1:01:11 Webinar
Ask Us Everything About Pure1 AI Copilot, MCP Server and Fusion
This month we’re diving into how AI and automation can help infrastructure teams simplify storage operations at scale. See how Pure1® AI Copilot, the Pure1 Manage MCP Server, and Everpure Fusion™ work together to help you understand what’s happening across your environment, connect AI tools to operational data, and simplify management across your fleet.
This webinar first aired on 18 September 2026
The first 5 minute(s) of our recorded Webinars are open; however, if you are enjoying them, we’ll ask for a little information to finish watching.
Click to View Transcript
Again, my name is Kevin Rixon. Welcome to yet another Ask Us Everything session. Hey, you know, it's, it, it's a Friday, and we're always happy for Fridays, but the best thing is that we've got a great show for you here today. We're gonna be talking about AI Copilot, but frankly, the main thing that we're gonna be
talking about is how you can leverage AI in general within your Everpure systems, and just helping make the management of your data and your storage that much easier. So with me today we have three very important guests. We have Anand Madhavan, who is our director of product management. We have Ameeral Shaw, who is the, product manager on, Pure Copilot, and Russell Cagle,
who's a tech lead on, Copilot. So welcome, gentlemen, and we're so happy that you're all here. Thank you. Thank you, Kevin. Nice to be here. Absolutely. All right. Well, let's do a little bit of housekeeping.
As you know, the way Ask Us Everythings are set up, the entire content is driven by all of you. So get into the Q&A and, put your questions in. At the end, if we go to the next slide, we can see that, the, main area that we're going to, focus in on or, I mean, the, we're, we're going to have a
drawing for the f- folks who actually pose questions. So you get a chance to get some really cool prizes, choice of a trucker hat or a really cool Carhartt and Everpure co-branded lunchbox. So, please, you know, again, the whole idea of Ask Us Everything is you all are asking the questions and we're answering, and we've got a great prize for you if you, if you enter those
questions in. So with that, I think now we'll turn it over and do our first poll, 'cause we wanna gauge where everyone is in their Copilot usage these days. So if we can put up that poll and give everyone a minute to, take a look at, how familiar they are with AI Copilot.
If you're using it, what are you using it for? And then the stretch question, are you using it or looking to use it with any other AI, that you're using in your organization? Okay. So with that, I would like to s- kick off our discussion, with our three esteemed guests.
And for our first question, Anand, it's great that Copilot can answer, you know, all sorts of different questions and stuff, but, you know, can you tell me a little bit more about what it really does, what it was designed for, and, and, and, and what it, what it does? Give us a quick introduction on that.
Yeah, sure. So, I think it's best explained with, with kind of a little bit of an infographic here. The, the core DNA of Everpure is simplicity. And, and we have tried to bring simplicity, via Pure1 Manage, over the years through our Evergreen architecture, Fusion and, and so on and so forth, right?
And our AI Copilot on Pure1 extends that journey towards simplicity. And it's the perfect combination of simplicity with i- with intelligence, right? And what we mean by that is knowledge and information can be very scattered, across, you know, various different systems and knowledge base and your, your fleet, technical resources, user guides, et cetera, et cetera.
And last year we launched the Copilot, to, to help customers to get to the answers quickly and reliably through a single interface where they don't have to remember all the different CLI commands. They can just ask the question in the, Copilot and, they can get the answers. So, so that's the core kind of value proposition of AI Copilot.
It is more than just getting the answer. It has a reasoning engine behind it, which It has a reasoning engine, behind it and, it can go into your fleet and the knowledge base and, and find the right answers for you. So it's meant really as, you know, a, something that could go along with the CLI.
So f- for those folks who already know how to use the command level interface and, or, or, or already have scripts and things like that, it's not replacing it, it's just yet another tool, but in this case with the, with natural language, so that maybe those folks who aren't familiar with the CLI interface or if they have to jump in in an emergency or something like that can still get things done, w- without having to know the ins and outs of,
of Purity and, and how to run it, right? Yeah. So, I also want to point out a, a distinction here, right? Our Copilot today is, to, to some extent read-only, and it can find the answers for you, it can do the reasoning for you. It will not execute the commands for you.
And, and we have done that for a reason. You know, we want to make sure that with all kind of security and hallucinations, so we want to be very clear where we draw the line. So it is, it is handy tool for customers to get to answers very quickly, and get, you know, answers like which CLI command should I run if you want to do
replication on, from one area to another. It doesn't actually do the replication. The- Way we are thinking about it is we have our sister product called Workflow Automation, where customers can define deterministic workflows. Think of this as if this, then do that.
And the copilot can execute the workflows that are created deterministically for certain use cases. For example, if you want to do a, a workflow to send an alert or workflow to, upgrade your array, et cetera, et cetera. So we wanna keep the separation between the reasoning and probabilistic nature of the
copilot and the deterministic workflows which can execute, commands for you. That's a great, that's a great point. And I, and you know, I think when we think of what, our AI tools in general that we're using in our day-to-day lives do, I, I, I think having guardrails like that and making two different systems, has a lot of different benefits, especially 'cause as you said, the
workflow automation is something that can be used in a lot of different ways, and getting to it from copilot is one way. But yeah, copilot is basically your easy button for finding on-- finding out what's going on with your system. That's great. Yeah.
