelihood of a breach and the damage it causes. We have spent decades trying to reduce that likelihood, and now we have to get equally serious about reducing impact. Keep prevention strong. But the smartest place for your next dollar might be AI recovery. And by the end of this talk, I hope you’ll see why.

Now put those three things together and resilience needs a new paradigm. Security and recovery can’t live in separate worlds anymore.

Our answer is one continuous loop. Discover what matters. Detect what’s wrong and recover clean. And get bette every time you go around. And it comes down to three questions you need to answer.

What must we protect? How do we limit the damage? And can we recover cleanly and in time?

Now I’m going to walk you through all three and we will follow Acme the whole way. So the first question what must we protect? Now most teams would say: our data. And that’s where the trouble starts.

Acme’s business doesn’t run on data. It runs on one application: Order Hub. When it’s down, Acme isn’t selling. And protecting it takes two things. Knowing what you actually have and knowing how protected it really is. Let’s start with what Acme has. Order Hub is a lot more than a core database. Its storefront runs on Azure. The order services run as containers on AKS.

The orders database system of record sits in Acme’s own data center feeding the ERP, and it’s people work in Microsoft 365 with every receipt stored in SharePoint. After an attack, the data is useless. You need the network configuration, the keys in Azure Key Vault, and above all, your identity providers. If you don’t recover Active Directory and Entra ID first, nothing else comes back.

That’s a hybrid estate, and protecting all of it is one of our biggest strengths. One platform to see it all and bring it all back across cloud, SaaS and on-premises. And seeing it all starts with your most sensitive data. So Commvault finds it across the whole estate: PII and sensitive information and anything Acme defines with its own rules.

Now Acme knows what it has and what matters most. Which raises the harder question how protected is Acme really? It learns three things. First, Acme hasn’t protected its key vault keys or network configuration. Second, some workloads may rely on cloud snapshots. And for Acme, those copies aren’t air gapped or immutable, so an attacker who gets into the account can just delete them.

And third, Acme has never tested recovery for its identity systems or most applications. Acme is more exposed than it thought. Commvault shows every gap and keeps track of each one as Acme fixes it. Now that’s the estate we’ve protected for years. What’s new is AI. And most companies haven’t mapped it at all. Ask Acme how many agents it runs in production, and the answer is three, all on Microsoft Foundry.

Commvault runs discovery and finds twenty. Seventeen more that the teams have built on their own. Some on a Friday afternoon with real credentials and write access to Acme. Acme protects the three it relies on, and triages the other 17. But the business logic is not centralized on any single application anymore. On employees laptops, desktop agents now run skills and workflows and act on their own. Commvault flags the one nobody has approved, and protects the skills and workflows people build and Claude today, with OpenAI and Microsoft Copilot coming soon.

And when an agent goes wrong, the damage doesn’t stay inside the agent. It writes bad data straight into your systems. Now, most teams can’t see exactly what the agent touched, so they roll everything back to yesterday and lose a day of real business with it. So today we are announcing Commvault AgentRecover with built in AI auditing and governance.

It shows you every agent on supported platforms what it’s connected to, including tools, MCP servers and data, and what it changed. And it lets you roll back the agent itself and the impacted data without rolling back everything else. We’re starting with Amazon, Bedrock, AgentCore and Microsoft Foundry, and more platforms are on the way.

Now let’s return to that Tuesday. The clearance rule went into the system prompt. We know what the agent did. What matters most is what happens next. AgentRecover shows Acme exactly what happened. The system prompt change and every catalog item it changed after. Now you fix only the data and the agent breaks it again. And you fix the agent and the catalog stays broken.

So Acme rolls both back together, the agent to its previous prompt and the catalog to just before the change. And the order system has never changed. So not a single order is lost. Prices are right, products are back, and the agent is working again.

Now Acme undid the agent’s mistake and the rest of the business never stopped. That is the difference between rolling back data and recovering the business. Remember, that agent belonged to one trusted internal team and it did all that damage with access it already had. Acme assist is different. It’s external facing, and it handles transactions and PII data all day.

And that brings us to the second question. How do we limit the damage, before there’s anything to recover?

Now CIOs want to give agents more access because that’s where the value is. And CISOs want to give them less because that’s where the risk is. You can’t choose between value and safety. So you have to limit the blast radius. Take any request coming into Acme. It touches the agent, the model, its memory. It reaches into your data and application services and to act assumes an identity..

At every step the blast radius gets bigger. So you need control at all three layers. What an agent hears, what it sees and what it can do. Start with what an agent hears. Maria is an Acme customer, chatting with Acme assist. Her payment just failed. So she does the obvious thing. She pastes her card numbers into the chat.

