The Next Step For AI Agents? Salesforce SVP Says They May Start Making Purchases

“A procurement team could say, ‘Hey, agent, as long as you’re within this budget, here’s my credit card. Go ahead,’” says Tyler Carlson, Salesforce’s senior vice president and head of product for AgentExchange and partner experience.

Procurement teams may soon feel comfortable enough to leverage specialized artificial intelligence agents not only to write and assess requests for proposal, but also to make purchases on humans’ behalf.

That’s a near possibility, Tyler Carlson, Salesforce’s senior vice president and head of product for AgentExchange and partner experience, told CRN in an interview.

“A procurement team might buy a specialized agent or they might set Claude or ChatGPT on a long-running goal task to write an RFP, go send that RFP out to a bunch of different vendors, collate all the responses, use the requirements to determine what the right scoring and best fit solution for a business need is,” Carlson said. “A procurement team could say, ‘Hey, agent, as long as you’re within this budget, here’s my credit card. Go ahead.’”

Growing AI tool usage, Salesforce’s decision to decouple its customer relationship management (CRM) software from its user interface to open up further flexibility for users and ongoing advancements in the San Francisco-based vendor’s AgentExchange marketplace are some of the biggest opportunities ahead for the vendor’s partner ecosystem, Carlson said.

[RELATED: Salesforce Dreamforce 2026: CEO Benioff Touts New AIforce Interface Layer]

Salesforce Invests In AgentExchange AI Discoverability

Sean Catlin, executive vice president and global head of strategy at Québec-based Salesforce solution provider OSF Digital, told CRN in an interview that the solution provider has built out a methodology to differentiate the company from the rest of the channel.

This methodology is aimed at delivering value in weeks and transforming clients into agentic enterprises in months.

Other solution providers can “get to use cases and pilots really, really quickly, but the scale doesn’t really happen afterwards” with other solution providers’ methodology. “We’ve developed a whole practice with defined roles. We’ve got specialized training and enablement that we’ve been scaling up our team globally. We’ve been using that with customers. It’s really exciting.”

Salesforce’s Carlson told CRN that some architectural changes made to AgentExchange under his watch to take fuller advantage of the AI opportunity include better answer engine optimization (AEO), generative engine optimization (GEO), including better allowing indexing by AI agents and agent access through browsers.

That “is also going to help all of our partners get more discoverable in those surfaces,” Carlson said. “So you can start asking questions like, ‘What are the best consultants for Salesforce in this industry?’ And it will be able to read our data and make recommendations and actually converse with the AgentExchange through Claude or ChatGPT or Google AI. So that’s cool.”

Here’s more of what Carlson had to say on how Salesforce’s innovation and strategic choices in the AI era are benefiting its partner ecosystem.

What’s your message to Salesforce solution providers coming out of Dreamforce 2026?

For the partners—particularly for the SI partners—what we’re doing with AIforce is like a whole new surface area for them to think about how they bring Salesforce solutions to their customers. … We’re getting a lot of interest from the community on how they can use the Headless Toolkit, the asset that we created, to build experiences like what we’ve done with Salesforce and Claude. … It’s effectively an SDK (software development kit) for building all kinds of new experiences with Salesforce.

We’re seeing this reinvigoration of the ecosystem in a way that I think mirrors like the early days of Force.com (the cloud-based platform-as-a-service (PaaS) for building and running custom applications on the same infrastructure as Salesforce CRM) and the platform, which is really exciting.

We’ve got a bunch of ISVs building all kinds of new experiences like Lovable, as an example, which you wouldn’t have thought traditionally of as an ISV of Salesforce.

But they’ve built an amazing experience where you can connect Lovable to the Salesforce HeadlessToolkit, our APIs (application programming interfaces) and MCP (Model Context Protocol) servers.

You can build your cockpit or your dashboard or your app on top of Salesforce data. And then you can take that and publish it back to Salesforce as the runtime for the React app that you built on the platform [for example].

Then all of the SI partners can go and do that kind of custom work and implementation for all their customers.

Are you seeing a convergence in partner business models—services partners developing software, vendors adding services?

We are seeing a lot of SIs that are building more package solutions, durable IP (intellectual property) that they can bring to our marketplace.

