Video: Building Faster with AI? Don't Put Your Salesforce Data at Risk | Duration: 2696s | Summary: Building Faster with AI? Don't Put Your Salesforce Data at Risk | Chapters: Welcome & Introduction (4.56s), Salary Survey Introduction (53.68s), AI Data Risks (121.915s), Guardrails and Autonomy (330.365s), Six-Point Safety Rubric (592.365s), Preventing AI Slop (750.71s), Recovery System Architecture (814.905s), Recovery Scorecard Rubric (965.715s), Data Management Loop (1088.13s), Live Demo Setup (1323.925s), AI Agent Mistake (1398.78s), Selective Record Restoration (1655.49s), Data Recovery Framework (1870.155s), Try It Free (2197.88s), Q&A and Closing (2271.635s)
Transcript for "Building Faster with AI? Don't Put Your Salesforce Data at Risk":
Hello, everyone. Welcome or welcome to another Salesforce Ben webinar, this time in collaboration with Auto Rabbit, and thank you all for your patience whilst we, got started today. Just do let us know in the chat box where you are joining from. It's always lovely to see how global our audience is. And whilst you're there, do have a look at our q and a tab as well. We will have time for questions at the end of today's session. So if you do have any throughout, then, please do submit them in our q and a tab. This session is going to be recorded, and the recording is going to be shared on our follow-up email a couple of days, after the event. And you are in, listen only mode today, so you won't be able to come off of mute. Amazing. If we, move to the next slide, please. So just, very quickly from us, our Salesforce salary survey is now live. So do, scan that QR code on the screen to take part in our salary survey and join thousands of Salesforce professionals who are sharing their insights and being part of the story. It will be closing next Monday, but, the more insights that we have from, you, our community, then the more we can understand about the current state of global salaries, in demand skills, and also job satisfaction. And, also, as a nice thank you, all participants do have the opportunity to join our optional prize draw, and five winners will be picked at random to receive a $100 gift card. So, I'll also share the link, to the salary survey as well, so you can just look over that, as we get started with the, webinar. Amazing. If we, move to the next slide, please, I'd love to, invite our speakers for today's session to the stage. So we've got Prasanth and Naresh, the vice president of products and director of product management. So lovely to have you both here. I will hand it over to you to start the session. Fantastic. Awesome. Thank you very much, Salma. This is a very exciting topic today. Quite frankly, one of the most, you know, the... I'm... The hottest topic that you have, in the industry right now. You know, AI is smart at one side... On one side, but is AI dumb enough to make data mistakes? And if it is dumb enough, you know, what... How do you what do you do in order to sort of, get back, to your golden state? So that's what we will be talking about today. So my... I am, Prashant Samdwalla. I am the VP of products here at Otterrabbit with me. I have Naresh who is, who leads our, data strategy here at Otterrabbit. So before we go deeper into, into the session, let's set the agenda. So, we wanna talk about what the last few months has, taught us about. We wanna talk about how, you know, how important having a backup strategy is and what do you actually look for. And for you, the audience, I wanna give you something that you can take back to your organizations and also, you know, not only, like, protect your organization, but also, like, grow in your own careers using the tools that we will leave you with. Right? We will talk about, like, a scorecard system that you can employ at your organizations that can really help you, sort of get ahead of, you know, AI and Salesforce. Force. Now, you know, I'm pretty sure all... Like, just like us here at Otterrabbit, almost all companies are... Have a directive from the top to use AI tools to get more productivity, to do more, to have higher cognitive abilities than, than, you know, the current workforce has. So with that, you know, with that kind of pressure that is coming from the top, how do you prepare yourself? How do you make sure that, you know, you're not at blame when AI makes a a actual mistake? So that's what we're gonna talk about. So, Neeraj, can you give me the next slide, please? Thank you. So what has history taught us? A little bit. So over the last couple of months, we have seen multiple, events, incidents that where an AI agent, actually was having production control and there were recorded events of data loss. Right? For example, a cursor agent deleted a production volume for Saster, and they basically said a rollback was not possible because, you know, the scope was... The API scope given to the agent was too large. And, they initially said that rollback was not possible, but one crazy engineer found a roll... A backed up copy somewhere and they will be... They were able to restore it back. Same thing happened with, with Replit, where it deleted, you know, data that was not supposed to touch. Right? Like... So this problem is not just in Salesforce, but it is all across the globe. It's all across the different systems. The problem with Salesforce is Salesforce is an enterprise tool. So when an enterprise loses data because of an AI challenge, that is usually not published in the news. So I bet you these things have happened in the, in the Salesforce world, but we just don't hear about it as much as we hear about a cursor or a replant. So remember that number, nine