AgentExchange: Salesforce's Bet on the Agent Economy

Salesforce’s AgentExchange is more than a marketplace — it’s an architectural statement about who will control the agent economy. For PMs building on or competing with the Salesforce platform, understanding its mechanics, incentives, and strategic implications is quickly becoming essential.

  ·  7 min read

AgentExchange: Salesforce’s Bet on the Agent Economy #

T here is a moment in the lifecycle of every platform when it stops being a product and becomes an economy. For Salesforce, that moment arrived quietly in early 2025 with the launch of AgentExchange — a marketplace not for software, not for integrations, but for autonomous agents and the discrete skills that power them. It is a subtle but consequential distinction. When you browse AgentExchange, you are not shopping for a tool. You are hiring a digital worker.

That reframing changes everything: the packaging, the pricing, the trust model, the competitive dynamics. And for product managers — whether you’re building on the Salesforce platform, competing with it, or simply watching from the adjacent space of enterprise software — understanding what Salesforce is constructing here is quickly becoming essential context.

What AgentExchange Actually Is #

AgentExchange is the official marketplace for Agentforce, Salesforce’s AI agent platform. It launched in Spring 2025 and allows independent software vendors, consulting partners, and internal Salesforce teams to publish three types of artifacts:

Agents — fully pre-built autonomous workers scoped to a job-to-be-done: an SDR agent that qualifies inbound leads, a field service agent that handles scheduling and escalation, a compliance agent that monitors contract language. Buyers configure and deploy them within their Salesforce org without writing code.

Actions — discrete, reusable capabilities that an agent can invoke: querying a third-party data source, sending a structured Slack notification, triggering a downstream workflow in NetSuite. Think of these as the verbs in an agent’s vocabulary.

Skills — composable reasoning modules that give an agent domain-specific judgment: how to interpret a renewal risk score, how to triage a support case by sentiment, how to negotiate a scheduling conflict. Skills are the hardest to replicate and, over time, will likely be the most defensible unit of value on the exchange.

The marketplace runs natively inside Salesforce. Discovery, procurement, installation, and billing all happen within the same environment where the agent will eventually operate. That’s not a convenience feature — it’s a lock-in vector, and a deliberate one.

The PM’s Lens: Why Marketplaces Are Hard #

Building a two-sided marketplace is one of the most genuinely difficult product problems in enterprise software. The chicken-and-egg problem is real: sellers won’t invest in building listings without confident buyer demand, and buyers won’t trust a sparse catalog. Salesforce has an unusual structural advantage here — it enters the market with an enormous installed base of buyers already inside the platform. The demand side is pre-seeded. The challenge, as it almost always is, is supply quality.

Early marketplace listings tend to cluster in two failure modes. The first is commodity depth: twenty variations of the same SDR agent with superficially different configurations, competing purely on price, eroding margin for everyone. The second is trust debt: a few low-quality listings that produce bad outcomes erode confidence in the entire catalog, making buyers hesitant to try even the genuinely excellent offerings.

Salesforce’s response is a tiered review process with Salesforce Labs certification for top-tier listings, plus a ratings and review layer borrowed from the existing AppExchange playbook. Whether these mechanisms are sufficient to sustain quality at scale remains an open question — AppExchange itself has accumulated significant catalog debt over the years — but the institutional knowledge is there.

The Strategic Frame: Platform or Infrastructure? #

The more interesting product strategy question is what Salesforce is optimizing for with AgentExchange — and there are two coherent, competing answers.

The platform thesis says AgentExchange is primarily a retention and expansion mechanism. Agents deepen integration into a customer’s operational workflows in ways that static software cannot. An SDR agent that has been running for six months has accumulated conversation history, tuned its escalation thresholds, and been woven into the sales team’s daily rhythm. Ripping it out carries a switching cost that no SaaS contract renewal conversation can fully capture. On this reading, the marketplace is a moat-builder, and the economics of individual listings are secondary to the stickiness they generate for the core CRM.

