Speed and Scale at Reduced Cost | Agentforce Operations Keynote, Dreamforce 2026

Please welcome Chief Executive Officer, Agent Force Operations, Salesforce, Simon Mulcahy.

Hey!

Welcome, everyone. Welcome. I know you're all probably on day whatever, 3, 4 of Dreamforce, and many of you Looking pretty shipshape considering what you've been up to last night, I'm sure, for the concert. I hope you had a great time. I am super, super, super excited today to be talking to you about Agentforce Operations. As the voice of God said, Simon Mulcahy, CEO of Agentforce Operations. And this is about you. This is about your back office. And this is about us and a revolution in that back office which has not seen the level of innovation that it needs to for many, many years, but we're going to change that. And we're really talking about the back office. And if we get this clicker to work, we'll move the slide forward. How about that? What we're really talking about in the back office is all of you have ERP, core master data, and around that you've got modules— Coupa, Ariba, Workday. The list goes on of these kind of core modules, and they're really kind of your core systems, and you spend an enormous amount of time hardening those. But around that is a whole other universe. It's a Wild West, and this is where the real work happens. This is people emailing. This is people spreadsheeting like you have never even seen before. And this is where people are swivel chairing between systems, rekeying information, and capturing things that really drive your business. And this is how work happens today. And it's not just a little bit chaotic, it is absolutely insane. And the problem with this is that it's slow. This operates really at the pace of humans. This operates with human levels of accuracy. And what happens to all the data? Well, that data is locked in spreadsheets, it's locked in emails, it's locked in call logs, and it's not available for your organization to make optimum decisions in the world of AI today. And this is just everything we've come to accept. We've come to accept the cost of this, the sheer cost that if you want to improve a process or make the process run faster, you throw people at it. There's incredible amounts of leakage because the sales that are not happening because paperwork is stuck on people's desks. The cycle time of your business is operating way too slowly. How can you increase those, those cycle times? And how do you deal with the risk exposure that you have? Because the data that you need to make good decisions is locked away in spreadsheets that you can't get access to. These are the costs of holding— of you holding yourself back in the old world. We all thought AI was going to come along and Midas touch this, magic away all of that as a problem. But here's what's happened. It's basically 2 weird solutions that have been presented to all of you. On the one hand, you have the consulting universe, which is whispering into one ear saying, we've got you, we can do this. We will put our best 100 consultants on site. We will work with you. We will cleanse your data. We will rationalize your applications. We will give you the best 3-year roadmap. towards identification you have ever seen. In the other ear are the amazing frontier models, incredible technology, and they're whispering in the same— in that ear, they're saying, give us your data, throw it in the hopper there, and the big flashing green light, headbutt that, and magic will happen. And it feels like they're the two options that you have. Every CIO and COO we speak to is kind of saying, is that it? Are they the trade-offs? There's a third way, a third way that is what you really want, that people are really asking us for. And that is, how do I harness the power, the best of AI, the speed of AI, while also having the auditability, traceability, predictability that I've got used to from the world of software? And how do I not have an army of people from outside of my organization crawling all over me? How can my team do more, faster, cheaper, better? That's the power of AI. The best of AI, the best of software, and the best of your team. That's what people want. And that honestly is why we're here today. And why take my word for it? Why not hear from one of our great customers, Honeywell Technologies? Everyone, please welcome Heidi Smith.

Morning, Simon.

Good morning.

Thank you for the welcoming committee. I appreciate it. Absolutely.

This is how we, uh, this is how we roll. Um, well, first of all, first of all, thank you, Heidi, for being such an amazing partner and amazing customer and helping us continue to build this, uh, product which we're talking about today. But maybe let's start, let's kind of wheel back Honeywell Technologies is going through an incredible amount of change. You are transforming in every corner of the company. Can you talk a little bit about what's going on and the decision you made to— or not the decision for us, we'll talk about that in a minute, but more your decision-making process to kind of think through what technologies you use?

Yeah, so if you followed Honeywell, you know, we've just split into 3 big standalone companies. Honeywell Technologies is really on this journey to reimagine our processes end to end. What does true transformation look like for our businesses? How do we make them as efficient as we can? How do we move things along as quickly as we can with AI and the solutions available to us while keeping a human touch and really making sure that we've got people in the loop where they need to be, uh, involved in that critical thinking and strategic decision making and accelerating things where it makes sense to do so. So we're really looking for those solutions that can bring those things together, that transformation, human touch, and the speed of AI where it's appropriate.