And speaking of that, here are some examples of questions you can ask. The kind of sky's the limit in this case. The-- over the year, we, with partnership with, Amirul and Russell, the team has done a, a fantastic job of adding, as many items as possible. So, for example, when we launched, we couldn't ask questions about, you know, virtual
machines or alerts or anomalies. So one thing to remember is what was not possible six months ago, it probably is already there now. So, you know, just don't, Make sure that, you know, we are continuously evolving. This is not related to a specific purity release.
This is on the cloud, so, you know, we are releasing pretty much like twice a week. So there's a lot of innovation happening continuously, and we've expanded the breadth and the depth of the copilot, the last few months. My eyes went directly to the lower left there. Why is my array performance slow?
I mean, obviously here at Everpure, we're used to blazing fast performance, but I'm sure every, storage manager out there can identify with that's the number one complaint that they get no matter what system they're on. And getting any-- a quick and easy answer to that one is probably like seventy percent of their, of, of their headaches removed.
So that's great. That's great. Well, it's great that copilot can answer these kinda questions. What's the vision and where do you see it headed? You know, where, where is it going? What are the next, evolutions for AI copilot?
Yeah. So if you look at it, you know, twenty twenty-five last year, we started this journey. We started with knowledge search. We integrated various different knowledge sources, your technical user guides, the, the peak, knowledge systems and all the different, FA and FB, technical documentation.
And then we expanded quickly to the fleet insight. So all the different Pure1 APIs, for FA, FB, be it recommendations, alerts, anomalies, VMs, et cetera, et cetera. So you don't-- You, like, you, you can not only ask questions about a specific document, but you can start now asking questions about your fleet.
You know, what is the latency of this array, or how is the replication doing from one array to another, et cetera, et cetera, right? So that's how we started in twenty twenty-five. And this year we have expanded to, what we call as the case assist, the support center. You know, we have amazing support, but there are times when you want a quick answer or, you
know, there are times when the answer to your question or the solution to your problem could be answered by your copilot. So the copilot, tries to solve the problem, but you always have the option to talk to our, our technical support engineers. And that's a kind of a, a journey that, we have embarked on, this year, and it's a
multi-year thing. We'll continue to evolve and develop that. The other thing this year that we are, we have invested in is the troubleshooting assist, where you not only get an answer, but it helps you investigate problems. And this again, there's obviously, there's like various types of problems. We are starting with a subset of problems this year, and that is going to be a journey.
And in addition to that, we, we, we can allow you to do whatever copilot does on Pure1 within your systems, and Russell will talk more about that, how we can do that using MCP. And the future is where what we just talked about, right? How do we make copilot more actionable and much deeper integration with our workflows
engine, whether that is to, you know, create a, create a case, create a ticket, you know, upgrade my array, for example, or, you know, you know, or run commands on your array. How do you do that in a deterministic way? And that's kind of our future and leading to where the copilot can be more proactive. Right now you need to initiate the conversation with the copilot, with a question
or a, or a, or, or a, or a problem. And you can start seeing copilot to be a lot more proactive, where it recommends what you should be thinking about and what you should be, doing. And far future is the autonomous, and that's kind of a far, far future and but who knows with things kind of innovation happening so rapidly, far future could be three years from
now or two years from now. So we'll, we'll, we'll keep an eye on it. But, this is kind of our roadmap on how we think about the vision of the copilot. Think of this as your high-performing teammate. Where, you know, it's an extension of, of, of what you do and who you are. Love that. Love that.
And, yeah, I love how you, talked about the case assist and the fact that it's not really, you know, taking over for support, and, and, and you still have that, live agent interaction. That was something that Kaz touched on in our, anniversary Ask Us Everything last time around, that, you know, we, we, we want AI to help us and augment, but, you know, none of us feel
it's, it's there and ready yet. We, we, we'd rather trust more of the easy buttons that we put into the system in general to help you manage it rather than, take over completely from, from live support. So that's great. Well, hey, if y- anyone still needs to answer the poll, jump in and do it 'cause we're gonna close that in just a second, and then we can
bring up the, bring up the answers. So last call. Last call on your poll, and then we can shut it down. So, Ameeral, can you tell us a little bit more about the troubleshooting and triage assist capabilities?
Sure, Kevin. Thanks, Anand. As Anand was talking and just, saying that one of the biggest pain points that customers have reached out to us is help with troubleshooting and helping them solve, triage issues. So as of today, customers need to piece together information from documentation, CLI, and try to make, sense of it.
And triage assistant hits the nail on the head with the exact, issue that customers face a problem with. So all that they need to do is give it a symptom that, "Hey, why was my array slow?" Or, or, "I see that this array was slow last night. What happened?" And it works with you as a high-performing, teammate, to get to the issue at hand. So it'll start with the preliminary checks,
looking at was there any alerts that kicked off last night? Was there an upgrade going at that time? And if no, then it goes to the second step and looks at the baseline to see was, was there a spike in the latency at that time? And then map it to the volumes that got busy and then give you a conclusion, the evidence
behind it, and a recommended next step. So if we zoom out a little, the value here is time. The investigation usually takes one or two days. You create a case, you wait for the support agent to come back. But now all of this can be done in a matter of one to two minutes by just talking to AI
Copilot, which is your high-performance, companion. So if you look at the slide here, what it says is a single interface for all your debugging issues, which is was there a problem? What time was the problem? Why did it really happen?
It cuts across different dimensions, on, volume slowness or array slowness or capacity fullness, or was there an upgrade issue that slowed down, your array? And it's completely self-serve. So the customer asks the question and gets an answer in minutes. If, we just go to the next slide.
