Today on the Salesforce Admins Podcast, we talk to Avi Shah, Senior Director of Product Management for Salesforce AI. Join us as we chat about Agentforce Grid, a new way to coordinate data, automation, and AI agents.
You should subscribe for the full episode, but here are a few takeaways from our conversation with Avi Shah.
Salesforce Admins deal with data, automations, and AI every day. But how do you make everything work together in a way that makes sense? That’s why I sat down with Avi Shah to talk about Agentforce Grid, a spreadsheet-like tool for creating AI workflows.
“Agentforce Grid is, in our opinion, the fastest and easiest way to build AI workflows,” Avi says. “You have columns for your data and the actions you want to take with it.” Some columns are AI-based, enabling you to run prompts or agents you’ve built, and others are action-based, allowing you to update records or call an invokable action to send an email.
Put it all together, and you can build complex AI automations that transform your organization’s workflows.
As Avi explains, Agentforce Grid gives you a simple, spreadsheet UI to perform powerful transformations on your data. You can pull things from Data 360, uploads, or even the web into a data column.
Action columns give you a way to act. You can run prompt templates, agents you’ve already built, or inline prompts. Not everything needs to be an AI step, however, you can also perform more deterministic actions like formulas, updating records, or invoking flows.
All of this makes more sense when we talk about actual use cases. For example, you can use Agentforce Grid to assist with case categorization, working with a list of cases, a prompt column to analyze them, and another prompt column to look at those analyses and categorize them based on theme, priority, or issue.
Avi has also seen customers take advantage of Agentforce Grid for transcript and session analysis for customer-facing agents. You can use the prompt column to analyze, classify, and extract information from transcripts to make sure that everything is working the way you want it to work.
Be sure to listen to the full conversation for more from Avi on Agentforce Grid. And don’t forget to subscribe to the Salesforce Admins Podcast so you never miss an episode.
Mike:
Welcome to the Salesforce Admins Podcast. Today I’m joined by Avi Shah, and we’re digging into something that’s bigger than just a new interface. It’s a new way to design how work gets done. We’re talking about Agentforce Grid. Now, on the surface, it looks familiar, but under the hood, it’s where data, automation, and AI agents start working together in a single workspace. And that’s the shift. This isn’t about managing records faster, it’s about orchestrating outcomes.
Avi’s going to walk us through how Grid lets you move from reacting to data to actually activating it. From running prompts alongside your data to triggering actions in real time, to designing workflows where humans and AI play their role. We also get into some real use cases like case categorization and pipeline updates, but through a little bit of a different lens, not a “how I built this,” but how does this change the way my org operates?
So if you’ve been thinking about how AI fits into your day-to-day as an admin, this episode’s for you. Let’s get Avi on the podcast.
So Avi, welcome to the podcast.
Avi Shah:
Thanks, Mike. Happy to be here.
Mike:
Yeah. Well, let’s get started off for those that haven’t met you. Tell me a little bit about how you got started in the Salesforce ecosystem and what you do at Salesforce.
Avi Shah:
Yeah, I can’t believe it, but in February, I actually just passed my seventh year at Salesforce.
Mike:
Wow.
Avi Shah:
And I’ve had the opportunity to work in a bunch of different places inside of the company. So I came over and started on the essentials team, which way back in the day was the old name for our SMB/CRM, which is now known as Starter. Worked there for a little while. When we bought Slack, I moved over and started leading some of the Slack app integration, so was the product lead for Sales Cloud for Slack. And then spent a little bit of time on Sales Cloud, helping them kind of think through their Gen AI strategy and put together an initial vision around top of funnel prospecting. And then the Agentforce team spun up and had an opportunity to follow some of my Sales Cloud team over to Agentforce.
And that’s where I’ve been now for the past couple of years, and currently the proud product owner of a new product called Agentforce Grid.
Mike:
Yeah. Well, that is top of mind for us today because I’ve taken a look at it and I think it’s got a lot of really cool uses for admin. So tell me more about Agentforce Grid.
