Salesforce Just Helped Microsoft. You're Picking The Winner Of A Trillion-Dollar Fight.

On October 27th, Salesforce made Microsoft's agents much, much easier to use. And on the same day, Oracle said it's building a way to bring its own apps, its Oracle apps, into ChatGPT. Rivals are opening doors for rivals. Nobody does that out of generosity. Every one of these companies is chasing the same seat. They're chasing the software you depend on to get work done across everything else you own. If you're running a team, you're picking who gets that seat right now, one agent at a time. Your company is about to be in the middle of a trillion-dollar fight. The nine biggest cloud companies are on track to spend $957 billion on capacity this year and more than a trillion and a half next year. And I want to talk about it today because I think we need to understand how some of the long-term dynamics in tech are starting to shift. For a long, long time, companies that ran our B2B software, that ran tons of the computers and software we use day-to-day, were complementary. They had different roles, they had different contract types, they had different customer types, and everybody kind of kept in their lane.

Microsoft and Salesforce are great examples here. That is starting to become more contested. And it's not that anyone is saying I'm picking between the two directly on traditional software, it's that it's becoming contested around the leverage that AI agents provide. And that sort of brings me to the core takeaway here. The thing I want you to remember and think about for the rest of this video. AI agents are changing what we think of as competitive. They're changing what we would describe as the high ground in the economy at large. And that means that these traditionally separate companies, like Microsoft and Salesforce, are finding themselves in odd competition moments, even if they're still largely complementary, because the agent economy space is so valuable. Think of it as the high ground in a battle. If you have the high ground, it becomes easier to do anything else on the landscape. In this world, the high ground is who owns the agent relationship. Who owns the relationship that mediates the rest of the computing experience that you have at work. Everyone's betting there's going to be one big answer to that. And everyone wants their company to be the big answer to that.

Everyone thinks the future of computing is agents. The future of computing at work is B2B agents and agents that can use your work computer. You're going to do much much more computing that way, and they want to own that relationship. It's not that they want to own every agent, it's that they want to own the control plane, the experience setup, the framework that allows you to use agents across your computing environment. That's the trillion-dollar battle, and we're going to get into it today. If you're new here, I'm Nate Jones. I spend a lot of time taking these kinds of announcements apart and thinking through what they mean for how we actually work. So, please subscribe if this kind of analysis is helpful to you, because the choices these companies make are going to shape the tools that you and I get to use, whether or not you're the person who's buying them. Why is Salesforce choosing to help the competitor's product work better here? There's an answer to that. And but before we get to there, we have to cover the wider landscape. Because on the same day, OpenAI and Oracle announce an expanded partnership as well. Oracle is developing a workspace that brings its business applications and data into ChatGPT.

That workspace is still in development. But the direction that it's going is very clear at this point. Because you're going to be able to keep Oracle's applications, and yet you're going to be able to get more of the work done through OpenAI. So, again, we see this pattern of companies helping each other. These companies are helping each other reach their customers while also competing for a bigger role in those customers businesses. And that's becoming a very common pattern across the valley right now. First, let's start with the way a lot of companies bought software in the past. Microsoft sold you Outlook, they sold you Excel, they sold you Teams. Salesforce has sold you the system where your sales team manages customers. Oracle might have sold you the finance applications. These businesses already overlapped and competed a bit, right? Microsoft has its own business application. Salesforce also owns Slack. It's not that there's ever been a clean agreement that everyone would stay in their lane because that's anti-competitive. That's just not how it works. But a customer could buy from all of them and often did, and each could have a substantial relationship with a different part of the customer company, and that was super common. But now let's fast forward.

Let's look at a real job that customers need to finish and see how the line is blurring between traditionally distinct roles. Closing the month does not respect the boundary between spreadsheets, between emails, between accounting systems in the finance department. Selling to a customer doesn't stop at the customer record. People have always crossed those boundaries in the past. People have always read the email. People have checked the numbers. People have asked someone what changed and decided what to do next. A lot of what made software useful happened in the person moving between the applications. You and me, right? As you ask agents to do more of that work, somebody has to supply the software that puts that agent layer together. Which information can it use? Which instructions are current? Who can approve the actions? Where does that result go? The agent needs answers to those questions even when the information and the people involved are very spread out across multiple different products.