seconds. Nine seconds is all it takes, for an AI agent to, cause, irrecoverable damage to your system. Next slide, please, Naresh. Alright. So with that said, you know, AI has already demonstrated that it can cause incredible amount of damage. Right? So what is the problem? So what... So the problem is not the model itself. Like, you know, and probably came out with, Claude five... Fable 5.2 yesterday, which is apparently one of the the best intelligence that is available in the market. So the model itself is not dumb, but the guardrails around it are actually not ready for the intelligence that the model has to offer. Right? So this could be that, you know, your MCP tools have a large variety of permissions. They have too much autonomy, and I'm, and I'm guessing... This this happened with me. I'm pretty sure this, happened with most of us on the call. When a AI agent, when you plot code or your codex is asking you permission, hey, do you allow this? Do you allow this? For the tenth time or the eleventh time, you don't even read what's on the, on the message. You just simply click, oh, yeah. I allow it. Right? So what happens is, because of that bias, there's too much autonomy that is given to an agent. So we have to have systems in which we have, you know, we are able to recover to a previous state, when we give too much autonomy, to an agent. Right? And there's no undo button for agents. Like, one of the examples that I've seen demonstrated in the recent past is an AI agent or a swarm of AI agents who are accessing multiple links on the on the Internet and one link suddenly downloaded a malware. Right? And, you know, the AI agent caught it in the right, in the right moment, but there was no sort of undo URL. Like, how did you come to that URL? Can you undo that? That path is not available. So if that path is not available in your harnesses or in your guardrails, what do you do? Right? So this is what we're gonna be talking about. So the premise is that AI is very, very smart, but the guardrails are dumb. So how do you make sure that you have the, you have the right guardrails where you secure the human and you secure the system? So that's what we'll talk about more. Neerish, next slide, please. Alright. So, like I said, Salesforce is not exempt from this. So right now, you're able to use AI agents. It could be your agent force. It could be your Cloudforce. It could be your codex, Cloudcode. Whatever you're using in whatever fashion, you are able to connect, a Salesforce instance to your AI systems, to your AI harness, and they are able to run inserts, updates, deletes, upsurge on your org, on your metadata. And there is a sudden ripple effect that is felt across, you know, your entire organization. Some new intern, some new admin decides that, hey, I'm gonna I'm gonna complete all my Salesforce backlog today and starts wipe coding. And let's say that they're not using the right, the right guardrails. You can end up, with a very dysfunctional Salesforce arc as, you know, like, in a in a short duration of nine seconds. Right? So how do you make sure that you sort of control that? Right? There's another problem with AI, which is, the multi step agent chain can leave steps one to n minus one. But when... Let's say that if n fails, there is no native rollback for the o... The... For the n minus one steps. So let's say that you're building, a functionality, and that takes about 10 commits. Right? And that 10 commits is across, you know, data and metadata. So let's say that the ninth commit fails. Right? Today, the harnesses, the systems that we have do not have the way that they can roll back the the earlier nine commits. So what do you do in that kind of situation? Right? So this is what we will be talking about more, but I want to set the premise that Salesforce is... By opening up Salesforce to your AI agent, you are, you are unlocking the ability to make changes at the speed of thought. And what are the guardrails that we need to have? Narish, next slide, please. Alright. So here's the takeaway. AI made Salesforce rights cheap and fast. Undo must become equally cheap and fast, or agents will outrun your battles. Right? So remember this. And now let's talk about the rubric that I promised you, the tools that I promised you that you will be able to take away from this webinar. So next slide, please, Naresh. Sure. Alright. So I would say, as a Salesforce admin or a Salesforce developer, if you're managing a production org, these would be sort of your, six, let's say, six, tools that, or six places or six systems that you have to keep in mind to make sure that you have a very reliable data risk averse system. Right? So the first and foremost rubric, point is the speed gap. Are you able to recover as fast as an AI agent can write? Right? The second one is the blast radius. What can the AI, permissions actually touch? Right? The third is reversibility. Can you undo selected records without overriding good work that has already been done? Provens, which can you see what has changed when and by which user and by which, agent. Hygiene. Now this is a big, big problem in Salesforce. And if you're in Salesforce as long as me, you would know that Salesforce systems can soon, like Salesforce Salesforce production org can go out of, let's say, data hygiene sync, extremely fast. And there are often cycles that Salesforce admins and Salesforce developers take in order to just clean up data. So with AI, when you're able to make, changes at the speed of thought, you're also at a risk of adding junk data and junk metadata faster than you're able to clean it. And the last thing is trust. Can