Avi Shah:
Yeah, absolutely. So Agentforce Grid is, in our opinion, the fastest and easiest way to build AI workflows. And so when you think about Grid, Grid looks and feels like a spreadsheet. We have kind of two types of columns in Grid. We have data columns and we have action columns. So data columns are pretty simple. They just let you bring in data, whether it’s uploaded from CSV, imported from Salesforce, so bringing in account or opportunity or case records, or even being able to bring in data from data cloud. And once you’ve got some data inside of the Grid, then the next thing is, well, you’re going to want to take action. You’re going to want to do something with it. And so our non-data columns, we have these action columns. Some of them are AI-based, so being able to run prompts and agents. Others are action-based. So being able to do things like update records in CRM or call an invocable action to send an email.
And so long story short, we’ve got all of these different column types and we’ve seen customers kind of stitch them together in really interesting ways to create kind of these AI-powered workflows.
Mike:
Yeah, no, that’s really cool. I mean, so from what I’ve seen, you can have data inside or outside of Salesforce. So it doesn’t all have to be in one place. When you’re building that Grid, if you have some data like account data in Salesforce, but also are bringing in a spreadsheet, you can kind of work with both of those data sources, right?
Avi Shah:
Yeah. Absolutely. We have support easily bringing over any data that’s native to Salesforce. Obviously, by supporting data in data cloud, we ultimately also support all of the data cloud connectors. So if you bring data into data cloud, you can then bring it into Grid. To your point, you can upload it from a spreadsheet. And we also have a new web search capability, which is really cool, which actually also allows you to bring in data from the web.
Mike:
Wow. Okay. Tell me more about the action columns, because I’m assuming these will also use some of the prompt templates that we’ve built for other Agentforce things, right?
Avi Shah:
Yeah, absolutely. So when you think about the action columns, we’ve got a bunch of different things that we support. So we’ve got AI-based action columns, and those are things like being able to run your existing prompt templates, run any of your existing Agentforce agents, as well as we have an AI column, which allows you to just define inline prompts. So if you don’t already have a prompt template that’s been built and published for a specific use case and you just want to use an LLM kind of in the flow of your work or in the flow of your Grid, you can do that really easily by just using the AI column to author an inline prompt. So those are kind of the AI columns that we have.
And then we also recognize that a lot of users, when they’re building out these workflows, not every step needs to be an AI step. So we’ve got a bunch of other action columns that they can use to help do kind of more deterministic things. So we’ve got a formula column, which just allows you to build formulas using Salesforce expression language. So formulas that output things like strings or bullions or numbers or dates, like all of that is supported in that column. We’ve got an update record column, which allows you to take any value from the Grid and just write it back to any of your records instead of Salesforce.
Mike:
Wow.
Avi Shah:
And then we also have an invocable action column, which just if you have standard or custom invocable actions, you can call those directly from Grid. So honestly, it’s really pretty vast in terms of what you can do.
Mike:
Yeah. I didn’t know about that last part. So the one thing that always comes to mind when you have anything en masse for a Salesforce admin is, what’s the security, what’s the guardrails around it? So I’m assuming that admins are going to be power users of this, developers, users within an org. You mentioned the ability to actually write to the record. That is totally profile and permission-based, right? So this isn’t going to give somebody more access than they should already have.
Avi Shah:
Of course. No, it’s a great call out. And yes, Grid runs entirely in the user context. So even when you’re trying to execute a prompt template or an agent, if you as a user don’t normally have access to that prompt template or agent, you can’t have access to it in Grid.
Similarly, if you’re trying to either pull or write to data inside of CRM or data cloud, it first does a permissions check on your user to make sure that you have access to that entity before you can actually interact with it.
Mike:
Yeah. And it’s funny, we just assume that that’s the case, but I always feel like unless I say it, it doesn’t become real. Unless I just remind myself, “Nope. They don’t have permission. They won’t have permission. We’re good.”
So I’ve thought of different use cases admins would have for this. What are some of the use cases that our customers are coming up with in terms of how they’re using Grid and kind of this AI action power?