Right. So Russell, do you wanna talk about the MCP, or if you have any questions about triage, happy to take them. Yes. Hello. So, the Pure1 Manage MCP Server is a way to give your AI assistant access to the data that's in Pure1 Manage.
This sounds pretty dry, but the, the effect of this is actually very transformational. I know s- speaking for myself, my workflow has completely changed over the last nine months, and I use my AI assistant for almost everything. And the more stuff that I can plug into my assistant, I use Cloud Desktop, the more stuff that I can plug into it, the more powerful it is.
And so I use this to do And I think y- you might be going through the same thing of, like, I use it to troubleshoot problems, make recommendations, and generate reports. And so AI is very good at synthesizing data from multiple sources, and the more sources you give it, the more powerful it is. And this is just one element in your stack.
So you can add Pure1 Manage data, and then you can add other You know, for every other MCP server that you have, you can add, other parts of your stack like vSphere, your network, your observability, you know, things like that. And then it becomes a very powerful assistant for, doing the things I just described. Great. Thanks, Russell and Ameeral.
Two fantastic examples of how we can go beyond just the standard questions that we might be used to with, the AI Copilot into the area of really helping determine what's going on in our systems at a deeper level with, with triage, and then with MCP Server, then opening it up to other, AI systems to be able to help, you know, get, get in and, and, and understand more.
Again, it The whole idea of this is just to make your job as a, an IT professional, someone on the front lines, making sure your data is available to whoever needs it, you know, making that job a lot easier. We have a ton of questions. So let's go back to Anand and, cover off on some of the Copilot-specific
questions that we have. One of them is, a pretty simple one, from John, "Can we ask Copilot what our current load percent is," and I think this is the kicker, "on all of our arrays?" Yeah, I, I see why not. So the copilot is designed to answer questions about whatever you see on Pure1, unless there's a bug. But the, the expectation is that the copilot
should be able to answer questions, about latency, bandwidth, IOPS, and load. But if for some reason you don't find it, we'll we can definitely take a look into it, but it should, it should answer. Great. And, you touched on this a little bit, but I, I wanna make sure we're really clear.
Travis was asking, do any safeguards exist to ensure the AI copilot recommendations are accurate, explainable before providing them? Yeah. I'll, I'll give a high-level overview and then Russell can, talk more about it. So we do have extensive evaluation, platform to, to framework to make sure that the
copilot answers are accurate. And you have, the thumbs up and thumbs down button that you can give, feedback, and the team takes a look at it, very intensely, on a daily basis. But over to you, Russell, if you want to speak a little bit more about the evaluation framework. Awesome.
Sounds great. Okay. So- I think Russell, Russell may have a, a few more details about how we do the evaluation on the, for the- Yeah for the accuracy. No, I think you covered it. Okay. Perfect. Well, I have one
for Russell, from Paul. Are there questions only for the MCP for now, or will it be able to make changes in the future? We are working on being able to make changes, but that requires, like, an order of magnitude better safeguards. So, we are being very careful with rolling
that out. You can sort of mix and match things. So if your AI assistant has access to tools that can make changes, then, it can certainly use the data from the, the MCP server to guide its decisions. Got it. Okay. Travis also asked: How will this Pure1 AI copilot understand environments outside of
Pure Storage, VMware, Azure, other infrastructure platforms? I think, this is where the MCP can be super, super helpful. The Pure copilot, the scope, the eye that it has is for the VMware Pure appliances. Of course, we do have some insights into the, the VMs, if you do have VM collector. So anything that you see on Pure1, copilot will be able to have its eyes on.
But the, the, the more effective way is for you, if you have your own agentic systems and we You can connect our Pure copilot, for example, into Cloud and a VMware copa- VMware MCP server and, you know, your, your Azure MCP server all together, then you can start correlating one to the other. So that's, that's the main kind of the, the benefit of us exposing, the copilot as, as an
MCP server. Got it. Okay. And Alex has asked, do I use copilot CLI to use the MCP server? It might be for Russell. It's not necessary, but it is an option.
So I, I assume you're talking about, like, the Microsoft copilot CLI, and that is an MCP client, but you can use any MCP client to connect to our MCP server. Got it. All right. So let's see. Let's go ahead and bring up the, the poll results now and see exactly where everybody is.
So looks like we're kind of split down the middle there. About a little more than half have used it, AI copilot sparingly, and almost half of you have never heard of it, so this is a well-timed session for sure. And for those of you who are using it, looks like it's really across the board. Troubleshooting, number one, which we talked about, is, is probably the number one thing
that system admins are, are, are worrying about. I love to see, though, that there's fleet-wide questions in planning and forecasting from, you know, at least 20% of you. And then again, we've, we've still got quite a few folks who, need to still go in there and, and take a look at it. As far as the MCP servers, we're We got a lot
of people who are still thinking about it, some folks who are not even planning on it, but that's still a minority. So gentlemen, any, any thoughts on the poll and any recommendations for folks maybe how to get started? Let's start with, uh- Yes Anand on, on copilot. You know, for those who haven't tried it yet,
what's the what's, like, the easiest way to, and, and simplest way to just, you know, get, get started, maybe get a feel for it, see if it, it works for them? It's, it's pretty simple. So when you go to, Pure1 Manage, it's right on the top left corner, you will see the AI copilot as a tab.