That's what brings these companies into much more direct competition. Microsoft can say, "We already have your documents. We have your conversations in Teams. We have your employees working today." Salesforce can say, "Hey, we understand your customers. We understand the processes you use to sell to them and how you ring the cash register. The AI labs can say, "People are already asking our product to do the work. Just connect the applications they need." Those are different angles, but they're all coming for the same position. Let's look at Salesforce for a minute. Salesforce is the original B2B. They basically invented the B2B SaaS category. In that world, almost every startup I have ever been at or talked to would absolutely kill for their distribution because they are working with sales teams and companies across so many continents, so many scales, so many verticals. And those companies in general have absolutely no interest in rebuilding all of the stuff that Salesforce gives them just because a new model got better reasoning. That's just not a a relevant commercial question. So, their distribution advantage is real. But, the question of the future is not just about distribution advantage. The question of the future is about how the agent relationship is mediated. So, I've talked about the fact that Salesforce has done a great job opening up their data set so agents can work with Salesforce data. I think that's true.

The question of the future is how do we manage and control those agents so they become an enterprise scalable approach to intelligence. That is the question of 2026 and 2027. That is the question that is getting these deals done. Everybody is trying to figure out how they can meet customers where they are with the systems of record that they already have and give them the flexibility to control agentic experiences through corporate relationships they already understand. And that's exactly why Salesforce leaned into the control plane solution that they talked a ton about at Dreamforce. I was there. It's also why Microsoft has been leaning into a control plane solution. It's why Oracle is now starting to talk about it. Everybody is talking about it from their angle because everyone understands that it's less about me and my personal agent in 2026 and 2027, and it's more about how agents are an intelligent fabric across the business, and we have to be able to manage them as a corporate asset. Who gets to do that? That is a highly valuable position, and that's something where you can be very agnostic about which agent does what. You can have Claude, and you can have ChatGPT, and you can have Quan, and you can have any agent you want, but you have to have a common control plane to make it work, and that is what everybody is trying to figure out how to do well. And so, if we come back to the bridge at the top of

the video and how Salesforce and Microsoft are partnering together, things start to make more sense because that bridge works with Copilot Studio agents, and the customer doesn't have to rebuild their agent to make it participate with Salesforce. So, Salesforce that way can participate in coordinating the work even where Microsoft supplied a good part of that work through Copilot. And so, in that sense, it might seem like they're supporting a rival, but really it's a way to earn more responsibility from the customer by both Microsoft and Salesforce showing that they understand where the future is going. They understand that customers want to have choice here, and they understand that they're both encouraged and incentivized to give customers choice and control. And this is also why better models don't automatically end this competition. When you're an individual developer, you might just say, "I want the best model for the application. I don't have to think about it." But in an enterprise environment, a more capable Microsoft agent could make a process that's coordinated through Salesforce more useful. Both companies win. Salesforce could gain from that improvement without having to supply the model, and Microsoft meanwhile has every reason to offer the customer more of the surrounding system itself. And so, they can both benefit and yet compete hard over the next purchase. And in that sense, the customer relationship is really being negotiated product by product as we move into the agentic economy. And I know you've heard the phrase agentic control plane a lot in

this conversation. I should explain it. It means software for overseeing agents, sometimes including coordinating their work. And for this argument, the valuable position I'm talking about is quite specific. Your employee brings an agent a job. It can get the relevant information, it can get the right tools, it can apply the company's rules, and it can bring that work back for the employee to review. This could be you. Maybe you're the employee, right? And so the agentic control plane can help determine when and where that work can happen, how that work happens, what the bounds on that work are, and essentially how much envelope the employee has to be productive. The control plane shapes our experience in the workplace, the way Windows shaped our experience in the workplace in the 1990s. Like everybody had Windows at work in the 1990s. Many of us still do, but it was absolutely dominant back then, and that's kind of what I mean by the control plane. But we don't have a Windows for the control plane yet.