Avi Shah:
Yeah. So honestly, the really cool thing is we’re still in the pretty early innings of this, and customers are coming to us with new use cases for different lines of business every day. And so some examples, right? We’ve seen customers be really interested in doing things like case categorization. So what that looks like is they’ll bring in a list of cases and then they’ll use a prompt column to basically analyze those cases. And then they might use another prompt column to look at the analysis and then categorize the cases based off of theme or priority or issue. And then they might actually either write those results back to the case records so that they can do really easy reporting on it, or they might just export it into a CSV so that they can analyze it in another system. So that’s a really cool one.
In a similar but different vein, we’ve seen things around transcript and session analysis. So as a lot of our customers get up and running with either Agentforce Service Agent or some of the other customer-facing agents that we have, something that’s really top of mind for them is being able to audit and kind of QA those agent conversations. And so we’ll see them bring in those actual session transcripts from Data Cloud and then again, use things like prompts on top of them to kind of analyze, classify, extract, like, competitor mentions and do a bunch of kind of like really cool ad hoc analysis to better understand what’s going on. Other use cases are things like meeting an account prep, pipeline and data quality audits, general like CRM hygiene and enrichment. It really kind of spans the gamut.
Mike:
Well, that was… I mean, the demo I saw, I’ve seen a couple. One was kind of like a follow-up to a conference when you get your leads and using that to kind of parse through before you really put them in the system.
I think the second one, the part that has me really jazzed about it is it’s a tool where I can, for lack of a better word, kind of have like a workbench that I can work with the data, I can query, I can make sure that the stuff that I’m about to put in is actually updating a record because the biggest problem I had a long, long time ago was you go to a trade show, half your customers would be leads, and then you put them back in the system as leads, but they’re already customers. And this to me feels like I could really get a bunch of data together and manipulate it and make my data better without it having to be on a system or outside of Salesforce.
Avi Shah:
Exactly. I mean, I think you nailed it, right? At the end of the day, what customers are really responding to here is, number one, being able to easily access their data. But then to your point, being able to use just LLMs in a really kind of easy, lightweight way inside of the safety of Salesforce. They don’t have to take their customer data and dump it into ChatGPT or into Claude to talk to it. They can do it right from Grid. So it’s really cool.
And that, I absolutely agree that admins and ops leaders are going to be kind of some of the power users here, but we’ve also started to see a little bit of traction from more of the business user who just wants an easier way to be able to use LLMs on their data.
So a really cool example is someone was responsible for generating status updates basically based off of like an entire pipeline view and they had to send those out every other week. And this was, there was 20 to 40 open opportunities inside of this pipeline view every time. And it would take them a couple of hours to actually go through and look at the latest activity, draft the status report and send it. And instead, what they did was they just built a Grid and they brought those opportunities in. They used a prompt to do the analysis. They defined that prompt once, ran it across all 40 in a single click, then used another prompt to, based off the analysis, draft a really quick status update, and then they just had it connected to the actual deal channels in Slack. And using an invocable action, they just sent out the status reports every other week. And then now all they have to do is they just go in and they refresh the Grid every other week and they’re done.
Mike:
Ah. So that was exactly where I was heading next, because the first part is always, “How am I going to use it?” The second part is, “What happens when I have to step away?”
So terminology-wise, let me get it straight, do we call them Grids that we create so you can create multiple Grids?
Avi Shah:
Yes. Exactly.
Mike:
Okay. And are Grids like list views, we can save them and they persist or no?
Avi Shah:
Yes, you can. So you can think of a Grid as like a file, like a doc. And then just like a spreadsheet you can have multiple tabs, you can have multiple tabs inside of a Grid as well. And so we call a Grid as kind of like the spreadsheet file. And then each tab we’re referring to as a worksheet.
Mike:
Perfect. Okay. Well, that’s nomenclature that a lot of people are used to.
Avi Shah:
Yeah. We wanted to make it feel familiar. That was the goal.
Mike:
Right. If I’m rolling this out as an admin, obviously I’m going to use this a lot, but I could turn to some power users. You mentioned sales. If I had salespeople that are really good at, say, creating list views or creating basic summary reports, creating Grids for them will feel the same, right?