your green... Can your leadership green light agent right back? Meaning, the recovery that also happens, through AI. So these are the tools that I wanna, that I wanna give you guys today. Like, as a Salesforce admin, these are the things that you should be thinking about before allowing your organization, to access Salesforce via cloud or codex or any different AI system. These are the things that you need to have in place before, the first right happens on your Salesforce or using an AI agent. Right? It could be as simple as a a e, taking a picture of an org chart and saying that, hey, Claude, you know, go ahead. Read this picture and update the, contacts in my sales. Before that happens, you have to make sure that these six things, this checklist of six things are checked off. Raj, next slide, please. Alright. So we often say this in Salesforce, junk in is junk out. So you have to make sure that your AI is not creating junk data, meaning duplicates or stale records, or leftover, pileups. Your AI is not creating junk flows or junk, objects, fields. And even, you know, before the AI era, we had this problem with humans. Humans used to create, a lot of junk fields on important objects like accounts or contacts or opportunities for some trial and error, function, but they often forget to to remove that. The same thing will happen with AI if you are not careful. So we wanna make sure that, you know, you you are having the right controls to make sure that the debt does not grow, agents are not using junk data, and, you're not getting more and more AI slop. Right? So to prevent AI slop, you have to make sure that there's no junk data. So that's sort of the formula. Alright? So let's talk more about the recovery part. Nourish, next slide, please. So let's talk about... If you have to build an architecture, and we did, where you have to, create a reliable... Let's call it a time machine so that you can go and do write backs to your Salesforce org after an AI agent has done some damage. So let's look at how an architecture would look like. Narash, next slide please. So this is a very good representation of how, a good recovery system should work. So when your agent force or Cloudforce or MCP agent is making those DML calls to your Salesforce arc, there has to be a system that is taking time snapshots. So the t zero, t one, t two, t three, and t n that is able to take backup snapshots at regular intervals or even, you know, with a continuous stream of data, and able to reliably store that. Now what happens with... Because of that is, let's say that you realize after five days that, you know, there is some junk data or bad data, in your Salesforce system. So you you should be able to sort of reach in to that particular times, time snapshot and recover the data that you need. You'd also need a reliable system that takes that delta, that small piece of data and is able to stitch back. What object does it go into? What is its parent? What are the childs? From the highest object to the lowest child, how do you create that, you know, the the the the schema that is required for Salesforce to accept it? So those who have worked with data loaders like me have have this experience that you have to work with multiple CSV files, multiple mappings, and, you know, multiple objects and figure out what, you know, the relookups of multiple parent and child... Between multiple sheets of parent and child object records. That is the world that we have lived in. But we cannot live with manual data loads anymore. We need some automation where the system is automatically calculating, okay, what is the data that needs to be put back into Salesforce or put into Salesforce. So it's a very, very important that you have not only the time based snapshots, but also a system that deeply understands Salesforce schema and is able to put together their files in the format required for Salesforce to understand. And this is the system that we built. We would love to show you a demo in a little while, so, stay on for that. Anurag, next slide, please. Alright. So in terms of, what we built in terms of a recovery, like I said, the... You will need a rubric to understand how good your backup and recovery system is. Again, this is just for, you know, heuristics. We're not gonna be, these are not, computed numbers. So let's say that, if you have, a reliable backup, right, that is able to take backup at a regular interval or consume a stream of data, as it changes, then you know that you're doing a five on speed. Right? If you're able to understand, like, for example, the opportunity record has changed, so does that mean the, account or contact record has changed? Right? So if you're able to understand the blast radius, then you also get a five. And can you recover from the snapshot, like, for example, if you... If there's an opportunity record that changed, are you able to reliably put that opportunity record back if you found out that that was bad data or malicious data or just junk data? How fast are you able to find out if, your system is, creating junk or not? So you need a reliable compare operation to see if, certain before and after state of data is, qualifies our junk or not. Right? So I would say that, you know, for all of you on this webinar, use this... Screenshot this. Use this scorecard to go into your organizations and figure out if we have the following, you know, if if you hit five on the... On your... On this rubric. Right? On speed and backup, on blast radius and compare, on reversibility and restore, and on and on and on. So this is a very valuable tool that we can offer you today that you can take in and