Everyone's trying to figure out who that's going to be. Let's think about the experience that a customer might have with your business today. Let's say they were promised something and it was supposed to be delivered yesterday. And today, they're talking about their issue to a different employee or maybe even a different agent. Does that memory of the promise the customer had survive? Does the person handling that know what was agreed upon? And can they act on that and carry it through? If your software makes that really intuitive, you have a good reason to keep paying for that software. And that's the kind of relationship that these different players would like to have with you as you begin to bring more and more agents into the workplace. And if you want a specific example of how this might look, Salesforce's agent context engine, which is currently in preview, aims to supply shared business context to agents on many different supported platforms, including agents built elsewhere. Microsoft's site queue products are a bit related, although not quite the same, which makes it hard to compare.

But, they bring together information or context about how people work of business data, the definitions and rules agents need to interpret all in one place. These are efforts that are overlapping, and they're really aimed at making existing information and customer relationships inside the business more legible, more useful for agents to work against. And those definitions that we're talking about, the way I'm talking about this really matters. A spreadsheet, for example, can contain the correct revenue number, but an agent might use the wrong meaning of revenue when it talks. A customer record can be accurate, but miss a promise made in an email yesterday. So, connecting the systems gets you access to that data, but someone is still going to have to establish what the information means for a decision in front of you. The business still has to supply that judgment. And vendors want to make sure that they are there at the point of attack, where the system supplies that information so that humans can make good decisions.

And that's why they're all attacking it from different angles. And that's why the space in the adjunctive economy is so valuable right now. Now, let's bring OpenAI and Oracle back into the conversation. Their announcement describes a finance employee using ChatGPT to investigate costs and understand purchasing decisions and identify what is left before month end close. The The proposed workspace they're working on together would bring information from finance and procurement and planning applications all into that work that the employee could do, subject of course to company permissions and approvals, which are common restrictions on control plans. It's supposed to be an experience that helps finance professionals in particular get their work done more effectively. And Oracle has really good reason to make that easier. The customer gets way more use from the Oracle applications they're already paying for with this partnership done. Open AI also wins because they gain a chance to become where that employee works through the problem. And if that works well, the next request might start in ChatGPT 2. And you can add another source and bring in another department and you can help in another part of the financial close, maybe, right? And so that's how you take an initial play around finance that makes sense for Oracle and ChatGPT and turn that into a larger role. Now, Open AI has been pursuing this in other ways, too. They launched Frontier back in February. I covered it then, connecting

business systems and context and agent work and permissions all in one place. The Oracle news is just a fresh example of the overall push in that enterprise direction. It also complicates the idea that the established software companies will simply keep the most valuable position in the stack because they have the data. An incumbent, in this case Oracle, might help a new competitor, Open AI, become more useful because the partnership has benefits for both of them and in the process change the dynamics in the race overall. And owning the data doesn't settle where the work has to happen, either. Microsoft Fabric is a good example here. Microsoft's Fabric allows you to reference external data that lives elsewhere in place as you do work in the Microsoft ecosystem. So you don't necessarily need a giant database migration before another supplier can become really important for you. The customer is in charge. The customer is king. The customer can retain existing systems and change the way people use them pretty freely. And that lowers one of the biggest barriers to this kind of competition spreading across the B2B enterprise space. Now, there's a smaller version of this competition that I've already seen pop up and it helps us understand why a customer relationship can be strengthened so much when this agentic piece goes right. I talked about a recruiter I'd spoken with whose story

I I anonymized, right? Her agent pulled candidate profiles, drafted a short list, and passed it to the hiring manager after she reviewed it. Over time, she taught it how she screened people, what different hiring managers wanted, and which candidates actually got hired. When changing AI providers came up for her, her manager had not understood how much work she'd put into that arrangement. The candidate data, remember, was only a part of what she needed to tell that story. There were her instructions, her corrections, the skills she'd built in the way the tools work together that were all part of the story. And keeping access to just the raw data would not really preserve any of that. This is the part of the customer relationship that these companies have reason to care about. This is where I want to get very precise here. The useful knowledge that accumulates while someone does the job, where does that live? How does it get used next time? How much do you have to reconstruct if you move?