And so how did that lead to you, like, looking at Salesforce, looking at Ruggello? Why us?

Well, first of all, who knew that Salesforce was in the back office? So we learned about Ruggello many months back and quickly were able to see how it applied to many different pieces of our business. Supply chain, agent force operations being the first place, which we've— we can talk a little bit more about how we've deployed use case-wise so far, but there was just an instant fit to how do we take Regrello, automate some of the workflow pain points that we had, the manual handoffs and touchpoints. When we have things sitting on a shelf in our supply chain world, it's dollars and cents. When those cycle times slow, working capital, there's all kinds of impacts to people, to dollars and cents, and to the P&L. So it made sense. It was a really good fit for us when we saw the capabilities of it. We're excited about the roadmap, the really good things that we've seen this week, and look forward to seeing how we can extend it across other parts of our business.

So Heidi mentioned Regrello. It's remiss of me to say, so the company before acquiring was called Regrello, which stands for red, green, yellow, in case anybody's got a quiz on that. Good, that. So back to you, Heidi. Why— can you talk a little bit about the experience you've had with Regrello and what problem you're solving? With the technology?

Yeah. So the use case we have in production today is with our supply chain partners, and it is all about unconfirmed POs. So every week we have somewhere in the neighborhood of 9,000 POs that go unconfirmed and are waiting on somebody to do something with them. That's money. It's time that we could get back doing something else, deploying those resources in ways that could help accelerate Realizing revenue for the company. So that was the first thing that we tackled. I'm very proud to say, and the team that actually did the work in part is in here, you can talk to them if you have time later, but we did a technology project at Honeywell Technologies in less than 2 and a half months to get this going with Regrello. The speed at which we were able to implement this has never happened before in our technology organization. Monica, am I right?

Yeah.

So, yeah, so very proud of that.

Any, any other kind of insights that you've— as you've obviously, you've had a lot of experience with technology over the years. What's kind of the difference, would you say? You said speed. Is there any other kind of insights you'd share with this audience?

Insights are valid. So Simon showed a slide a second ago that talked about cycle times. That's been a huge one for us. Insights, what it's done for us is allowed us to unlock value that other parts of our business see. And we're excited to take this to places like finance and legal. Where can we unstick manual processes with really straightforward workflow automation and associated AI? We haven't started down that journey yet, but we're very excited to do so.

Heidi said one very, very important word here, journey. This is not some nuclear bomb that goes off and the light switch flickers and suddenly everything has changed. This is a journey. Heidi, I can't tell you how happy we are to have you as a partner and how excited we are to be on that journey with you. Everyone, ladies and gentlemen, please thank Heidi Smith.

Thank you. Thanks, Simon.

Very much. Yeah, I appreciate it. Thank you very much. Thanks.