just go and compare if you have the following in your Salesforce org. And if you don't, then, you know, we will leave you, by the end of this webinar, we'll leave you with something where where you can actually go and do a free trial, of our product. So, Maresh, next slide, please. Alright. So, to think in loops. Right? Because the... This, AI agents... In the era of AI agent, everything is sort of a loop. So, if we have to put the features that you need to satisfy the rubric that I just showed you. Right? So there needs to be a continuous schedule backup that takes the backup that is able to do a search and compare, that is able to prove restore. And what is proving restore? And, you know, for those of you who are on this call, you know that in in the world of Salesforce, putting data into Salesforce in the format that it has is actually a very, very hard job. Right? So how do you do that? And the... Probably, we also have, other things that, you know, AI systems need. Like, for example, seeding. When you have new features and functionality that are being delivered on a, you know, at a extremely high velocity, how do you test for that? Now you cannot use your Salesforce production data in order to test, for those functionality in sandboxes. So you need a reliable seeding mechanism that creates static data, that creates that mass production data in order for you to prepare sandboxes, so that they are ready for, you know, high velocity developer testing. Finally, you also need a reliable archive system that is able to give you a view back into your Salesforce. The reason is, as AI continues to evolve, one of the things that will also change is the amount of the volume of data in your Salesforce org. So your storages are gonna be... Start getting, fuller and fuller. Earlier, the problem is the best AEs and the best, Salesforce operators, Well, let's just say, you know, they were not very timely in updating Salesforce, and they probably would... Are not putting all the data that is required for them to put in. But because of AI, they could simply take a screenshot, upload a meeting recording, and that uploads your Salesforce organization. What that loop basically means is the volume of data in your Salesforce would drastically increase. When your volume increases, your storage is full, your Salesforce are slow. You need to offload storage from your Salesforce are to an alternate storage, say an S3. Again, have a view back system so that you can view those S3 records in your Salesforce as part of the object as, you know, and leave your users with an interface where they don't feel that this data is coming from outside of Salesforce. Right? So you need a reliable archive system. So the rubric shows you the rubric that will prepare you for AI will show you that these are the features and functionality that you desire, before your Salesforce org is ready to be unlocked, for AI. Right? So, Naresh, next slide, please. Alright. So this is what we've been talking about. You have to have a reliable system where you're able to practice without, you know, destroying your Salesforce org. So before you're moving data from production to your sandbox, you have to have a reliable seeding mechanism. This is something that we cannot stress enough. Please do not use production data in your sandboxes. And if you are in a regulated industry like health care, insurance, banking, then by bringing production data into your sandboxes, you are breaking, compliance. So please make sure that you have the right systems, the right tools. If not, AutoRabbit can help you solve those problems. But if you're not using AutoRabbit, please have those systems in place. Please use the rubric that we have showed you a little while ago in order to have a comprehensive data strategy around your Salesforce. So I think it's time for a demo. Naresh, can you give me the next slide, please? Alright. So I'm gonna hand off to Naresh. Naresh comes with, incredible, decade or more experience around Salesforce and Salesforce data, strategies. There is no better person that I know of who can talk about Salesforce data strategy. Naresh, over to you. Thanks, Prashant. I mean, so far we have gone through the theory of why we need, how AI agents can make bigger changes in your Salesforce logs and how it can impact your production data and all of that. Now let's see that in action. We don't want too much of theory here. Like, we'll be, talking about the scenario where we'll take a baseline of the data that is already there in your, Salesforce org, and then, we'll use Cloudforce to create some... To make some changes to your Salesforce data. And after that, we'll do a compare operation to exactly understand what are all the fields that got changed and, what's the impact that change has created. And then we'll restore it back to Salesforce. And finally, we'll verify whether, the intended, change is restored back as we, as we wanted. Let me just Ashans, open with that number. Nine seconds. That's how long it took an AI agent to wipe out the company's prediction data. Nine seconds. That is less than it takes to find your phone when it rings. So I'm I'm I'm going to do something a little uncomfortable today. I'm going to make a mistake, not a fake one, a real one, the kind that happens on an ordinary afternoon to a smart person using a smart tool like Clark. And then we are going to find out what it takes to undo it. Meet my sales talk. Real deals, real deals, and let's look at the opportunities. This one, Cobalt here, Cobalt Analytics, is one of my biggest customers, and their annual is due in eight days. Today, I'm a sales manager, and