Dreamforce 25 - welcome to Agentforce! How would you like to settle your bill?

One of the recurring questions I’ve had throughout this year’s Dreamforce has been how Salesforce plans to monetize the agentic revolution it’s proposing? At the end of the day this is a business and Salesforce is not in the agents game for the good of its health - value has to be created for shareholders and Agentforce is going to have to make its contribution to the bottom line.

Going into the conference this week there was a presumed wisdom based on public domain statements about the traction Agentforce has had since its launch at the 2024 Dreamforce just over a year ago - 12,000-12,500 deployments, at varying stages of maturity and scale, with some 6,000 of those being paying engagements.

That’s impressive growth from a standing start a year ago when Agentforce was first went on generally availability, but it does mean that the majority of users are yet to cut a check for their agentic dabbling, although Chief Operating and Financial Officer Robin Washington talks about a 400% increase in Annual Recurring Revenue (ARR) from Agentforce, around $440 million.

Now, a ‘try before you buy’ policy is fair enough with such a radical pivot around a new tech like agentic AI. A repeated cri de coeur from Salesforce CEO Marc Benioff has been for enterprises to start experimenting with agents and their potential, ASAP, so making it free for them get underway with pilots is entirely sound thinking, strategically and tactically.

But it does beg the question of how this is managed in practical terms? At what point does Salesforce look at an ‘experiment’ and determine that the user in question has now got themselves a working agentic system that is plugged into key workflows and contributing to the revenue generation of the business? And when that moment comes, what happens next? Is the customer presented with a bill to start paying from here on in?

Opening the kimono

Earlier yesterday I asked that of Washington who talked about how Salesforce has been looking into new pricing mechanisms and models, but for more detail on this it was left to Chief Revenue Officer Miguel Milano to ‘open the kimono’ on how customers should expect to be funding the agentic transition moving forward.

Milano is one of the so-called ‘boomerang’ breed of Salesforce management - individuals who were with the company before, in Milano’s case from its early days, left to pursue new career moves, but have been wooed back by Benioff into top slots in the corporate hierarchy. So he’s someone who’s seen a lot of change in Salesforce over the years and has watched its operating models morph and evolve over time. In common with every other Salesforce exec this week he’s quick to argue that the agentic revolution is one of the biggest he’s encountered:

This is unprecedented. I’ve been 25-plus years in Sales. I haven't seen anything like this coming at me ever in the last three decades...This reminds me a little bit [of] the 2010 to 2020 decade, where I joined the company in 2011. It was kind of lucky moment because at that point every company wanted to go to the cloud. They wanted to go to use CRM applications in the cloud. And we were the lucky recipient at that time because we were the best platform.

Flash forward to 2025 and now Salesforce is on the cusp of “the next ‘golden decade’”, he suggests:

What is happening is that every customer wants to become an Agentic Enterprise. Why? Because they want to go faster; they want to grow top line; hey want to drive productivity; they want to reduce cost; they want to increase the NPS (Net Promotor Score), the customer success; they want to make the lives of the customers better, more available, more pro-active; and then they want to empower the employees. They know that AI is going to enable that. 

But what they - and the rest of us - want to know is how all this is going to be paid for, Milano concedes:

We do demos. That's the easy part. And then fortunately or unfortunately - I think it's fortunately because it gives customer confidence - we do pilots. Most of the time we do pilots because [users] need to prove that the agent can operate, can work, can engage with customers. And then once they feel that they're ready to go and deploy not just one agent, but many agents, they need the right commercial construct. 

But: 

One thing we've learned is that customers, they don't really understand. There's a lot of unknown....’How much am I going to consume, how much I'm going to pay you?’.

Options

Worry not, says Milano, there are multiple ways to contract commercially with Salesforce and it starts with “meeting customers where they are” - in other words, trying to deal with them on their own terms and meeting their commercial needs:

There are customers that say, ‘Listen, I don't want to get confused with consumption [models], I want seat-based licenses. Well, we created seat-based editions, like AgentForce One Edition or Agentforce for Sales or Agentforce for Service. That essentially is, if you have already 10,000 users of Service Cloud, I'm going to give you the superpower of AI and agentic for those 10,000 users. So we upgrade you to Agentforce for Service and now you have a limited access to Agentforce for employee-facing agents. That's pretty cool.

Another approach is the consumption model, although this is, as Milano notes, not to everyone’s taste:

There are customers that want to go into the consumption world, but they just want to buy fuel credits. They are a bit more scared. They still don't know [whether to] pay-as-you-go or do a pre-commitment like you do with hyperscalers, [where] you commit certain amounts, but you only pay when you consume the revenue. We have that [option].

And then there are new pricing models that are emerging that are tailored to the new agentic realities, such as flex agreements. These came out of a starting point of assuming that if work done by humans is shifted over to an agent, organizations may end up with fewer humans doing certain jobs. Over the past year, thinking around this has moved on as Milano explains:

A year ago, we didn't know that really humans and agents need to be working together, and  that there's many more things that humans are going to be doing. But we wanted to build agreements, flex agreements, where if a customer believes that by shifting to agents, they're going to have less humans, they don't have to pay for the Service Cloud licenses or the Sales Cloud licenses. We give them the option in the contract. We call them flex agreements where they can use the investment in the seat-based licenses to fuel more consumption.