Okay, so what was behind all of that? Well, Heidi talked about the transition that Honeywell's on, the value that they're seeking, the focus that we have on one process to begin with. What's behind that? Agentforce Operations powered by Ruggello, which you all now know what that stands for. What is Agentforce Operations? Well, it is a platform to transform back office processes, starting supply chain and finance operations. If you can describe your process— I didn't say application— your process, your end-to-end process, We can digitize that in minutes. Not only that, but we instead— we remove the human from program managing it and put in place a super agent that then orchestrates work from the beginning to the end of the process. And that super agent then brings in specialized back office agents to execute work and maximize automation. Also calling in humans where the humans are needed for value-added work. And not only that, as you will see, hopefully, and blow some of your minds, we also make it really easy for you to integrate this into your back office systems. Now, what's really, really exciting about this is that this is not designed to be another platform that IT has to build code. No. This is a tool that is the perfect tool for IT and business to work together. It requires very little effort from IT, and it maximizes the business's ability to control its destiny and supercharge its processes quickly. And why do I say that? It's because we built this on-site at very, very demanding companies like Dell, Daimler Truck, Flex, Microchip, and many, many more. Big global, very complex businesses. That, ladies and gentlemen, is Agentforce Operations. And the business value that Heidi was talking to is being experienced in every single one of the customers that we work with. And what is that value? Well, it starts with the mechanical value of just saving cost. Cost and leakage we target, and this is tens if not sometimes $50, $60, $70, $100 million of savings. It's amazing how much money there is that you're leaving on the table. But that's just the start. As Heidi said, it's about cycle times. You don't cut your way to being a business for the next 150 years. You need to build a business that can operate faster. Your cycle times need to improve. Your decision-making needs to be informed with better decisions. Which, by the way, that means that all that data that is lost and hidden in spreadsheets needs now to be available to you. So these are the types of value that we see our customers getting. We point all of that kind of energy and our excitement and our innovation into the back office, and we focused it originally in the world of supply chain. We were born during COVID when everybody freaked out in the world of supply chain. That's where we focused. Michael Dell in particular was championing us into this particular area. And you can see these processes, these families of processes and the pain. If you know anything about the back office that these, these processes represent, these are intrinsically human-first processes in the old world. And as you're going to see in the new world, they look and feel very, very different. Now, the world of supply chain took us neatly also into the world of finance operations. And so we're also very, very heavily engaged there. And as you heard from, from Heidi, legal, finance, sales ops, marketing ops, these are all logical areas for us to go into. So take invoice audit, for example. Invoice audit is a really great place to start. Why? Because how many of you receive tens, hundreds of thousands, sometimes millions of invoices? And how do you check them? You probably have humans checking them, and maybe your human resigned, so you had to get a third party to do that. And what do they do? They do gross error checking, not every single one. But in today's world, that's ridiculous. Why would you not check every single cell of every single invoice? Why would you not check even the handwritten ones? But you don't need a human to do that. You can do that perfectly. And not just a 3-way match or a 4-way, why not a 10-way match? That's how we look at just that one process. Imagine that thinking applied to all of these processes. This is the world of AI first. So we were acquired by Salesforce in October, and actually by about February, we spooled up a team inside Salesforce in the Digital Excellence team. Basically, business IT is a team of 3 people. And 3 people, they go into each business unit of Salesforce. So finance, for example, which is actually going gangbusters now. Our CFO is very, very happy. And they basically do an enablement and a buildathon. And then from that, the team gets inspired. And now the team, as you can see on the right here, the team then goes and identifies the use cases, prototypes the solutions, An IT team of 3, remember, enables the business, hooks it up to the back office, to the data, and ensures the governance is correct, and off we're running. Since February, 50 processes across 8 functions. This is not a onesie-twosie, one process taking 3 years. It's a ridiculous old-school way of looking at the world. This is the speed of how fast your organization can operate. This goes viral. But with careful IT governance. So I'm really excited to basically run off the stage now because vastly better people are going to come and show you this technology. And we're going to show you how we automate your human-first processes into the world of AI. We're going to show you how we make it really easy to integrate that into the back office. And then we're going to show you how your processes should always be continuously improving in this world of Six Sigma kaizen-type thinking. So with all of that, I am very, very excited to introduce you to Amy.

Thank you, Simon. All right, get the rest of my posse up on the stage and then we'll start. So workers have been buried in non-added-value tasks They're spending time opening up documents from suppliers, from customers, from vendors. They may be chasing down updates for projects that they're working on, and we need the business to move faster. And so we're extremely excited to introduce you all to Marshall today, your supply chain and finance operations agent. Marshall is able to build your processes As Simon said, in minutes, Marshall orchestrates the work across humans and specialized back office agents, and Marshall executes this process all headlessly while creating a clear audit record for compliance. Now, we wanna bring this all to life with a demo, specifically with one of our wonderful customers that you just heard from, Honeywell Technologies. Who's looking to automate their purchase order lifecycle. And what I'm gonna show you today are 3 things. The first is how they digitize their process in minutes. We're not talking about months of work to get live. The second is how agents are doing all of the work, freeing up Honeywell's team's time. And the third is how it's built for scale. Now, before we dive in, I want everyone to give a round of applause for Tom and Sama over here at the demo desk. So, Tom, you ready to rock and roll? All right, perfect. As Simon just said, if you can describe your process, Marshall can build it. So let's take something every company does— buy things. Honeywell wants to buy parts, so they create a purchase order or a PO. I want 1,000 parts for this price delivered on this date. But until the supplier actually confirms when it's coming, Honeywell's questioning, well, will it come? How much will come? And when will it come? Now, one unconfirmed PO may not be a huge business problem, but when we're talking about a company the size of Honeywell, as we just heard from Heidi, This is thousands of unconfirmed POs at any given point, and that's driving revenue leakage. Production lines can slow down. Customer orders can also be delayed. And in the old world, each week buyers were spending hours comparing their spreadsheets to POs, chasing down suppliers, updating those spreadsheets and systems. Well, Honeywell said no more. They wanted to drastically reduce the amount of unconfirmed POs that they had. They wanted to improve their business, their bottom line, and ensure that we— they were delivering on time to customers. So what did they do? They flipped the switch with that process. They went from human first to AI first. And so that's what we're doing right now. I'm simply describing the process. We're letting Slackbot know, because Slackbot is actually communicating with Marshall. We're uploading a standard operating procedure, so if you have any documentation, process flows, Marshall can use that. But let's face it, not all of us do. That's okay. Marshall is trained on hundreds of thousands of back-office process flows, so whatever you have, Marshall can go to town. And here we see he's starting to build the process. But let's look behind the curtain to find out what exactly Marshall is building. First, he's taking all of that messy business context and he's building out a complete end-to-end executable process. He's starting to build the stages and the tasks, the actual structure of the process. Next up, He'll start generating the forms and all of the data. What data is needed to start the process? What data is captured along the way? And ultimately, what data is needed to complete this process? But that's not all Marshall does. Marshall's also identifying areas to automate the work, bringing in specialized back-office agents. You can see him doing this right now.

