Agentforce, Bespoke Pricing, and the ‘Death of Pay-Per-Seat’: Our Talk With Manny Medina

Of all the top voices in the Salesforce ecosystem, perhaps few are more qualified to talk about AI agent pricing than the former CEO of sales execution platform Outreach, Manny Medina. 

Manny, who is now running AI agent Revenue Engine company Paid, spoke to Salesforce Ben about AI ‘hype’, bespoke pricing, Salesforce’s newest changes, and much more.  

Outcome-Based Pricing Is “Inevitable”

We asked Manny about Salesforce’s recent pivot away from a pay-per-seat model, and more towards a consumption-based pricing model.

Salesforce launched Agentforce in October last year, offering the AI suite at a rate of $2 per conversation.

Agentforce has seen several tweaks and upgrades in the following months, with Agentforce ‘2.0’ and ‘2dx’ being later released. 

The most recent news, however, is the launch of a revamped pricing model, introducing a new Flex Credit system, with ‘actions’ now costing $0.10, and packs of 100,000 credits available for $500. Customers are able to allocate AI spend through the Salesforce Digital Wallet.

Manny, whose new company Paid handles AI agent pricing, subscriptions, margins, billing, and renewals, told Salesforce Ben that there was a trend of moving towards outcome-based pricing, which is “inevitable” for two reasons. 

He said: “One, it’s very easy to sell. If you go to a customer and you tell them, you only pay when you get results, it sounds like an IQ test. How do you say no to that?

“The second part is that it creates a business model differentiation that sets you apart from the competition. So from a competitive dynamic environment, you are preventing your competitors from coming in and saying, ‘Look, I can do the same but cheaper’, because you don’t know what the same is.”

Manny added that there is an “inevitability” in terms of how agents will come to be priced.

He said: “Because if you think about what the agent does, the agent is taking up whole human tasks and executing them on their behalf. So whether it’s taking over the entirety of the job or taking over a good portion of the job, they are in the business of doing what the human was doing before, freeing up time for the rep to do something more valuable and, you know, more human. You would be more relationship-driven.

“The fact that it moved from seat to consumption was kind of expected. What was not expected was expanding the definition of consumption to outcomes. That’s a part that I think caught us all by surprise, including me, because I’m in the business of selling the future of monetization for agents.

“Most of it is outcome or workflow-based. When I reach out to Salesforce about this, they’re like, ‘Yeah, we’ll never do it. ’ So to see that happen on the other end is quite astonishing.”

We asked Manny why it has been so hard for some companies to get pricing right and what the correct approach to this would be.

The Paid CEO said that the correct approach would be for the agent creators to understand the business problem that they are solving, helping them understand what exactly is valuable to their customers. 

This means they can then start putting a price around what is valuable to the customer, Manny said.

He added: “If you look at the other companies that have been ahead of the game, such as Sierra, Intercom, Zendesk, or even recently Decagon, they’re all in the business of customer service and customer support. They have also come up with some kind of outcome-based pricing. 

“If you look at what outcomes are, they’re either tightly defined, in which it makes it easy for you to wrap your head around it, or they’re bespoke, such as Sierra, in which for each customer, there is a set of outcomes that they care about, and you can instrument and then measure and charge against them.

“So I feel like this trend of going to outcome-based pricing is inevitable.”

Unrealistic Expectations Around AI? 

Opinions around artificial intelligence are often quite divided, with evangelists claiming the technology will change just about everything – and for the better, no doubt – while skeptics often point to the potentially devastating effects AI could have on the labor market and human civilization generally. 

These are both extreme visions, but there is a third possibility – that the technology is, perhaps, simply being overhyped, at least in its current form.

We asked Manny whether he thought AI agents in particular really are as transformative as some people are saying, or if there are unrealistic expectations being created about their potential. 

He said: “I think that they are more transformative than people expect. So I don’t think that even our expectations are realistic – but on the low side, not on the high side. Everything changed with OpenAI’s strawberry release. I think it was renamed 04, when reasoning got involved in what LLMs can do. 

“Now you have the birth of fully agentic workloads and workflows to be addressed by agents themselves, and it will start with the most common, most pedestrian, least amount of thinking involved, such as scheduling an appointment, finding leads, or sorting out your ICP… resolving tickets.

“If you think about a ticket, you look up the resolution and you inform the user what it is. But I think it’s going to evolve into something far more complex and far more human-like than what we have right now. 

“And as it does, we should expect interactions between the agent and the customer to go up – that consumes a lot of credits, a lot of tokens. So the vendors will have to come up with pricing scenarios and monetization scenarios where they can cover their margins, but also deliver the expected result and the expected experience to the customer.”

Death of Pay-Per-Seat?

While there is a new flexible pricing model for Agentforce, there is still a pay-per-usage element nonetheless. 

We asked Manny about his thoughts on whether the tech sector is going this way generally.  

He said: “What has changed between the SaaS world and the agentic world is that all the information that you need around the work that you’re delivering lives in the code. There is no human involved to deliver the work; it’s the agent doing the work.

“If all you do is observe what the agent is doing, you can then say the agent is delivering this amount of ROI, the agent has performed these jobs, the agent will cost you this much, and the margin on the work performance is X, Y and Z. You can take those components and then put it on your application to show your customer at all times an invoice and the bill and the application itself, all the value that they’re getting from the customer consistently. So that’s what we are.”

He added that a model that allows customers to move credits between different agents and between seeds makes a “ton of sense”, so you can align your consumption of different parts of the cloud to what you paid for. 

We asked Manny his thoughts on whether there was any sense in having a one-size-fits-all approach to AI agent pricing, or if the answer to the question of pricing would always be some variety of ‘it depends’. 

He told Salesforce Ben: “I think that the answer doesn’t need to be ‘it depends’. As we all know, Salesforce is one of the highest-performing sales organizations on the planet. If you are a good seller at Salesforce and somebody asks you for the price, you know that a good seller would say, ‘Well, let’s take a look at what’s important to you before we price.’”

Manny said that he thinks that current pricing approaches are “artifacts of the past”. 

He added: “Our real differentiation is going to be in the way that we engage with our customers, in our story, and in the way we price and package and go to market with that situation. If we’re smart about it, it will be a huge boat because we will get into our customers’ business so deeply that even our pricing is aligned with it, and nobody else will look the same.

“So will Salesforce do it at some point? I don’t know. It’s a big ship. But what I’m seeing is that the smaller players, like Sierra, for instance, are doing this several times a day. Like, this is not even a question.”

Final Thoughts 

Manny Medina spoke with authority on the topic of agentic AI pricing, offering us some fascinating insights and predictions into the future of the sector. 

Salesforce’s apparent recent pivot away from a pay-per-seat model to consumption-based pricing and flexi-credits with Agentforce could well be a blip, but it also may be an indication of broader changes to come in the SaaS world and tech sector generally. Our full interview Manny will be published soon, so stay tuned!

What are your thoughts on Manny’s predictions? Leave a comment in the section below.