She’s done nothing wrong. But now the number could end up in the model, the logs, the agent memory. And once the agent has heard something it can’t unhear it. So Acme needs two things. First, a record. Agent Trace captures what Acme agents have heard, said, and did. When legal or a regulator asks what an agent saw, Acme has the answer for audit and e-discovery.

And second, and even better, stop it before an agent hears it. Coming soon. Agent Guard applies Acme own rules, PII, custom patterns and even regular expressions to every AI conversation and agent call. Now that covers what an agent hears.

Sorry, I need a sip of water.

Yeah, okay. Now meet. John.

Now, unlike Maria, he knows exactly what he’s doing, and he’s going after the second layer. What an agent can see. He pretends to be Maria and asks Acme assist for her orders. Now, if an agent believes him, it hands over the address, the purchases, and her payment details.

No hacking, just a simple conversation. Now security teams call this a confused deputy attack.

A prompt guardrail won’t stop it. You have to enforce access at the data itself. And that’s exactly what Commvault Data Access Manager does, down to the row and column. So Maria’s records stay hidden, and John sees only his own orders, with payment details masked. You define a policy once, and it’s enforced across your databases, warehouses and data lakes. Every access is logged and your applications don’t change.

Now, John was trying the front door. The bigger risk is the door that nobody knows is open. That’s the third layer and the most dangerous. What an agent can do. And that comes down to identity. Every agent needs an identity to do its job. Just like an employee has a login, the agent has an account.

We call these non-human identities, and they’re exploding. Spread across Active Directory, Entra ID, and Okta, each with its own permissions. Almost nobody can keep track of who can do what. Now, in every company, a handful of accounts control everything else, who can log in and what every other account is allowed to do, etc.. Security teams call them tier zero.

You get into a tier zero and an attacker doesn’t have one system. It has all of them. Now Acme Service Desk agent has one account in Active Directory and another one in Entra ID. Neither one was meant to be powerful, but with a simple permission change, either one can end up in tier zero. That’s why today we are announcing Commvault Attack Intelligence.

Now it discovers everything in tier zero across Active Directory, Entra ID, and Okta for people and agents alike, and it maps every path that leads there. It detects the moment tier zero changes and it helps you roll back risky changes at scale. Discover, detect, and recover applies to identity. Now here’s how tier zero expands without anyone noticing. Late on a Friday, IT takes a shortcut to fix an access problem.

To fix it, the admins make a few group changes and by accident they give that service desk agent admin rights. Now nobody gave those rights to the agent directly. It inherited them through a group inside a group that is almost impossible to spot. But one small change and the agent can now take over the whole network. Commvault shows exactly who made the change when they made it, and what it looked like before and after.

Then it shows why the agent is now tier zero across Active Directory and Entra ID. That’s what makes it so hard to spot. Your identity systems are connected, so a small change in Active Directory quietly spreads to Entra ID. And agents don’t have checks people do, like multifactor sign in. So whoever takes over that agent now holds the keys to the business.

That is how ransomware ends up on every machine at once. Arlie, our AI assistant, explains the risk and recommends undoing the change. The administrator approves, and the change is reversed and the path closes. On the dashboard, Acme tier zero exposure drops back to where it was. Nobody used that path. Now notice what Acme did. It controlled what its agents hear, what they see, and what they can do. And it reversed only what went wrong.

That is how you limit the damage before there is anything to recover. And that brings us to the third question. Can we recover cleanly and in time? Now Acme has done everything right on prevention, XDR, network security, identity protection. But even the best prevention won’t stop everything. Most teams turn to disaster recovery.

But DR was built for outages, where you can trust your own systems. In a cyberattack, you can’t. And replication copies the attack straight into your recovery site. What you need is cyber recovery, and that comes down to three things. First, it needs to be clean. Second, it needs to be full-stack. And third, it needs to be proven before an attack, not after it.

Now let’s start with clean. Your EDR is watching what is running live. It isn’t scanning the backup you took three days ago. If you restore an infected backup, you just restarted an outage. That is what Commvault Threat Scan is for. When your security team finds out the new threat, the first question is, is it in our backups, too? And how long has it been there?

Threat Scan answers that in three ways. Our own entropy detection spots signs of encryption. Hyper threat hunting checks backups against Google threat intelligence for known malware, and deep inspection looks for threats that nobody has seen before. Together, they help you identify the last known clean backup and how far back to go. And today we are bringing Threat Scan to Microsoft 365.

For Order Hub, it shows which files were hit and when the infection likely started now. But a clean copy is only the start. You still need to bring the whole application back. And that’s the second piece full stack. A full stack recovery is hundreds of steps, across teams, services and now agents, in strict order, and often when you’re starting from nothing.

That’s why today we are announcing Commvault Recovery Studio.

Super happy about this one.