You'll see a search box just like how you would see it on a, on a, on Google search, and you just start typing your question. And, if you don't know what questions to ask for, we do have a prompt library which, helps you with, specific questions that, that you can ask. And, and we're also, launching what we call is the autocomplete, where you if as you start
typing, if you type, "How do I..." then you will start seeing the possibility of what you can ask. You know, you can say, "How do I upgrade my array?" or, "How do I disable SafeMode or enable SafeMode?" Or, "How do I, you know, you know, find out a specific CLI command for to do something?" So it tells you the possibilities, but the It's pretty simple.
Just go to Pure1 Manage and start typing your question. Excellent. Excellent. So for those of you who haven't tried it yet, that's the easy button for jumping in, and then, you know, I, I, I think it will be very similar to, what we've seen, from a lot of the folks who have used it.
You try it a little bit. I know this is my experience with most AI tools. You know, for some things you wind up being underwhelmed, but for a lot of AI, it's, it, it becomes such a boon, it really does change the way at least I think about managing my work and the tasks that I need to perform versus those that I can offline, which, which
is always great. Another, question came in from Travis. He is, he is our star so far. So a lot of you other folks need to start dropping in more questions in the Q&A. Are there plans to have Copilot correlate storage issues with vCenter events, such as host changes, data store latency, you know, basically just getting it even more plugged in
with, VM-level stuff? Hemant, can you- I know this is something that Kaust touched on last week as well, and we kind of touched on it You, you touched on it earlier. It's like, well, you know, right now we want to make sure it manages the Everpure systems, but yeah, what are the plans for at least within those parts of the ecosystem that we
know most of our users have? I can take a dig at it. Yeah. So as of today, Copilot has some coverage as, to the vCenter events, but we'll have to build more on that. For the host changes, that is what the triage, looks at, if there are any issues with the host, host latency, volume latency.
And the next level of deeper analysis will be VM-level performance wherein we try to understand what kind of workloads are running on it, and we try to pin it specifically to which workload and which VM is causing the issue. Yeah. The main go- the main goal that we, we will try to get to from a triage perspective, Travis, is one, if you do have some
degradation, right now you can find, which workload, and we are expanding to which host may be connected to that workload which is causing the issue, then eventually expand to, VM-level insights into which VMs are connected to, the array which, where the issue may be or the volume where the issue may be. But yeah, so that is the plan.
But if y- beyond that, if you want a lot more correlation between your vCenter, the second thing that we, that we plan to do later this year, actually early next year, is if you have, the VM analytics, collector, then, which does have the entire topology map, we plan to integrate it that, with the Copilot as well. Great. Great.
And, there was another, follow-up from Paul on OpenStack. Will we be able to use, You know, w- a- a- any plans to support that for VM monitoring? I don't know. I, I haven't thought about that, but that's a good input. We- Okay will certainly consider that. Yeah. We, we will use that along with all the,
actions in the poll that we a- folks were definitely looking for a lot of, hey, if it could do stuff, that, that, that's a great roadmap plan for us. So, so is the taking a look at the OpenStack stuff as well. This one probably for, for Russell, but, just wanted to make, you know, let everyone know about the answers in the poll, around using, the AI systems with
the MCP Servers. We've got there again, you know, quite a few folks who haven't tried it yet. So do you have any, just like we did with Anand and Copilot, any advice as to how to, s- you know, get started, test the waters, take a look at it, figure out what might make sense?
Yes. I can actually present a little bit of a demo. Ooh, we love demos. Yes. So this is one I made earlier, but it, gets the idea across. So what is powerful about, these AI assistants, especially if you're using a coding, assistant,
and this is, ChatGPT, but it also includes Codex, which is, OpenAI's coding assistant. And so it can modify itself. And so what I'm showing here, I went in and I gave it access to the documentation, and I said, "Read the documentation and install the Pure1 Manage MCP Server." it actually kind of ran into a bump be- I said, "No, no, you need to use the client ID," and then it's installed
it for me. So once I installed it, I The, This one. So once I installed it, I said, "Test it out. What's my most overloaded appliance?" And then it gave me an It, told me, "I don't know which organization you're talking about." this is our multi-org support, for our larger customers.
Right. But once I specified the organization, then it was able to answer it. So, what I'm showing here is just that, you can tell your assistant to set it up for you. Or you could go t- That's awesome. Or you could go to the settings and, connector and just pretty much, you know, add the auth,
au- authentication, and it's pretty much like- Right it's not Either way, it should be pretty simple. Yes. Yeah. Yeah. Ta-ta-talk about single pane of glass now from the same tools that we're using for all the other things in our day. That's, that, that's fantastic, having that type of, you know, visibility right into our
systems, and again, through natural language. I, I, I think that, that could be a huge boon to folks. And let us know, you know, in the Q&A or the comments if that's, you know, scary or, "Oh, wow, I could actually use something like that." We'd love to hear that. Okay, so keeping on that same, uh- You know, vain, Rajendra was asking how to use
Claude/Cursor to talk to well, Copilot in this case, to talk to arrays, but it sounds like the MCP server might be the way to do that. Yes. Which question? so, Rajendra was asking how to use Claude Cursor- Okay, yeah to talk to Copilot, but then, you know, then to talk to arrays.
So what is, what is our answer in that? Yes. You configure a Cursor or Claude, to connect to a r- what's called a remote MCP server, and that's, what I was showing before that, like, if you have a coding assistant, then it can do it for you, or you can go through the, your settings.