Wow.

You don't see that every day. And it's not just that. This has been revolutionary for our customers in putting it into their hands. Customers like Dell went from 100 to 300 processes in just 9 months. This baby goes viral. And let's face it, every customer that uses Agentforce Operations doesn't just want to build their process, They want to iterate on it over time. And so we've made it incredibly easy for process experts to do just that. Maybe I want to change an assignment from a human to an agent. Guess what? All I have to do is talk to Marshall. Hey Marshall, can you change that assignment? That was hard.

Done.

Marshall takes care of it. Maybe I look at this and say, we've got a lot of agents completing work. Let's bring a human in just to look at what the agents have completed. Once again, I just have a conversation with Marshall. This is awesome. This is how you go viral. This is how teams get enabled because I don't have to understand where to click, how the system works. I just have to talk to Marshall. What I've just shown you is how we've taken a legacy human-first process, And Marshall has transformed that into an AI-first business process blueprint. This process is ready to run. It can run weekly, it can run thousands of times a day, an hour. So let's see it in action. Starting this process is just as easy as building it. I'll return to Slackbot, because once again, Slackbot can communicate with Marshall. And I'll say, hey, Slackbot, let's start an unconfirmed PO. Type that in. You can, of course, do this from Teams, from email, and anywhere the work actually begins. And so, the first thing I want to draw your eye to is how Marshall is now starting to orchestrate this process. He's orchestrating towards a timeline. He's bringing in the specialized back office agents to execute tasks. He's chasing down updates, chasing down information. All to ensure that this process completes reliably and predictably each and every time. The second thing that Marshall is doing is creating a complete audit trail for all the work that is happening. Every stage, every task, every agent interaction, all auditable. Every decision, every handoff. And frankly, this is what we should expect when we're running enterprise-grade processes. This was critically important to Honeywell for SOC compliance. But that's not all. The third thing I wanna draw your eye to are these specialized back office agents. I think we all know that if you make calls to an LLM, you get probabilistic results. But if I'm running financial calculations, is mostly right okay for my organization and my back office operations? Definitely not. And so that's why these agents are built for predictability and accuracy. They're built on thousands of innovations to deliver the speed of AI with the certainty, the accuracy, and predictability of software tools. And now within the unconfirmed PO process, these agents take care of a lion's share of the work so that when human judgment is needed, We then bring in the Honeywell buyer, the person that's managing, unfortunately, all these unconfirmed POs. But now it's fortunate because the agents are taking care of the lion's share of the work, as mentioned. And so when they're ready to review the work, whether that's in email, in Teams, or here in Slack, the agents have already done all the aggregation, the synthesis, the analysis. Everything is teed up for this buyer so that she just sees the conclusions. She has to click. Done. That's how simple work should be. Now, what I just showed you— we'll hope for a slide. What I just showed you is how Agentforce Operations is revolutionizing the back office in how we're automating processes. But we're doing a lot more than that. Let's now talk about how we're integrating into your core systems that run your business. So welcome to the stage, Meredith.