Avi Shah:
Yeah. Absolutely. So we actually have a really cool feature inside a Grid that allows you … If you’ve already done the heavy lifting of creating a list view with certain filters, if you go into Grid, you can actually just bring in that list view directly. So you don’t have to go through and recreate the filters and do it. If you wanted to, the componentry’s the same. It’s clicks, it mirrors the same kind of patterns that the list view filters do, but rather than recreating it, you can now just select the list view and bring it in directly.
Mike:
Oh God, stop making things so easy.
Avi Shah:
That’s our goal.
Mike:
I was like, “Oh, well, we’ll just create the list view and then you got to recreate it.” Nope. Just bring it in, just make it super … I just like the idea that I could foresee a lot of … I think the use case where somebody’s classifying cases, I think that’s interesting, but I mean, salespeople always deal with a lot of different permutations, I will say, of requirements of, “I’d like a list or I’d like this to do this, this and this, and then one column to export out.” That to me feels like the biggest grab because I know as somebody who’s built sales list before, that X factor is the ability for Agentforce to add that prompt in. And I mean, you could create a list and do a call down sheet super fast because you don’t have to read through all of the activities. It could just give you, “Here’s how you need to greet this next customer based on the last interaction.”
Avi Shah:
Yeah, absolutely. And honestly, the web search capability strengthens this even more because LLMs, they are doing a much better job of having access to more recent data. Obviously, that used to be more of a problem. But in terms of doing basic research around a key account, you can now use more of the premium models in Grid to say, “Hey, go pull the company description, do something more sophisticated like pull the 10K.” And you don’t have to actually have a predefined agent built to do this. You can kind of do it a lot more ad hoc in Grid and obviously make sure that the results are grounded and cited, but we’ve seen some success with that already.
Mike:
Well, very cool. I know it’s just rolling out, but what are future plans for Grid if you can share anything?
Avi Shah:
Yeah. Absolutely. So we’ve got a big month coming up. So Grid is available today in live beta, but we are going GA on the week of April 6th. So that is coming up pretty quickly. We are really excited to kind of show this off at some of the upcoming forums that we have.
But in terms of the roadmap, I alluded to it a little bit before. So one of the things that a lot of our customers ask for is, “Hey, I’ve built a grid, it’s working great. How do I schedule this? How do I run it programmatically? How do I have this thing run every time a new lead comes in or every time a new case comes in?” And so those automations are something that is on the very top of our priority list. We will have a solve for that coming in the next couple of months, if not sooner. So automations is a big one.
Another thing is we’re really leaning into this idea of even though it’s easy and familiar to build grids today and it’s all clicks, not code, we want to make it even easier. So rather than forcing you to go in and configure a grid, we’re going to lean heavily into AI so that you can just describe what you want, whether that’s to a native grid agent or whether it’s to Claude. We’re building out a Claude code skill right now. And through the APIs, these AI assistants will actually be able to go in and build your grids for you. They’ll be able to edit them, they’ll be able to answer questions around them. And so really excited about kind of both the automations as well as the AI-first approach that we’re taking.
Mike:
Yeah. I mean, with Agentforce for setup, I mean, I feel like we’re closer and closer every day to just telling Salesforce what we need from it, and it either building it or returning a result back in terms of an application or data. So I would assume long, long down the road that Agentforce Grid will also be buildable by Agentforce Vibes, right? Because if you could vibe code something, a grid could be included in that application.
Avi Shah:
Yeah, absolutely. I mean, I think we’ll have a Grid MCP. And so we’ll have that exposed to all the other agents and tools inside of Salesforce so they will be callable and you’ll be able to interact with them however you want. But I think again, like the thing that we love about grids are: run them in the background, build them in the background, but if you ever need to click into a Grid to see what happened, you always can and you’ll get that very familiar spreadsheet-like view with basically a sell-by-sell rendering of what happened across your workflow. And so it’s just a really kind of safe way to build and it’s seen a lot of customers feel really confident really quickly doing it.
Mike:
Yeah. Well, I mean, I think with everything AI, the ability to audit it and the ability to do the one thing I didn’t like doing in math class in middle school, show my work.