But it works with any MCP client. Great. Okay. And then, Ameeral, I think this is for you. Tron had asked, "Does each on-prem array have their own Fusion MCP server?" So in, in his example, if I have 18 arrays, do I need 18 MCP servers?
No, and that's the beauty of Fusion. You can have 20 FAs and 20 Flash Blade, but it's just one fleet. So all you need is one Fusion MCP server, and it can talk to your whole fleet, and your whole fleet could include a bunch of FAs and bunch of FBs, and you only need to set it up once.
And as Russell showed us, it's as easy as pointing to the A- AI and saying, "Can you set it up for me?" And you're done. Excellent. Well, see, again, all roads lead back to Fusion at some point because it, it especially if you're talking a multi-array environment, it is an absolute boon, and it is the direction that we're taking all of our, our
management within, Everpure. So whether it's Pure1 or, you know, the Copilot, MCP server, et cetera, we always have an eye toward not just producing a feature or capability that works on one array and one array only, but how does it work across fleets? And then of course, you know, the reusability and other things, even if you have very small
fleets, within your organization, Fusion still has great, great efficacy. So I, I always love it when we can show how that's a boon to folks in their, in their day-to-day, just like using natural language. Okay, here's an interesting one. Why would an organization pick on-prem MCP versus PureCloud MCP?
And- I'll take at a high level crack at it, and then Russell and Ameeral, you're the experts here. So the, the PureCloud MCP is it gives the ease because you don't have to do any setup. It is already hosted on our side, and all that you have to do is go to your cloud and, and, and cursor to, to connect to it.
And you get a full view of your entire fleet, and you can ask pretty much any question that you would ask on Pure1 Manage, Copilot, you can, you can ask those questions. The on-prem MCP is for those customers, either you may not be on cloud, let's say you're a, you're a dark customer, or you have some security concerns where you want everything to be inside your LLM and you don't want to use our LLM, and you want to
have the full control, from a security perspective. Then you can install the Fusion MCP server on-prem, and, you get pretty much the similar benefits, from the questions perspective, but you, you get to have that completely under your air-gapped environment. But Ameeral and Russell can add more.
Yeah. I'd just also like to add that the Copilot MCP server, you know, it offers different tools, and it's, it You don't have to have Fusion running to use it. So it's like a different flavor of the same thing, and it's read-only for now. Yeah. And adding on to that, for the Fusion MCP,
it's, good for dark sites, as Anand said, and for users that want to do actionables and make changes to the code since it's open sourced. Let's say you have, VMware MCP or some other MCP and you want to integrate it. So, that's one way you can, use the open source code, make changes to it, and, do actionables through it.
Great. And Ameeral, let's stay with you because John had a really good, question. So, he's understanding that, Copilot can be used to analyze, you know, replication health, so it's, it's basically a replication failover type, scenario. RPO risk, disaster recovery readiness, you know, that it, it, it can show John what's going on
there, but can it also proactively identify issues that would then, possibly prevent the successful failover? So, as of today, Copilot can look performance issues and get to the root cause behind it and recommend the next steps. But going to specific replication analysis, that is one of the roadmap items this year,
and, that's very good feedback for us. Great. Excellent. See, again, that's another reason the These Ask Us Everythings have benefits for both of us. All of our, our, our, our technical-minded, you know, frontline, customers and users
who join this, also provide, not only get great info, but they provide us a lot of key info onto how you're actually using your Everpure systems and what's important to you, so we love that feedback. I know that, that's the only way I'm able to rope three really important tech guys in on a Friday morning for an hour, you know, is the, the The That's not their day job, let's put it
that way, you know. We're always looking at them going, "Where the heck is this feature?" here, here's another good one. Ian writes, "AI Copilot, does it have any limitations on ratcheted storage unit?" Okay, I am not familiar with that one, and I've been in storage for 20 years.
So in terms of finding info, info versus making, changes. So hopefully someone can decipher. I have never heard ratcheted storage. Yeah, I don't know too much details on that either. But in general, Copilot, the Pure1 Managed Copilot, stays away from making changes, but
it can find the information, but we can dig deep into the ratcheted storage unit and happy to get back on this. Okay. Yep. Excellent. And, Ian, if you've got more that you want to provide more color, to help us answer that better, that's fantastic.
Boy, our, our, our rock star Travis. So, "Everpure's emphasis on sustainability and environmental, responsibility as AI workloads continue to grow and consume resources, how are we ensuring that AI Copilot and related AI initiatives are being developed and operated in a way that aligns with those goals?" And I know that might be a little outside of the,
Bailey wiki U3, but I think it's, it, it's a concern that a lot of people have, for sure. I mean, AI feels like it's eating the world in a lot of ways, and not in a good way. So any, any, color behind the scenes that you all can provide? If not, I can jump in with some color too. Yeah, I can, I can tell what I know.
So we are definitely tracking the number of input and the output tokens used through the AI Copilot, which is bubbled, to the sustainability team. So everything that we do is definitely tracked both from obviously cost perspective, because we need to make sure that, you know, it is, sustainable from an ROI perspective. But at the same time, the, the, the LLM usage and the OpenAI usage, that we have is bubbled
to the, to the sustainability team, which goes into the broader sustainability report that, that we generate every year. But there could definitely- Right be more. This is what I know. Yeah. Yeah. And I, I was just gonna add, 'cause I, I know
that this question has come up sort of in general, if you will, you know, how do organizations like Everpure that are so focused on sustainability and have been from the beginning And, and what I keep hearing is, you know, again, obviously we don't have full control over the LLM stack and how they're using their own technology, but we know that from our own standpoint, we are incredibly efficient, I guess is the best way to,
to put it, and very mindful in how we design the systems both that our, our customers use and the ones that we use internally. You know, not just with a, a checkbox of, hey, we hope it's sustainable and, you know, then after the fact, crowing about how great we are. It, it, it's a design philosophy within all of our systems, and we'd be remiss if we weren't
taking a look at that. And I think, Anand, the example that you just gave about the, you know, the transparency and the accountability that we have on how many tokens are we using, you know, and then refining our processes so that, you know, and obviously there's a, there's an economic incentive, like you said, but, it, it, it has some real green output to it as well.