I'm going to start with a stat that frankly is a little bit scary. On average, enterprises use 957 applications, and it feels like even more when you can't remember your password. And out of those 957 applications, only 27% are integrated. That means that 3/4 of your business, of your data, of your systems, of your processes are living in silos. That's a huge issue for any department, but it's especially a huge issue in the back office because the last mile of any process is integrating into core systems. I don't really need to tell you this. You know, integrations are a necessary evil of big business. And if you aren't integrating them, you are likely paying a human to go in between SAP and Oracle and Coupa and a million other systems. It's slow, it's expensive, it's manual, and no one even likes doing it. So we asked ourselves this question. What if the need to integrate didn't hold you back? And we know this is hard today, because today, you really have 3 options. An RPA, but it's brittle. It breaks when a UI changes. Custom integrations and code, though sometimes necessary, are expensive. They take a long time. They require IT work. And has anyone ever sat on an IT backlog and just waited for your process or your project to be picked up? And finally, computer vision. Computer vision is amazing technology, but it's not 100% accurate, and that's not good enough for the enterprise. So I would like to introduce you to Systems Integration Agent. Systems Integration Agent is just absolutely incredible technology. We're going to take it a little slow, because I really want you to see it, and I want you to understand it. Systems Integration Agent does a couple things really, really well. First, it works anywhere, headlessly. All it needs is access to a sandbox and a login. Secondly, it's live in hours, not months. It self-learns, so it doesn't need code, it doesn't need an API, and it doesn't need RPA. Next, it executes deterministically so that you get accurate results every single time. And finally, it self-heals. So So if a UI changes, like they do all the time, it quickly goes back into learn mode and fixes itself. I don't want to talk about it too long, though. I really want to show it to you, and I'm going to show it to you in the context of a customer, Freeport-McMoRan. Freeport is not exactly a household name, but it's an incredible customer. Freeport is using Systems Integration Agent to integrate supplier onboarding. Freeport is one of the world's largest copper mining companies. So just to put this into context, and just to put this in— give you the picture of how many suppliers they have to onboard, they have mines all over the world. This means they need to get people all over the world, They need to get equipment. They need to get copper from one mine to a supplier. So they have a lot of suppliers to onboard. So I'm going to show you this demo, and I'm going to tell you about Systems Integration Agent in 3 steps. One, I'm going to show you how it learns systems without code, API, or RPA. Next, I'm going to show you how it executes work deterministically. So that it gets it right every time. And then finally, I'm going to show you how it self-heals. But before I show it to you, we need to take a look at the old way of what this looks like— looked like for Freeport before Systems Integration Agent. So let's take a look. What are we seeing here? We are seeing a video at 20x times of one of Freeport's very best analyst inputting information. They're updating forms, they're clicking, they're filling out fields, they're pressing a button and waiting for it to spin and actually load. And this is really hard. Before Systems Integration Agent, it was taking them hours and hours of this manual work. We won't stay on this too long because some of you might be having flashbacks to when you were an analyst and you were working in SAP. and inputting this information. But for those of you who aren't aware, this is real. This is what it's like, and this is a human doing the work. And guess what? They can't do it at 20x speed. So, let's fast forward. Let's go to today. Let's show you what this looks like with Systems Integration Agent. Who wants to see it? All right, let's do it. Okay, step 1, magic trick number 1, because it really does feel like magic. Systems Integration Agent self-learns. You give it access to a sandbox with a login. Think of it like a classroom, and Systems Integration Agent goes in and learns every business rule, every process. It clicks around, and the LLM is helping it reason and figure out exactly what to do. It even learns conditional rules, like if the vendor's in the United States, We have to have a tax form. That is just the least surprising thing of this entire presentation. Okay, so now that it has learned, it's going to store all of this information. So as you can see, this takes about 90 minutes, and for Freeport, they actually gave us access to their systems on a Friday. Everyone headed out for the weekend, and it was ready to go on Monday. So here is bringing me to my next magic trick, and that's that it executes deterministically. And it does that because after the learn phase, it locks in an MCP. And this is your library of MCPs where it's learned all the business rules and all the processes. So let's see how this looks in action. Let's look at that second magic trick, execute deterministically. We're going to go back in Slack, because while you're seeing all this stuff in the background, really, your team kicks off in Slack. So, Marshall calls Systems Integration Agent to get started, and now Systems Integration Agent is filling out all those fields. It's going in between systems and pulling and inputting data. And while it's a demo, you can see that Sama, my wonderful demo driver here, is hands-free. So Systems Integration Agent is doing all of that really, really difficult work that the analyst was doing before. As you can see, this used to take hours and hours of time, and now Sam is hanging out. She's doing other work. She's working with customers. She's doing more value-added work while Systems Integration Agent works in the background. But creating the supplier isn't the only step of this process. Now we have to create a purchase order in Coupa. And Systems Integration Agent has already learned how to do that. So we are going to go into Coupa, and Systems Integration Agent gets started. But it looks like there is a small snag. And that snag is that the UI changed, like they do. But it's no problem for Systems Integration Agent. Because we go right back into that self-learn mode, and it heals. It completely takes care of this and continues to execute the process exactly as planned. Okay, so I have shown you how Systems Integration Agent learns the process, how it executes it deterministically, and I've shown you how it self-heals. All right, process done, process complete. Let's take a look at some data. Don't take my word for it. All right, so we're going to take a look at some data. So, we did some testing on our own, and we did 250 hours of benchmark testing on easy, medium, and hard processes. And what you see here in the first column is Systems Integration Agent. This is the best of AI with the best of software. And then in the next column, you're seeing a frontier model alone. As you can tell, accuracy, 100% correct every time for Systems Integration Agent. For frontier models, they do pretty well on easy processes, but on hard processes, they actually sent the multiple emails to the wrong supplier.