We did a big uplift for consulting partners on the AgentExchange recently. Reinvigorated the listings, brought the new partner program and competencies and all of that. And we created effectively what is a partner container. [They can list] their consulting services, but also if they want to list ISV applications, they can all do that under one partner container.

We are starting to see some of our SIs go in that direction—build out accelerators, packaged IP, AgentExchange apps. Deloitte is one that did that with us (for example).

What are the benefits of Salesforce solution providers investing in their AgentExchange presence?

The opportunity now with what we’re doing with headless and the opportunity for what the partners are building on the AgentExchange, whether ISV or SI, to drive more usage of the platform.

[That] is creating a flywheel for the partner and the reps (Salesforce sales representatives) to care about bringing in the right partner for their customer, bringing in the right ISV solutions for their customers, so that it’ll drive additional consumption and usage and value for the customer. … Our aspiration with the AgentExchange is … to build a similar marketplace for a SaaS (software-as-a-service) product like the hyperscalers have.

And as customers pre-buy usage, commit to usage, there becomes this mechanism where our partners get access to this pre-committed budget. Purchases through the marketplace can help them retire that usage and precommitted budget. … We launched AgentExchange commerce with custom private offers for partners to sell through.

In October … we will also allow partners to sell services and their core products that aren’t necessarily running on Salesforce through that marketplace surface as well.

How should Salesforce solution providers think about this theme coming out of Dreamforce that the platform’s value is beyond the user interface?

We have this opportunity now to think of these AI tools as a completely new interface experience for Salesforce and lots of other applications.

What we’ve built with the Headless Toolkit is basically that set of primitives in the platform that enable you to bring all of those experiences to life across any number of new interfaces, starting with Claude, Slack and Coworker. … We are going to be adding many, many more.

The interesting part about that for our partner ecosystem, whether you’re an SI or an ISV, you now have this platform foundation to build incredible new experiences.

And if you look back maybe two years when all of this large language model stuff started, we were thinking about prompt engineering and prompts almost starting to become a little bit of that new interface layer.

The problem with that is that everybody can prompt the LLM in their own way. And you get this sort of proliferation of different people asking different questions. And in the enterprise, that can be a challenge because when you’re running a process and I’m running a process, we want that process to have relatively deterministic outcome and results.

That’s where a lot of what we’re building into the Headless Toolkit and AIforce layer is this set of opinionated pre-built skills and plugins, which are effectively a wrapper around our MCP server with a set of pre-built skills that give everyone what effectively would be the right prompting for the LLM to understand how to do a task.

So whether that’s generating a brief for an account or helping you diagnose your sales pipeline with those pre-packaged skills, we’re giving you—this is the right way to ask the LLM to do this task on your behalf on top of our platform.

For our partner ecosystem, you can think about, we’ve just built [37 Claude skills into Salesforce] so far. We’re going to be building a whole bunch more. But the ISVs and SIs can take this layer that we’ve built and completely explode the possibilities. Verticalized use cases, industry use cases across all different kinds of things, building out things for their own apps and surfaces.

For the consultant partners, you can think about taking all of the implementation methodology that they have and packaging that up into skills to say, ‘How do you configure and use Salesforce better as a platform?’

There are a ton of really cool stuff that I think we’re hoping to see the ecosystem build on top of what we just announced.

Are we seeing Salesforce enter a new category with this technology stack?

If you look at that stack that we announced—AIforce at the top, then you have the agent layer, the context, the semantics, the apps and then the data layer—when you have all of your business processes defined and your permissions, models and all of the different business logic and workflow that you’ve generated as a business in one platform like Salesforce, it becomes business process infrastructure for the AI.

You can now point these AI tools at that layer. And because we’ve built this platform with metadata at the core and semantic understanding of your business, it’s almost a superpower for AI to understand your business, to understand what the rules are, to understand how you work in your enterprise. And that allows it to actually be a helpful operator on your business.

I think of this transition that we’re going through as Salesforce moving from the user interface layer that you interact with to your application and business process infrastructure that can feed any type of experience, whether that’s an experience that we’ve built, like Slack or the Lightning experience with Coworker, or Claude or any number of other AI tools.

What should solution providers know about the evolving pricing models for Salesforce’s AI products?

In general, the market hasn’t [decided] on one thing. … We have customers that want to buy in certain ways and others in other ways.