it's the end of the quarter. My pipeline is messy, and I have a new teammate who never sleeps. It's Clark. Connected, straight into my Salesforce m... Through an MCP. Let me show you how it's configured. You can see that there's a Cloudforce, MCP that is configured here through which I can do the read operations on the Salesforce data, write, delete, and update operations as well. For the read, there's no any approval gate defined here. It's always in yellow. And for write, delete, or update operations, it requires an approval. And let me also show you something that's there in Hotrabort Vault. You can see a baseline demo that I have taken, before doing any changes using cloud. Now let me go back to cloud and run this simple prompt. What it will do is it will just clean up my pipeline. Anything that is slipped past this close date is dead, and I'm just asking cloud to mark those opportunities as close lost. Let's see it in action. Cloud is identifying the tools to perform this particular operation, and it's using the Salesforce MCP internally to query the records that are matching the criteria that I have specified. And once that that is identified, it's just asking me for the approval to, query the records and to update them. Let's see what it identifies. You can see that it identified a few records, which requires an update. I'm approving it again, and I'm not taking too much of time. I'm just taking couple of seconds to approve that. That's how it happens in the real time as well. So what we're seeing here is it has updated a couple of deals here, Cobalt and Iron Leaf. I know that Iron Leaf is going to go out of, going to go out because they are building something in house is what we heard from them, but Cobalt is a real deal. They are going to, be in a renewal, in the next couple of weeks, and it's incorrectly updated now. Now let's go back to Salesforce, and let's assume that there is someone someone who have, who who are going to do certain changes to one of the opportunities here. So what they're doing here is they heard a good news from Harborview, and they're just marking it as closed one. Now what we're seeing here is there are a couple of changes, two different opportunities that were updated by the cloud once the prompt is received by it. And there is one particular opportunity that was actually updated by a real human inside the Salesforce org. Now let's go back to HotRabbit, and let me just run the confar. Now I'm running the COMFAR job to understand what exactly are the records that got changed either by cloud or by human after I have done, the changes, using cloud. I see that there are four records that got modified from the time I have taken this backup. And I can see that there is... It's not just the, scales that got changed. There are other fields also that got impacted by the changes that we have done using clock. You can see that the forecast category is impacted. You can see that, there are other, fields also that got impacted by the changes that we have done using CLOT. And there's also a chance whenever we make the changes using AI. It's not just the specific field that we want to change. It's always, additional fields that are going to be impacted and also the flows associated with it, the triggers, and everything that are associated with it, can also get impacted. Now what I'm going to do is I'll just... I know that Cobalt Analytics is not supposed to be restored or changed to closed lost because that's an actual deal that is still in progress. We have shared a PO with them, and direct to approve that. So what I'm seeing here is Cobalt is marked as closed lost, which is not supposed to be the stays of Cobalt. And let me just select this and restore it. This will ensure that none of the other three changes that we have done, either by some human that have done these changes or by someone using cloud are not going to be impacted by this, by this particular restore operation that we're about to do. I'm just selecting that one particular record, which is not supposed to be marked as closed lost. And doing some configuration here. I'm just clicking on restore now. And within a few seconds, I should see that the opportunity is back in Salesforce in its original stage. What it exactly does is it will just pick that particular opportunity record, identify the fields that got modified, and just revert them back to its previous days by taking the, previous snapshot from the backup that we have taken before making any changes using clock. Let's see if it is done. Yeah. It's done now, and I should see the cobalt. Yes. It's back to its previous days and from closed lost. And, all the changes that were done by AI, all the remaining changes are intact. Like, if you can see here, Iron Leaf was supposed to be closed lost because we know for sure that it is in closed lost days, and it's marked as closed lost. And Cobalt is back to its previous days. Alright. Everything you just watched fits into this one slide. There are five words on this, prepare, protect, capture, compare, and restore. Like, what I mean by prepare is before an agent ever touches production scope, what exactly is allowed to change. You need to basically let it practice on the seeded data, which is again a must. Even if you are putting it from your production, if you are seeding the data from your production, you need to ensure that the data is in a mask state to ensure that you are compliant all the time. My mistake today that happened is just an item walk. It should not be the first time an agent sees your real customer's data. That shouldn't happen in the real world. And the next thing is to