So that's a, that's a, that's a great topic, Travis. I think we could, That's actually a good, example for, a possible future Ask Us Everything, is just our, in general, you know, environmental, you know, what we design into the product and, you know, how we as an org do that. If y'all have, interest in that, please let us know in the comments or, or, or Q&A.
Let's see. Safe mode ratcheting is what Phil was, talking about- Yeah when it came to that. That the ratcheted server, so basically SafeMode. Yeah, it makes a bit more sense now. So, I can try to help answer that.
So- Yeah. Jump on in Copilot reads telemetry from Pure1, not the array's control plane, so the ratcheting should not restrict observability. So we would have information through observability, but, we won't do any actions through Pure1 Copilot. Excellent. And thank you, Phil, for
jumping in with a clarification. You did the, you did the stump the chump, and it worked. And I'm referring to myself, not my three esteemed guests. Let's see. Oh, here's a good one. Someone wishing to remain anonymous, "Can
Copilot be run on a live cluster to do log analysis?" I mean- So, so- Yeah today the, the, the input or the, the data sets that Copilot uses or the, the knowledge bases, right? So your, your peak knowledge base, the user guides, technical documentation, and all the, like Amirud said, all the phone homed metrics and, and the, the statistics that, that we
get from the array itself, right? We don't look into the logs mainly because it is, the logs can be really huge and it's, you know, again, going back to sustainability stuff, it can both take a long time to analyze the logs and also, take a lot of resources. What we are considering is if you have a specific time period or a specific log that
you want to analyze- we want to, either give you the option to upload that specific log file so that it can analyze, very contained version of the log file, but that's a future enhancement that we are considering. Excellent. Excellent. And I wanted to be, complete here.
I just noticed that Peter had asked in, in chat about using any of these AI tools, without an internet connection. And Anand, I think you touched on this when we were talking about having the, the air-gapped system, but, you know, just to make sure we cover that off, is that a truly air-gapped, I don't know, dark site type situation, or does this Do any of these, solutions,
will they always require some sort of internet connection? I think this is for the Fusion MCP server. Amul, can you take that one? Yes. I'm trying to find the question. Could, could you please repeat that? It's over in the, the actual chat as opposed
to the Q&A. Oh. Oh, I was looking at the Q&A. Peter was asking. Okay. Yeah. Could this AI ever be done on the appliance without an internet connection? So for the Fusion MCP, you, don't really need an internet connection. If, let's say you have an LLM of your own, a lot of orgs tend to build their own LLMs,
fine-tuned LLMs, so you can pass it on to that, so it never goes to the, internet. So it's completely in your control. If you want to use OpenAI tokens, you can pass it on. If you ar- want to use Claude tokens, you can pass it on. If you want to pass it on to your local LM, you can do that for sure.
Right. Okay. So so- sounds like it could work. Excellent. Yep. oh, Ian has a, has a good question. How can pure.ai Copilot help me with my upcoming Evergreen upgrade? Will it make a roadmap for my exec team?
Well, I've not tried that, but- I could try. I, I think it, it, it could work. So the reason I said that, it, it does have, the view into the, Evergreen subscriptions. For example, you can ask questions like, you know, "How much is my reserve commit?" "What
is my on-demand look like?" And, you know, what is "What will it look like in a few, few months from now?" So anything that you would see in Pure1 Around subscriptions, Copilot should answer. But the specific question, can I do roadmap, try it out, and I'm, I'm curious. I think it may give you an answer.
Yeah. Yeah. I would think if nothing else from the standpoint of, you know, w- what is the age on these systems and the, you know, what, what, what rev, if you will, when was their last because again, I mean, if we're, if we're By Evergreen subscription, in this case, the context I'm reading into it is, you know, a controller upgrade.
But obviously it could be any number of different things. I love, Anand, that you touched on Evergreen// One, because I think there would be real use to understanding, you know A- and we have the forecasting and planning tools already even, built into the system, that can be used without Copilot. But, you know, Copilot is a front end to that, I- Yeah I, I think really helps.
And also, like, if it is non-Evergreen//One, but kind of general upgrades- Yeah yes, it, it will tell you, you have 20 arrays, and these three arrays are the most important ones to upgrade because this may have a Purity version that has a CVE or it is really old or unsupported. So it does prioritize the list of arrays that needs immediate attention versus things that
could wait. Right. Also, I'd like to plug the MCP server here because AI assistants, one of their superpowers is being able to generate reports for you. So I think you might have really good luck if you, connect the MCP server and then ask your assistant to generate the report.
Yeah. Right. Exa- That, that, that's where my brain went to. I love the exec roadmap. It's like, half, half of our job sometimes is just explaining to bosses why we're doing what we're doing.