It’s a new way to orchestrate full stack recovery across cloud, on premise and SaaS. Now Woon talked about infrastructure as code. Think of a runbook as recovery as code. Acme’s runbook for Order Hub captures every system, in order, and every step that used to live in somebody’s head. And Acme uses the same runbook everywhere: to test in a Cleanroom, to rebuild a recovery environment when an attack is real, and to bring Order Hub back to production.

You define it once and it becomes a repeatable blueprint for recovery. Most teams believe recovery means bringing back the database or the VM. It doesn’t. Recovery Studio is designed to recover the full stack starting with the foundation. Cloud Rewind rebuilds the cloud configuration first, including the network keys Acme found unprotected earlier. Then Active Directory Forest rebuilt clean. Then the data and the application services, and finally the agents.

Your runbook isn’t a document anymore. It runs with approvals where you want them. Now Acme puts it to the test. Recovery studio runs Acme’s runbook in Commvault Cleanroom, isolated from production, using a clean copy from Threat Scan. And then it checks what the business cares about. Does the storefront load? Is the data right? Can a customer place an order?

And only when all of that passes, is Order Hub ready for production. Order Hub is back clean in about three hours. Three hours. And that’s where Acme starts. And it’s a real number. The vast majority of teams have never tested a full recovery. From here, Acme can optimize. And it sees exactly where the time goes: streaming data back and rebuild the rebuilding the Active Directory forest.

Now from a standing start, Acme will never hit Rajiv’s 45 minutes. So what if you didn’t have to start from scratch? We took the best idea from DR, failover, and made it safe for a cyberattack. We started with Active Directory, and soon we’ll be extending it to the full runbook. Acme keeps Order Hub recovered ahead of time, on standby. So when an attack hits, it fails over in minutes.

Well, inside the 45, to a copy it’s already proven clean. That’s not just faster recovery. We have moved the whole work before an attack. And that brings me to the third piece. Proof. Acme changes every day, and every change can quietly break the runbook and the standby with it. A recovery you tested last quarter proves nothing about today.

So Acme uses AI Studio to build an agent. It’s job: test Order Hub’s recovery every night. If something breaks, it shows the team what failed and why. The team tunes the runbook and the agent tests it again. By morning, the readiness report is in the inbox, the runbook works, the standby is clean, and Acme knows it can recover. Proven on a quiet night, not in the middle of an attack. That is proactive recovery.

Now let me bring it all together. We started with three questions. And you watched Acme answer all three. It found agents nobody knew existed. When a pricing agent went wrong, it rolled back both the agent and the damage without losing a single order.

It kept customer data out of the wrong hands and closed a path to its most privileged identities before anyone could use it. And it didn’t just restore Order Hub. It proved the whole application works every night. These weren’t three separate stories. They were one loop. Discover what matters. Detect what’s wrong, and recover clean. That is security and recovery on one platform.

And I believe the most exciting part is still ahead now.

Security is becoming AI-first this year. Every major security platform has put AI agents into the SOC. When there is a fix, these agents act on it. They block the attacker. They isolate the machine, apply the patch at machine speed. But what if there is no fix yet? Or what if the damage is already done? Then the only way forward is going back to the last version you can trust.

And that takes what recovery knows what changed, which copy is clean, and how to rebuild the whole application in the right order. For years, SOC and the recovery team have lived in separate worlds. But in an AI-first world, security isn’t complete without recovery.

Now here is our vision. Picture a Tuesday at Acme not so far from today. A routine software update carries a flaw nobody knows about yet, and lets an attacker in. Acme’s security agent catches it in seconds. But here’s the part I love. It doesn’t open a ticket. It talks to Commvault, agent to agent. Together they work out what changed, and do two things at once.

They spin up a Cleanroom so the security team can run forensics. And they fail Acme over to the clean copy its agent tested that night at two. Now, just think about what happened. There was no fix yet. A patch might have taken days, but the business didn’t wait. Agents had it back in minutes. Within the policies Acme had already set. That’s the future we are building towards. Security and recovery agents closing the loop together at machine speed.

Recovery stops being the last step. It becomes part of defense. And everything we have announced today in security and recovery lays that foundation. Which brings me back to the economics. Now, you can’t patch every vulnerability before somebody exploits it, but you can decide what happens to your business when they do. That’s why I open by saying that the smartest place for your next dollar may be AI recovery.

Let me take out the “may”. Prevention lowers the odds of an attack, but recovery decides what it costs you. So here is how Acme’s story ends. The first Tuesday nobody attacked Acme. This Tuesday, somebody did. And its customers never noticed. Making that ordinary is what I think about every single day, I really do. So you can say yes to AI and still have a quiet Tuesday.

Thank you very much. And enjoy. SHIFT, ladies and gentlemen.