protect, confirm that you have a recovery point always in place before the agent does any rights on your data. And remember, the backup that I showed you, that one step is the difference between a scary story and an ornament. If I don't have the backup, I could have been a situation where, I couldn't, restore all the impacted changes. It is just a... Just one opportunity, but in the real time, there's a good chance that it's thousands of opportunities and they all associated, leads and, and the notes associated with them and the other relational data. All of that could get impacted and we should be in in a state where we are prepared to protect our data from any unintended change from AI. And the next one is to protect. Confirm that, I mean, like after protect, it's the ability of the tool that you're using to capture every change that is happening in your environment, ensure that you have continuous backups in place, which will help you record back in time with... Whenever, something bad happens, you can always have a recording point that is just taken, a few minutes back or a few hours back just to ensure that you're going back in time. And, the next thing is the compact. Like, I was showing the, the compact gives you the understanding of what exactly changed in your org. It's not just even though, the, prompt that you have given is supposed to change the stays of that opportunity. We have seen that it changed the forecast and other, other, other fields on the opportunity, which we could understand from the comparison that we have done. And finally, the restore undo... I mean, the restore should not be all. Like, when we have a snapshot, it shouldn't restore everything back in time because, there's always a chance there's something intentional happened inside your Salesforce org. Someone... If you may have done that or even, AI might have done thousand different changes and all of them, let's say 900 are supposed to be done and only 100 are supposed to be restored. That's where the select to restore comes into picture, and the ability to restore only, only the change that is not intended to be there in your Salesforce log is very important to be recovered. And, if you remember, one line from this slide, let's make it the one at the bottom, writes admission speed, need undo admission speed. That's what you can do with our Rabbit Vault. And, now here is a question worth sitting with. Why did he isn't close to two deals? Because they were in the front of it, and and he isn't taxed on what it can see. So think about what your agent can see today. Ears of closed deals, leads probably nobody has touched since long time, And, the test record someone might have forgotten to delete. And everything... Every one of those is something an agent can read, listen about, and act on. That's why, we call archiving a high zen engine for AI. So with archive, Vault Archive, historical records more of live Salesforce stories. Your agents have less scale data to react on and your storage will comes down to. If you ever need that, data back, you can restore it from any past date. Like, you can have multiple snapshots of your backup, and you can always restore back, from them. Or you can also archive some data out of Salesforce, and you can restore it back whenever you want, either selectively for a specific set of records along with its relationships, or, you can do it for the entire backup as well. And, but there is always this part people usually push back on. If I archive it, can my team still see it? Yes. That's something that is possible with VaultConnect. Our care data shows up right inside Salesforce, and there is nothing to install. No custom objects, no custom scripts, nothing, is required to be there inside your Salesforce org. All you have to do is, to just install a managed package and, your auditors will get their history and your users, will get their required context. And this... All of this happens without restoring anything into your actual stories. And, also, what this means is less clutter to your gaze is AI agents, which ensures that, there is full history to your people while giving only the required context to your AI engines, which will not act on the steel data, but once you archive it out of Salesforce. So here is my challenge to you. Before, your team's next AI driven change, whether that's a new ASINforce action, a cloud connector, or a flow someone generated in five minutes, you just put a recovery point underneath it. You can start tonight, and that is going to be for free. You can just scan this QR code, and you'll get Ultra Rabbit Vault freemium. It takes three steps. Sign up, connect to salesforceorg, and run your first backup. The third step is the one that matters. It's the same backup I showed you at the start of the demo, the one you forgot until it saved my pipeline. A quick note. So, there's not going to be any surprises. The free tier doesn't include everything you saw today, but it gets you that first recovery point, and that's where every story like mine starts. And you can also recover, using the, using the freemium capital, but there are certain capitalities which you can only, you you can get only by upgrading from freemium. Just take a moment to scan it now, and we'll leave it up, while we switch to questions. Amazing. Thank you both, so much. As Naresh mentioned, we're gonna move into q and a now. So if you do have any questions, do feel free to, pop them through in that q and a tab, and we'll cover as many as we can. We did have a couple come through. So the first one is how should our data strategy change when we start using Cloudforce to update Salesforce