So I love that. Oh, this is a good one from Alex. What's the data retention policy, around AI Copilot? So most metrics, are about 60 to 90 days in the past, and it really depends on, on the, on the metric, but that's kind of a rough timeframe that you can keep in mind. And I, I think, Amul, you may have a little bit more details into which metrics, from a
retention per- perspective, but, you know, 60 to 90 days is a, is a good timeframe. Right. For m- most of the performance one, I think there's a 30 to 60-day gap with, three-minute, window. And for the other ones, it goes on for longer, for three to four months, and the window, for that would be, the freshness window for that would be around, 30 minutes or so, like the
granularity of how deep we can go into. Great. Excellent. All right. Well, we're not wrapping up completely here, folks, so don't tune out. We still have plenty of, of content here for you.
But we did want to do a final poll, so we're gonna go ahead and put that up, so that everyone has a chance to answer it before we do actually finish up, okay? So this is not your cue to go get your third cup of coffee or go to lunch or whatever. We're still here, but we wanted to make sure that you all had time, to answer this poll because these, these questions are really important.
Okay. Let's see. Someone asked what analytics are being done with the data from Copilot. That's a, that's a pretty open-ended broad one, but, um- I think we've given some good examples of what we can do with it, but anyone wants to jump in Yeah. So we, we mainly look at it from, two angles.
The, the main one is we look at it from an accuracy perspective. We talked about the evaluation framework, right? So we have a lot of analytics around, you know, where are the thumbs down happening and, and sampling the questions itself to understand, you know, where the Copilot may, may not be accurate so that we can go and address those problems.
That's kind of one angle. The other angle is to make our own product decisions and, and feature development because, like, there are a lot of questions you may be asking today to Copilot which we, we, we don't give an answer, where there's no coverage. So it helps us prioritize.
You know, there was a question about replication, for example. It helps us prioritize what's something new that we can build both into Copilot or as, as a new dashboard within Pure1 or, or any feature, for example, because if you're asking a question for which we don't have the answer, that is clue to us in terms of what is missing and where the gaps are, and that helps us, define our product roadmap.
Excellent. John asked, "What will storage admins of large international multi-data center environments find most useful for AI Copilot?" I can start with one use case, and then you can add. One of the things that we are adding as we speak, I think the next two to three weeks will be live, is, the multi-language support.
So, so far Copilot has been English only, and we are adding, you know, few more languages, and over time we want to add, like, 10 to 15 languages. So that is going to be a very huge benefit for international, you know, organizations where you may have, you know, subject matter experts who are in, in who speak different languages and get the same benefit from, from the Copilot.
Yeah. Over to you, Miro. There may be definitely others. Yep. Thanks, Anand. So one other useful, use case I see is fleet health report at that scale. When you're managing, like, 50, 60 arrays and, you don't want to log into each one and check,
how it's going, piece things together from different GUIs. So I think that was what Russell also showed as an example, through the Copilot MCP, which was, you can get an overview of your entire fleet as a fleet health report, and you can dig into it as to which ones need attention. So now it's a top to bottom r- rather than bottom to top approach.
Absolutely. Yeah. I also wanted to add one thing, which is that, it, speeds up onboarding because it has access to documentation. And so rather than you doing it yourself, you're going in and just asking the Copilot, like, "Well, how does this work? What does this error message mean?" and it has your Kontxtual information to go along as well.
And I'm just I'm gonna jump in again because of course that, that's exactly the use case where we designed Fusion for. And then again, all of this Copilot, the MCP server right on top of that, utilize it to go across literally your entire multi-data center, you know, data estate. And to your point, Russell, I love that you shouted out, you know, onboarding and things
like that. Between workflow automation in Pure1 and presets in Fusion, you can automate a lot of this stuff. And then that way when you're doing the onboarding through the knowledge transfer, 'cause again, you know, this is, this is read-only. It's, you know, it, it, it will at least be able to tell you which pieces of automation
can then be run or to know what's already being run in the background that's set to run so that you don't have to jump through, you know, 13 hoops this time. Maybe it's just one. You know, and most important, standardize that across this multi-data center environment. I mean, that's, that's the nirvana that we're shooting for.
And again, everything I just described makes sense even if you've got two arrays because we're always onboarding new folks or people change responsibilities. We're always forced to do more with less. You know, it's, it's never about we're trying to eliminate people. We're trying to make the people that are doing the job, their job easier because we know
you've got 700 things on your punch line, you know, your punch card for that day in your entire IT infrastructure. We're trying to make sure that at least your data and your storage is not the one that's, you know, taking up most of your time. Okay. That's I'm off my soapbox.
That's just That was the Doug. All right. Well, we're kind of wrapping up here. Yeah, sure, why not? Let's go back to Travis one last time. Do we have any statistics on AI Copilot being able to, you know, basically resolve issues, versus having to go to live agents?
And if not, I definitely wanna hunt that down because that, I think, just the same way that I know that, what is it? I believe currently 72%, it might be 74% of Everpure, tickets that are generated are generated by Purity right now and in all its different forms. But it would be nice to know the subset that
was done even before needing to, to have to deal with an agent, you know, through Copilot. Yeah. So, like you said, Kevin, right? So majority of the cases are created, are kind of proactively by We catch it before even you know it- Right and we try to fix it. Um- Fingerprinting and all that fun stuff.