Model cost.

So with Systems Integration Agent, we use the LLM to generate the MCP, but then it's locked, and it's just executing every time. It's not recalling the LLM, so you're not incurring the cost of tokens. And then finally, speed. This is just the beauty of software. It can move really, really quickly once the plan is locked. So in easy, medium, and hard processes, System Integration Agent really took the prize. So I now would like to call to the stage the CIO of Freeport, Robert Kovacs, to talk to us a little bit about his journey with Ruggello and Systems Integration Agent.

Hi. Thanks, Mary.

Thanks for joining us.

Yeah, thank you.

All right, so we haven't been working together that long, about back in 2025, so I want to go back there. For a minute, what were some of the challenges that you were facing that you wanted to solve?

Well, let me just step back for a minute. So with a mining company, you— most people may not think we're very technical, but we've been using AI and machine learning for quite a while. We come up with our own reagents similar to the way pharma does. We use it for drill and blast. So we, we've been pretty astute with the machine learning and AI and We knew that we wanted to improve our processes and really help out the G&A space.

Okay, incredible. What made you bet on Regrello?

So we did some research, and it seemed like one of the top ones, and we spoke to Aman and Simon back before the Salesforce was maybe a twinkle in their eye. I'm not sure, but—

It might have been the opposite.

Yeah, so— We also pulled together the 4 of the top consulting firms, asked them what they thought, and it all kind of pointed to Ruggiello. We just, we knew we wanted to get started. We wanted, we knew we could capture that value, particularly if we could integrate processes across functions.

Mm-hmm.

And I think the other key thing was culturally the team really worked well with us. You know, like they were about getting things done and without bureaucracy, just like we are. So Worked well together.

Great. And I understand from chatting with you earlier, you started with just a few processes and you have just expanded really rapidly. How did that work for you?

I think overall it worked well. We learned a lot. We did the first 4 and got to value very quickly, and then that effectively just generated a tremendous amount of demand from the business. We, we also then decided, okay, we have to put a little bit of structure because We want to avoid everyone ending up with like 1,000 Access databases and creating some technical debt. So we put a little bit of guardrails around it, but it's, it's been going great. We have hundreds of processes and a whole queue lined up of more.

So incredible. So you use Systems Integration Agent today. What are some of the use cases you're using it for?

There's just— we probably have it in about 60 cases. So it's, you know, you showed a fantastic one. What you don't see on that one is there was a point in there where the person just had to push a button and go out to Ariba and just wait. And it's like, well, how long do you wait? 10 minutes? We don't know. Just keep hitting the button over and over. And so it took that out. And then anywhere where you were working with something without an API, it worked just really well. And I think some people may have heard this the other day if you listened to Jeff, but We went live on a Friday. The SaaS vendor changed the stuff. We didn't know about it. It kicks off on Monday. We're watching it go. It paused. We're like, uh-oh, oh no. And then it just went, relearned, healed, and then kept going. So it was really impressive.