Actually the presumed rationalization of seats hasn’t been seen in practice, he adds:

We have very few customers asking for that. And for those that are asking for that, the seats are not going away. But you know what, it gives a sense of reassurance to the customers that they can have the flexibility. 

ELA thinking

But the big development as far as Milano is concerned is the introduction of ELAs - agentic Enterprise License Agreements. These were devised by Milano and Benioff during the summer when the CEO was residing for a month in Europe and meeting with CEO customers in the region. What the two men found in such meetings influenced their thinking about what was needed from a licensing point of view, says Milano:

I think customers are already past the phase, we call it the technology phase, where they were kicking tires and experimenting. And CEOs for the most part, they're tired. They're saying, ‘I just want to transform. I just want to use AI everywhere. We are clear that there are very few technology vendors that can do this for us. We want Salesforce to do it, but we are worried about the pricing’. After many meetings, we realized [there was] uncertainty of pricing. Predictability of cost was very important for CEOs. So we put together something very simple. 

That simplicity manifests itself thus - flat fee, unlimited usage of Data Cloud and Agentforce for customers:

They can deploy any use case they want for two, three years.They can ingest as much data as they need for those agentic use cases. Of course, we have some wording so that they don't go crazy with data ingestion for any other thing.

Such unlimited consumption agreements alter how Salesforce monetizes customers, argues Milano:

The reality is it also comes with a step change on how they use our software. They now are using our software as a Digital Labor platform. They're going to deploy many, many use cases. It has to do with CRM, but it doesn't have to do with CRM. I mean, we are identifying all our CRM apps, but there are so many other use cases that are adjacent to CRM that agents can do. So we're giving them the construct, the predictability, for them to go big with Agentforce and Data Cloud.

More, more, more

OK, so that’s the theory and that’s what came out of summer in Europe. How’s it working out in practice? As with all things agentic, it’s early days still, but Milano is pleased with the progress he’s seeing to date and has big expectations for what’s to come:

We put this together mid-July. We've already signed a dozen of them. {As of] Friday last week, we had 150 being negotiated right now in the pipeline. After Marc announced [ELAs] in his [Dreamforce] keynote on Monday, this number is probably going to be doubling very soon. All of them for this [second half of the year], a lot of them in October, many of them in Q4.

As to what’s in it for Salesforce and the bottom line, Milano is looking to a 10% uptick in Average Order Value (AOV):

This is a step change, in some cases a multiplier impact in our monetary relationship with our customers. I'm always saying, the only reason customers pay us 50% more, 200% more, 300% more, is because they're using us differently. They're using a different budget pool, which is the Digital Labor. They are driving tremendous value. 

Bottom line - how’s all this going to impact on the bottom line, as it were? At present, 40% of Agentforce and Data Cloud revenue comes from existing customers “re-filling the tank”, according to Milano, but things aren’t that simple any more:

What you probably didn't calculate in the math is we started the year only with a few thousand, 5,000, less than 5,000, paying customers of Agentforce plus Data Cloud...Now fast forward 12 months, at the end of [a] year, we have now more than 10,000 paying Agentforce and Data Cloud customers. We're going to finish [this year] with more than 20,000 paying customers, Data Cloud and Agentforce. So imagine next year, all these customers are coming to us and they're going to re-fill the tank. That's very, very exciting.

And all of this is having a beneficial impact across the wider Salesforce portfolio, he points out:

This is a multiplier effect. In most cases, the more agents [customers] use, the more they realize that they need our core apps to execute. I have customers that didn't have Sales Cloud or Service Cloud and because they have agents now, they are buying the core apps. So not only there is a consumption flywheel effect, but there is a product flywheel from core products to agentic products to more core products to more agentic products.

My take

The consumption flywheel is kicking in and it's becoming a bigger and bigger component of our total bookings - and next year is going to be even bigger. 

One of the promises of agentic AI for end user is not only greater efficiency and cost-containment, but, more importantly perhaps, a new revenue growth enabler. That’s also the case for Salesforce, according to Milano who declares:

Welcome to the next chapter of SaaS.

Now, a lot of this thinking and policy has changed and evolved over the course of the past year as the realities of the agentic thesis pitched by Benioff at last year’s Dreamforce were put to the test out in the wild. We may find that in another 12 months things have changed again as customer preferences and needs mature as Agentforce adoptions increase and enterprise needs become clearer.

But for now, Milano has provided a useful update and an assurance that Salesforce is thinking hard about how to monetize all this at a time when the short termists on Wall Street are increasingly demanding that tech firms “show me the money” when it comes to AI and getting shirty if they don’t like what they see. There are viable pricing models to make the agentic shift a sound commercial transaction for both buyer and seller - and almost certainly more to come.

Onwards!

Check out diginomica's dedicated Dreamforce event hub here.