Yeah right. So what we call as the proactive cases. Mm-hmm. And the reactive cases are where the customer comes and types, a problem, and then we try to understand what that is. And of the reactive cases, about roughly about 10 to 15% are the ones where the answer is somewhere either in the documentation Or, or it is, it's a straightforward answer from,
from the fleet. We can reason it out. And we are trying to expand that with the triage and the troubleshooting agent where you can, you know, sometimes, "Hey, this area was slow at, you know, 3:00 PM, why?" Right? I mean, like, those are the things that we want to be able to AI to go and reason. Either help or provide enough context to the support engineer so that they have all the
information and they can dig deeper before, they get the answer, and it, it reduces the time to resolution, right? Because they have more context from AI. And anybody else wanna jump in on that one? All right. Well, I wanna thank everyone, and, and first
just, you know, get any last thoughts that you all might have, words of advice, get excited for what's coming, whatever. The floor is yours. Russell, let's start with you. What are you most excited for? I love, like, you know, interacting with, like, where the rubber meets the road.
Like, talking to real people using this stuff and hearing their real questions is, It's like, yes, it's, very invigorating for me. Love it. And I'm excited okay, go ahead, Amul, because I'm, I'm super excited. Yeah, I am really excited about the vision as, Copilot evolves from observability to triage
to doing actionables, being an end-to-end companion. And one word of advice for our, guests here if, That would be just start. I know for some people it can be overwhelming, so if you feel like there's a lot, you can just ask your AI agent how to get started, but just get started. I, I, I love that one.
I just gotta jump in, 'cause, I mean, that's kinda how I was with AI in the beginning, too. It's like, "Oh, God, yet another thing I've gotta try out," whatever. And then you do, and then you're like, "Oh, my gosh, how did I exist before it?" now Mr. Excitement- Awesome back to you. So, so the main thing that I'm excited about is, I mean, I see that roughly half of you
have tried Copilot, and like Amul said, you know, I, I'm really looking forward to the, the rest of you trying it out, Copilot on Pure1. But what I'm even more excited about is I see there's a shift in how, in general in your organization uses AI, right? So, you know, you coming to Pure1 to use AI is one thing, but as you build your AI
infrastructure, I'm excited to see the, the MCP servers being a part of that holistic AI agent that you build, because every company is It's a matter of time. Every company's gonna build their own AI agent infrastructure and AI agent framework. I'm excited to see how you can correlate the Everpure signals with all the other signals so that you can get to the bottom of, the problems or information across multiple
systems, in your environment through these, MCP servers. Excellent. Perfect note to end it on. Thank you, gentlemen. This was I, I And thank you everyone in our audience. Fantastic questions.
You can tell a lot of people have been thinking a lot on this. Quite a few folks obviously have tried it out and were wondering, you know, where it's going next, can, can it help me with specific use cases. That's exactly why we do Ask Us Everything. Again, it's, it's, it, it, it's a chance to help everyone in, getting to their next step,
you know, basically. So if we can, bring up the, the wrap-up slides, walk through a little bit of housecleaning, 'cause there's a, a couple, a couple of fun things that we have that we can, let people know about. First of all, I wanna congratulate our raffle winners, Paul S., John B., and Alex T. Hardy congratulations, and you'll get, contacted about which prize you want to, want
to select. Also, yeah, again, if we can get those slides up, that would be great. We have the, QR code that you can scan, for our, We have a, a couple different things, actually. One of them is the Fusion Mastery Program, which I wanna make sure everyone knows about. We touched on Fusion quite a bit, and, what it is is basically a way of, getting credit if
you're already using Fusion for what you've already done. And for those of you who haven't tried Fusion yet, it's a great way to, you know, get a lot of great knowledge, try it out, and actually get a chance at some really cool prizes as well. So, you'll, you'll get information about that in your, your wrap-up kit.
And please, again, we host this and run it basically through the Everpure Community, which if you haven't tried it yet, is a fantastic resource. It's basically Ask Us Everything's, you know, in, in a more, you know, social media type, I guess, bulletin board type situation. Oh, my God, I just aged myself calling it a bulletin board.
You know, the idea It's the forum where, you can interact not just with Everpure, SMEs, but more importantly with your own peers and find out what's working, what's not. I mean, heck, people are f- sharing, workflow automation recipes and Fusion presets and all sorts of great stuff like that on it. So, definitely check that out as well.
I wanna thank, Anand, Russell, and Amul. You guys did a fantastic job. I really appreciate, all the, you know, great info that you passed on and, the energy that you brought. I wanna thank everyone in our audience, for being here.
We will have another, Ask Us Everything in just a few weeks, so please keep watching, for the next one. And until then, have a great day, have a great weekend, and we'll see y'all soon.
Anand Madhavan
Director of Product Management, Everpure
Ameerul Shah
Product Manager, Everpure
Russell Cagle
Member of Technical Staff, Everpure
Kevin Rickson
Product Marketing Director, Everpure
Got questions about AI-powered storage management? Get answers.
This month we’re diving into how AI and automation can help infrastructure teams simplify storage operations at scale. See how Pure1® AI Copilot, the Pure1 Manage MCP Server, and Everpure Fusion™ work together to help you understand what’s happening across your environment, connect AI tools to operational data, and simplify management across your fleet.
You’ll see how to manage your storage with even less effort, and:
- Get actionable answers in plain language with Pure1 AI Copilot
- Connect AI agents and tools to live Pure1 operational data with the Pure1 Manage MCP Server
- Simplify provisioning and day-to-day management across arrays with Everpure Fusion
- Move from insight to action faster with a more connected, intelligent approach to operations
We Also Recommend...
Personalize for Me
My Updates
No updates at this time.