A vendor that helps a company preserve that work and make it useful to more employees can become very difficult to replace for good reasons, because replacing it creates way more work for people who don't want to do it again and who want to stay productive. The recruitment story's a good example here, but at company scale, the opportunity is really to help entire teams work this way without any given person rebuilding connections and instructions on top of that data layer. And it doesn't happen automatically. Last month's correction does not automatically mean that it's going to make it into this month's hiring brief, right? Someone still has to maintain all of that. And the system that makes that maintenance manageable has a much, much more durable job than one that's just a chatbot that answers questions. Now, do you see why I think the stakes go way, way higher than AI subscriptions? Gartner forecasts about $1.46 trillion in worldwide software spending this year alone.

And that isn't all up for replacement by any means. And it isn't all agent spending yet, but it gives you a sense of the established scale of business that these companies are about to influence. And the work that is performed through all of that software that you and I use is the work of business itself at this point. It's hiring. It's collecting revenue. It's managing purchases. It's serving and delivering value for customers. Those are activities that businesses have no choice but to fund. If you help a company to do them extraordinarily well, there's massive commercial upside there. And if you already sell the tools involved, there's a serious risk in letting another supplier become the one the customer relies on to assemble the job. That's the high ground that we're talking about. A position in how valuable work gets done with the ongoing relationship and purchasing influence that that that follows from that. So, at this point, I want to compliment the companies involved, which sounds silly, but stay with me. There really were two ways this could shake out, and I started with a partnership conversation between Microsoft and Salesforce for a reason.

It is way, way better for all of us who are customers if these companies realize that they do better when they provide us more choice. And that's exactly what that partnership highlights between Salesforce and Microsoft. It's actually the Open AI Oracle partnership is similar. We deserve to have choice. I've been pounding the drum and saying the future of AI is in giving companies and individuals choice, and these kinds of partnerships enable companies to focus on getting agents at work delivering value, which is by itself a huge job. It's not easy. And they would really rather not have a difficult conversation about having to choose between two companies that they both value. And so, I am hoping that even though this is a battle over the agenda control plane, and even though companies certainly have an incentive to try and own and drive that experience, and should be competing for it, I'm hoping we continue to see what I would call frenemy competition. Good partnerships formed where it makes sense, that allow companies to choose. Because if we can choose based on our existing unique stacks, where we want to invest, we're going to be in a much much stronger position to pick control planes and agentic intelligence management services that work for our business. And so, even

though I think this is often marketed as winner takes all or winner takes most, that's a way that dynamic could go. I don't think it has to go that way. Finance may need controls that marketing doesn't. A specialist may be better at a particular job themselves because they want to decide how the pieces fit together. This is such a big market that there are going to be room for multiple winners, and we're going to have to see, probably in the next 12 to 24 months, if the dynamic shift and a true agentic control plane winner does emerge. But either way, you don't need the entire system to tip over into a single winner dynamic for this to be consequential. If you have competition for the different roles in the business, you are going to need to vote with your dollars for as much interoperability as you can grab in order to make installation of multiple agentic systems easy.

And that sounds like a mouthful, but it's really important. It's the top of mind for a lot of the leaders I I talk to because they need to have flexibility in their agentic configurations in order to prepare for the future because the space is moving so quickly. And that's why these kinds of partnerships are so important. That's why the MuleSoft partnership makes so much sense. This is where I want to come to some of the takeaways from this video. I know it's easy to talk about the high ground. It may feel very abstract to spend a lot of time talking about what these big companies are doing, trying to own the agentic experience for the next 10 or 20 years. What does it mean for us? Well, you have a tremendous amount of power in the system with your sets of choices. And I'm going to talk specifically to the people on this channel. By definition, almost everybody on this channel is someone who is leaning in on AI. I love that for you. I love that for us. That means that at your respective companies, by and large, I've actually validated this by talking with so many of you. You guys are listened to. You guys are people who have influence around AI purchasing decisions.