Yeah, that's great. What kind of value is Systems Integration Agent driving for you?

I think we're capturing— actually, we're capturing a value tracker on all this. And so we look at every month, what value we're getting and our forecasted value. And we're capturing, you know, some of it is we don't need contractors to do certain things anymore. That goes away as hard dollars. Some of it's soft dollars where like thousands of hours are freed up for people to go do other things. Other cases, it's things we could never get to because we didn't have the time. So like material deduplication, we're doing that. So it's just quite a few.

Okay, great. What's next? Anything else on the roadmap?

I think ideally, you know, we're heavy into the back office now. We'd like to pull this into operations. If we could tie it into something like mine planning, then we know we can drive a lot of value there. And then also maybe integration between maintenance and supply chain so we can better optimize our inventory.

Great. So our team has been so impressed by your journey to AI first. What are some things that you think some CIOs or some leaders in the room should know about getting started?

I think the main thing is just to get— start with the first few and get going, because once you understand what it can really do, then it just starts to click on all the different places you could use it. I think that's the first thing. The second thing is getting the business aligned is extremely important, because if you have the process and people aren't using it, it's, you know, it limits the value. So we have our top leadership aligned. We look at a queue of everything based on value. They decide which ones we're going to do, and we've got the backing of our most senior leaders to push this forward. So I think the business alignment is extremely important, and then putting enough governance around it to make sure you're doing the things you need to, but not enough to slow it down.

Awesome. We've heard that a couple of times, I think with Heidi too. Simon mentioned it. Just bringing together the business and IT in partnership. And then we want to be on that journey with you and be that partner for you too. So it's just been absolutely incredible working with you. Thank you so much for coming out, and thank you for your partnership.

Thank you.

Okay, so as we continue this journey to revolutionize your back office, we've shown you how to automate processes. We've now shown you how to integrate them, and now we are going to show you something completely brand new. We're going to show you how to optimize them. So please welcome to the stage Marcello, GM of Intelligent Process Optimization.