I don’t want to have to show my work. It’s, I got the answer, you don’t need to know how I arrived there.
Avi Shah:
Right.
Mike:
But we want the opposite from our AI. It’s, I need to know where you arrived and I need to know how you got there. So that’s great.
Avi, I’d love to ask you, outside of building Grid stuff for Salesforce and some other really cool projects, I feel like you’ve got to put your hands on a lot of fun stuff at Salesforce. Is there any hobbies or any nonprofits that you like to do on the side when you’re not building Grids?
Avi Shah:
Yeah. Couple of great organizations that I’ll actually just shout out quickly. So StreetCode Academy is a really cool organization based down in the Peninsula that’s really focused on kind of bringing more of a STEM and technical classes and boot camps to kids from underprivileged backgrounds. And so making sure that they have access so that they’re ahead when they get to high school and college rather than behind. And so as we’re all kind of in the tech industry, that’s a really cool one.
And then another one is, on the side, I’m actually a big golfer. And so a couple of great organizations, Youth on Course and First Tee in the Bay Area are awesome vehicles to kind of grow the game. And it’s been fun watching and participating in some of those events as well.
Mike:
Yeah. So this sounds silly. Is there a lot of golfing in the Bay Area? I’ve never seen anybody get off a plane with clubs.
Avi Shah:
It is one of the golf meccas of the world.
Mike:
Really? Okay. I mean, I’ve flown to Phoenix before, not to golf. But when I was in the baggage area, every other bag was a bag of clubs because there’s a lot of stuff. I mean, it’s great weather out there in San Francisco, so I can’t imagine it wouldn’t be great for golfing. It’s just not where my mind goes.
Avi Shah:
It’s got an incredibly vibrant public golf scene. And then just a couple hours south, the Monterey Peninsula is considered one of the most highly regarded golf zip codes in the world.
Mike:
Oh. Yeah. Well, that I can see. That I can see for sure. Well, that’s very cool. Thanks for sharing that.
Avi, appreciate you coming on the podcast. And let’s see. I’m sure people are going to be able to run into you at some of the upcoming Salesforce events and see Agentforce Grid. It’ll give us something to do over the summer in between perfecting our golf shot, right?
Avi Shah:
Yep. We’ve got a lot of work to do both with grids and on the golf course, so that’s summer coming up.
Mike:
Awesome. Well, thanks so much for being on the podcast.
Avi Shah:
All right. Thanks, Mike. Appreciate it.
Mike:
Big thanks to Avi Shah for joining us and walking us through Agentforce Grid, and more importantly, what it represents, because this isn’t just a new way to work with data. It’s a shift in how we think about workflows entirely. From managing steps to orchestrating systems, from doing the work, to designing how the work gets done, that’s really where admins are heading. If you’re already thinking about how this fits into your org, you’re asking the right questions. Start small, pick a workflow, look for where AI can assist and where you stay in control. I hope this sparks some ideas and you’ll share it with your team or another Salesforce admin who’s exploring what AI can actually look like in practice. And of course, be sure to hit that subscribe button so you don’t miss what’s next. Until then, we’ll see you in the cloud.
Today on the Salesforce Admins Podcast, we talk to Joshua Birk, Senior Director of Admin Evangelism at Salesforce. Join us as we chat about how Agent Script helps admins build more predictable and reliable AI solutions. You should subscribe for the full episode, but here are a few takeaways from our conversation with Joshua Birk. […]
Today on the Salesforce Admins Podcast, we talk to Andrew Russo, Salesforce Architect at BACA Systems. Join us as we chat about Salesforce Foundations and why it’s a game changer for solo admins and small orgs. You should subscribe for the full episode, but here are a few takeaways from our conversation with Andrew Russo. […]
Today on the Salesforce Admins Podcast, we talk to Eddie Cliff, VP of Product Management at Salesforce. Join us as we chat about Salesforce Foundations and how it can give you access to even more capabilities within Sales, Service, and beyond. You should subscribe for the full episode, but here are a few takeaways from […]