Be intentional around the agentic control plane. Think carefully about how you put interoperability into your purchasing requirements when you build or buy one. Because you have got to get that right. You do not want to be in a position where you are boxed in when you look at how you control your agentic experiences going forward. And this is why most of this video has been spent on companies that are not the labs. I To be very honest with all of you, everyone that I talked to in companies wants flexibility. They want strategic optionality. They want the chance to have multiple solutions in play so that they are not locked in with any given one agent experience. And that's why, you know, I'm not sure how this is all going to play out, but I think solutions like a Microsoft-driven control plane, a Salesforce-driven control plane, other control planes that are set up by other third parties are going to be quite popular. They're going to be quite popular because they give customers what they want. They want choice.

And so when you're thinking about how to bet these out, ask yourself where does the value in the system come from? How easily interoperable are these agents? How many options do I have to swap them out? How does this tie into my agentic harness? How does this tie into my memory and data systems? How does this tie into the tools my employees use every day? Are my employees already living in Salesforce every day? These are the kinds of questions you start to ask, right? And the more you are intentional about that, the less you are likely to regret your recommendation for some sort of agentic control plane. And if you're thinking, "I'm not there yet." Good news, save this video and play it in 3 months. You're going to be there. Like everybody's going to get to this point real quick. It's not going to take that long. Now, if you are not in a place where you think you can influence a purchasing decision, you can still advocate for agentic solutions that work for your team. And what this looks like on an individual basis isn't as different as you might think. You're thinking about data access. You're thinking about interoperability with other agents. You're thinking about workflows. And sure, it's at the team scale, but it's very, very similar from a problem space perspective. And so, you also can be intentional. And what I see over and over again is that the people who are advocating for themselves, the people who are saying, "Hey procurement department, this is what I want to use.

This is why. This is standing in the way of me being productive." The ones who are squeaky wheels, they get the grease, right? They get the budget. They get the procurement department to say yes one way or another. Maybe not as fast as you want, but they get them to say yes. So, where does this leave all of us? What is the takeaway that we can walk away with? Number one, be intentional with what you're doing. Don't just by default go with the thing that you use every day. Think about what you're doing, how it drives the business long term, how you're driving impact. By the way, that's a cheat code for like doing well in your career these days. And then number two, think about where you want that control plane to sit. Who do you want to build that long-term commercial relationship with? And I know some companies are going to say, "We want to build it." Make sure you can resource that. That is not a small build. Make sure you know how to invest to get there. Make sure you have the talent to get that done and maintain it over time. It's not a small thing. Number three, make sure that you understand that the reason that this is happening is going to shape your own product experience going forward. The whole reason this conversation is happening is because agents are native to compute. Agents are going to mediate much of the compute we do at work and shape it. And the question on the line is who gets to manage that agentic layer for companies. That is the trillion-dollar question. And once you

understand that, if you are anywhere in the software and tech business, you ask yourself, "And how do I play in that landscape from a product perspective? What am I putting on the table? And how does that play into a world that is agent-mediated?" And we'll do another video or two on that one, but that's something that you should take and think about. Because even if you're not directly in a B2B play experience, maybe you're in B2C, maybe you are making physical products, agents driving computing experience is going to change everything. Anything with a screen, anything with a CPU, anything with a GPU, anywhere in your product experience, it's going to change it. And so you have to think now, how do you want to set yourself up so that you are delivering an extraordinary amount of value to the customer, and you are making it easier and easier for them to compute. We don't want to be in a position that we're in right now. I know so many of us are seeing these videos, feeling like you're not caught up. I feel like people ask me all the time, "How do you keep up, Nate?" And they have this like look in their eyes and like you we don't know when we last slept, etc. It's stressful right now. Part of what we need to get to is a steady state where agents computing can help us to manage the burden of complexity that we're all running into from a scale-up perspective. That's just a reality we're getting to.

We need to get there with an awareness that our choices now have a lot of leverage across that future. The agents we pick now matter. The control plane we pick now matters. How we think about product that matters, too. And so this trillion-dollar fight is not just a fight that we can sit back, get popcorn, and look at from a newspaper perspective. It affects all of us. And so, let me know in the comments how you see that fight playing out in your computing environment, how you see it playing out across the contracts you look at, how you think about it from a product perspective. I'd love to hear it. Cheers.