records? Let me take that. This is what I've been talking about, in the beginning part of the webinar. You need to have a strategy. Right? So the tools that we will leave you with is a rubric, the... Which had... Would have the numbers. You know, I hope people have screenshot of that, during the conversation, but that is the strategy that you need. You need you need to understand, you know, the speed gap, the hygiene, the provenance, the reliability, and on and on. Right? Like, in across multiples... Multiple functions like backup, compare, restore, archive, whatever. Right? So that is what Salesforce admins and developers desperately need to be thinking of. Like, that is the strategy. And we have come up with that strategy after a long series of going back and forth, with multiple customers. So it works. Amazing. Thank you. Another one here. What should we back up before an AI driven change, and how do we confirm we have a usable recovery point? I'll take that. The, it... It's always good to take the entire data backup because we don't know what exactly is going to be changed by AI. Like, we were talking about, the AI changes that we have seen in the demo. We have just asked, Clark to make changes to the stays of the opportunity, but it internally changed the forecast and other, other fields on the opportunity. And, also, there's a good chance that it will, impact the other related, child and parent records as well. So it's always good to have a full snapshot available in place and especially for your business critical objects. All this ensure that there's a separate backup, which will give you, enough, like, space to recover from any unwanted situation. Amazing. Thank you. If an update affects related records across multiple objects, what should we consider before restoring them? Oh, Nice. That's for you. okay. I didn't get that. Can you please repeat that? If an update affects related records across multiple objects, what should we consider before restoring them? So it's like, ideally, the, stool should be... Tool or any script that you're writing should be intelligent enough to figure out the relationships, related records, and everything, while also ensuring that it gives you the granular level of access to select a specific child record, and it it should automatically identify which parent, changing which parent impacts which child record and things like that. All of those relationships should be clearly understood by the tool that something, only if the tool is intelligent enough to recover from a situation like that, it's always a easy, thing for you to recover from a situation where your related records are also impacted. Great. Thank you. Another one here. How can we create realistic test data without exposing sensitive You can either, customer information? use, tools which will generate synthetic data for you, but, it's always... It it needs to give you the... Give you enough context on the, of, it it needs to, very close to what you have inside your production with respect to its characteristics and things like that. And, also, there is one other option where you can just take off your... The Salesforce production data, but there is this cost associated with that. Always ensure that you're masking the confidential fields and the sensitive information and the CDPR data and everything before making, before using that in your lower enrollments. Lovely. Thank you. And then I think this is the last one that we have, for today. What should we verify after a restore before users resume their normal work? You won't take that question? Yeah. I think, before they, I think the time between, the restoration activity and the user, you know, logging back in, that time is actually very, very small. So it needs to, Sample checks that you can do is go into the audit trail of the log and see that they record. You have to at least have a subset of data that you think have been impacted that led to the rest of our activities. So a good check would be to go back and see if the data resembles its previous snapshots. And there are there are places where it will not. Like, for example, the system timestamps will not change. The system time stamps will, what I meant was it will change. Like, the last modified and the created date will not change. So so stuff like that have to be, considered, and this is more on a case by case basis. What I believe you with is, have restore drills. Get a reliable restore drill going in your organization every once in a quarter where you simulate a data loss event, what we call in the industry as DLEs. Simulate a DLE and then try to recover back and then have, you know, have greater so that you can check what percentage of data was reliably recovered. Right? And you need a good partner and a good, you know, a good, genius who understands your org, your schema, and your data. So that's what we'll leave you with. So, this is this is this is, something... This is a skill that you will evolve by practice, and, there's no magic green button. Amazing. Thank you both. So I think it's, time to wrap up the session now. So, thank you both again for your time today and for explaining, everything to us, and thank you everyone in the audience as well for, spending the past forty five minutes with us as well. As I mentioned in my introduction, a recording will be shared in our follow-up email if you would like to revisit any of the key points along with a link to, sign up to Autorabbits Vault freemium as well. Thank you everyone again for, joining us today, and hopefully see you at our next Salesforce Ben webinar too. Thanks, everyone. Bye bye. you. Bye. Thanks, everyone.