Thank you, Meredith. Oh guys, it's great to be here up on stage. So Amy showed us how we can automate our back-office processes, and Meredith showed us how we can integrate them with our enterprise systems so that these processes do not live in isolation. But how do we make sure that what we build is actually the optimal process? And by optimal, I mean a process that actually meets our stringent KPI targets, our time to serve, Our cost to serve, our first touch resolution rate targets. Those are the targets we need to operate towards when we run business processes, when we run our operations. And most importantly, how do we make sure that those processes remain optimal over time? It's a pleasure to introduce you to Appromore. Salesforce Intelligent Process Optimization platform. We built Appromore a few years back, and we got acquired by Salesforce just a month after Regrello. We are about to launch. We're going GA in the Salesforce ecosystem next month. What can we do with Appromore? We can help you discover your real processes. Before, Amy told us that when we execute our workflows in Regrello, they leave behind an audit trail. Well, that audit trail is a digital footprint Appromore uses as input to automatically discover the process. So we can actually see visually through a process map what work gets done by whom and when. The actual work. And not only from Regrello. We can mine processes from Salesforce, from SAP, from any other system and even desktop applications to capture the swivel chairing that Simon was talking about. And what do we do with that visual representation of the process? Well, we want to analyze it. We want to see whether there are still frictions like bottlenecks or rework loops and how they impact on your KPI targets. And guess what? That information is then used as input by our optimization agent to identify opportunities to optimize the process. You set the target, I want to improve my time to serve, I want to improve my first-touch resolution rate, and Appromote will find you the recommendations to optimize the process. And then deterministically, we actually simulate that recommendation to make sure that the solution we are proposing works against your data, so you can build confidence that that's the right way to optimize your process. From there on, we can continuously monitor that process. But let's see that in action. I'm going to show you how we can discover a process running in Regrello, how we can identify opportunities for optimizing that and simulate those changing— changes, and then push them back into Regrello so that we can build a new blueprint and run it from there on. Big round of applause for Vaishnavi that is going to run the demo for us. Thank you, Vaish. Vaish, let's put that to action. So let me start from Regrello where we left with Meredith. This is Freeport's supplier onboarding process. I just right-click on my blueprint and I click on Monitor with Appromote. And from there on, I automatically fire up a dashboard. Where I can keep track of all the metrics and key performance indicators that are relevant to me. And like Regrello, Appromora is completely no-code, so it's really designed to be used by practitioners like you to monitor, analyze, and continuously improve your processes. So here I can see how many applications I'm running, what's my time and cost to serve, and you see there on the right, I've got some KPIs. Well, the time to onboard, I'm not doing well there. I'm not meeting my KPI of onboarding suppliers within 3 business days. You also see the straight-through processing isn't good. That measures how much rework I need to do to onboard a supplier. Great dashboard. It's telling me that I've got a problem. I want to see where that problem is coming from. And most importantly, how I can fix it. So, Appromore, show me the process under the hood. So now Appromore is literally discovering in real time the actual process behind those numbers. Now I can see step by step how Freeport onboards their suppliers, from picking up a case from the queue all the way to fulfilling that case and closing it. And I can see in darker blue color those activities that are performed the most. But also, hey, Appromore, show me the duration. Show me where things go down. You see now in dark red, I've got bottlenecks. That's probably where I'm losing time. That's why I might not meet my time to onboard. And that's a bird's-eye view of my process. Hey, Appromore, show me all the details. Show me all the possible pathways that I'm executing when onboarding my supplier? Well, there is about 1,000 different ways of onboarding a supplier. Even if that process is being automated, there is still complexity that I can reduce. You see, that's impacting my straight-through processing. Now let's take all these insights and feed them into Agent Forte as my process optimization agent. Hey, Agent Forte, Tell me why I'm not meeting my time to onboard KPIs. So I'm going straight into asking the real question. So now Agentforce is going to tap into the tools provided by Appromore to root cause those KPIs and give me ideas, concrete ideas on how I can optimize my process. You see that it picked up on the key bottlenecks, on the key rework loops. It highlights them in the flow. And then down the bottom is actually giving me some concrete recommendations to improve my time to onboard, like automating the routing of cases on arrival or adding an additional AI agent to review the documents automatically. And it goes on and on. Well, you know what? Let's feed these recommendations now back into Regrello. Let's build an updated blueprint that takes into account those recommendations. And that's what Appromore has just done. It gives me a link, I can go back to Regrello, and from there on I can kick off a new version of my blueprint. And then I can continuously monitor that. So we've seen how we can automate our process, how we can integrate it with our enterprise systems, and how we can bring that into a continuous monitoring and improvement journey.

Thank you.

Back to you, Simon.

Excellent.

All yours. Thanks very much, Marcello. So there you go, ladies and gentlemen. Hopefully beginning to get a sense of just the sheer power of this technology. We're done with this old world of living in the dark ages with this Wild West of chaos processes that are slowing your business down, costing you a fortune. Now's the time to move to the AI-first world, and it's not difficult. This is not a big IT lift. This is a very, very exciting journey that all of you should be on. We're on the same journey, by the way, continuously innovating. The level of innovation that's going on in this team right now is actually, frankly, mind-blowing. We can't even think further than 3 months ahead. There's just too many things that are coming out. soon. The level of innovation is insane. This innovation is not just in the technology universe as well though, because you've got real businesses that you need to introduce this into. So we're innovating in the same way about how this becomes live in your organizations. We passionately hate the concept of POC, which I'm sure you are all in POCs and you will probably never leave those POCs either. It's a ridiculous nightmare for everyone. For us, it's a 1-day next-generation POC. If we cannot prove to you beyond any reasonable doubt in 1 day, run away. You won't, I guarantee you, but it's 1 day. It's 1 day is all it takes. Once we've done that, we move into a proof of value program, 3 to 4 months. Why? Because we want to take 2 processes live in your organization at speed with our team for deployed. To help you do that. Why? We need to help you see this together, show you how it works. But very quickly, you start to also see all the other areas, as both Heidi and, and, and Robert shared with you. That then leads to the new scaling this out, and you scale it out with your own center of excellence in different ways. There are different ways of doing it, but then now you've got the sheer power in your organization to very quickly and virally change how your business works and move it to the AI-first world. This is not a hope for next year. You can get going on this right now. With all of that, I just wanted to say thank you all, and I wish you all the very best for the remainder of Dreamforce. I want to say thank you very much to our team and Aman Naimat there, who's the, the founder, technical founder and genius leader of AI Labs at Salesforce.

Thank you all.

Thank you.