Our focus is really about trying to give as much flexibility and choice in the way you consume. So we have models where we’re bundling the AI capability into the user license.

We have models where you can pre-buy a bunch of usage credits. And then we also have models where you can pay-as-you-go for the usage credits.

We are seeing a lot of different customers pick and choose and blend those things together. You’ll start to probably see our customers create an amalgamation of those things—some licenses up front, some amount of committed spend on usage that they have direct line of sight to, and then maybe a fungible bucket of, ‘I think I can spend this money with you, and I want to commit to it in order to unlock a discount, but I’m not exactly sure yet what I’m going to spend that money on.’

That flexibility is really important for our customers. … I talk to all of our ISVs and they’re looking to us a lot of the time for guidance on pricing.

We’ve done a lot of experimentation with pricing over the last two years so that we can be the bleeding edge, so that they can see what works. We help guide them, and then they attach to our model and follow us into what’s next.

Sounds like maybe Salesforce customers can have more control over pricing while using bleeding-edge AI models?

If you don’t really know where the future is going, having the fungibility to swap between a license and a credit that is usable across lots of different surfaces is potentially useful for the customer in that context.

It is a little bit of a hedge.

What else is in store for AgentExchange?

The main thing I’m focused on with our AgentExchange marketplace is getting the commerce services up and running.

Next year, we’re going to be working on public checkout. We have private offer, custom pricing available now. Launching our external product support and external service support for that.

The really interesting surface area that unlocks for partners is, as we start thinking about them building capabilities that drive usage or consumption of Salesforce platform, how that potentially retires commitments that our customers are making so that we get that commerce synergy across our partner ecosystem and our direct sellers.

There are also some interesting things that unlocks with selling first- and third-party products together. So you can imagine a partner of ours that builds an agent that runs on our agent runtime, and the customer needs credits from Salesforce in order to operate it.

But there’s some IP that the partner has built, and they want to put all of that together on one quote to the customer with one bill, one invoice. That’s something we’re really focused on trying to figure out how to unlock for our ecosystem.

We are enabling our partners to sell services through the marketplace. So now, with that addition to the platform, you’ll be able to put together a service engagement.

It’s like an accelerator with some IP plus a professional services package. And they sell that to customers through our marketplace, which means they get the bill and the invoice from Salesforce, which also helps with procurement friction, getting in the door—especially if you’re a smaller consultant. Leveraging the Salesforce relationship with the customer and their awareness of us and our marketplace helps them land a customer they maybe wouldn’t have landed.

We’ve got a long roadmap of functionality to deliver on the AgentExchange. I want to make some improvements to how our search works. I want to make some improvements to how we surface our listings into the AI tools for how they discover things. I want to build our checkout infrastructure as MCP servers so that we can bundle that into other surfaces and let agents buy stuff off of our marketplace.

How are AI agents changing marketplace transactions?

We had to do a full revamp of the way that we rendered our listings on the marketplace because at some point in the past, before I took over this product, we started doing client-side rendering, which actually meant that the agents are basically getting a stub with the listing.

So they were not indexing it. They were not bringing it into their training set. They were not able to access it directly through the browser.

And so we have completely revamped the way we do that, moved everything to a more performance server-side rendering thing.

So now we’re actually serving all of that content up in a way that’s optimized for AEO, (answer engine optimization), GEO (generative engine optimization).

Which is also going to help all of our partners get more discoverable in those surfaces. So you can start asking questions like, ‘What are the best consultants for Salesforce in this industry?’

And it will be able to read our data and make recommendations and actually converse with the AgentExchange through Claude or ChatGPT or Google AI. So that’s cool.

Do you see agents doing transactions soon?

I don’t think it is a stretch to think that, to start, a procurement team might buy a specialized agent or they might set Claude or ChatGPT on a long-running goal task to write an RFP (request for proposal), go send that RFP out to a bunch of different vendors, collate all the responses, use the requirements to determine what the right scoring and best fit solution for a business need is. … A procurement team could say, ‘Hey, agent, as long as you’re within this budget, here’s my credit card. Go ahead.’

It wouldn’t be a stretch to think that you would do that kind of a task on an RFP. Maybe [instead of checkout], it comes back to the procurement professional and says, ‘OK, I found the right solution. Go buy it. Check out with a credit card for three or four licenses for something to try it out and see how it goes.’