The infrastructure thesis says Salesforce is positioning itself as the trust layer for enterprise agent procurement — the place where a Fortune 500 company’s procurement, security, and legal teams are willing to approve an autonomous agent deployment without a six-month vendor assessment. The AppExchange review process, for all its imperfections, established a pattern: ISV products cleared for enterprise deployment carry implicit Salesforce endorsement. AgentExchange extends that trust transfer to agents. On this reading, the marketplace becomes genuinely strategic infrastructure — the AWS Marketplace of agent distribution — and Salesforce’s revenue opportunity scales with the entire agent economy, not just its own platform.

These theses are not mutually exclusive, but they produce different product priorities. The platform thesis optimizes for depth of integration and switching cost. The infrastructure thesis optimizes for breadth of catalog, cross-cloud portability, and interoperability standards. Where Salesforce places its roadmap bets over the next eighteen months will reveal which thesis is actually driving the ship.

What It Means for Builders #

If you are a PM at an ISV considering whether to list on AgentExchange, the calculus is straightforward in outline and tricky in practice.

The distribution argument is compelling. Salesforce’s installed base means day-one access to a buyer pool that would take years to cultivate independently. The platform handles procurement, compliance review, and billing. For a seed-stage startup with a genuinely differentiated agent capability, that is an enormous acceleration of the go-to-market timeline.

The dependency risk is equally real. Listing on AgentExchange means accepting Salesforce’s data access policies, its review cadence, its pricing constraints, and its ability to build a competing native capability at any time. The history of enterprise software marketplaces is littered with ISVs who discovered, often too late, that their killer feature had become a standard platform capability. Agents are not immune to this dynamic — if anything, they are more exposed to it, because an agent’s value proposition is often tightly coupled to the platform’s own data model.

The strategic middle path is to treat AgentExchange as a distribution channel, not a product strategy. Build the differentiated skill or agent. List it for distribution. But maintain a product architecture that allows the same capability to run on other agentic platforms — Microsoft Copilot Studio, ServiceNow, or the emerging crop of standalone agent orchestration layers. Interoperability is optionality.

The Broader Signal #

mindmap
  root((AgentExchange))
    Marketplace Structure
      Agents
        Pre-built workers
        Job-scoped
        No-code deployment
      Actions
        Reusable capabilities
        Third-party integrations
        Workflow triggers
      Skills
        Domain reasoning
        Composable modules
        Highest defensibility
    Platform Dynamics
      Supply side
        ISV listings
        Salesforce Labs
        Partner ecosystem
      Demand side
        Installed base advantage
        Procurement trust layer
        Native billing
    Strategic Theses
      Platform thesis
        Retention mechanism
        Switching cost builder
        CRM moat deepener
      Infrastructure thesis
        Agent trust layer
        Cross-industry distribution
        Ecosystem revenue share
    PM Implications
      For builders
        Distribution upside
        Dependency risk
        Interoperability imperative
      For competitors
        Category definition risk
        Buyer expectation shift
        Catalog quality bar
AgentExchange strategic map

Zoom out far enough and AgentExchange is a signal about something larger than Salesforce’s product roadmap. The emergence of agent marketplaces — Salesforce’s, Microsoft’s, and the several venture-backed challengers building category-agnostic agent distribution layers — represents the beginning of a genuine labor market for digital workers.

That framing is not metaphorical. When a company procures an agent from a marketplace, they are making a decision that previously belonged to HR and operations: what capabilities should be automated, at what cost, with what level of oversight? The marketplace structure — catalog, ratings, pricing tiers, compliance review — is the mechanism by which those decisions get made at scale.

For product managers, this is the part that deserves the most attention. We have spent decades thinking about software adoption in terms of seats, licenses, and integrations. The agent economy asks us to think in terms of delegation: which tasks are we willing to hand to a digital worker, under what conditions, with what escalation paths? The marketplace that best answers that question — not just for the initial sale, but for the ongoing governance of deployed agents — will define the category.

Salesforce is betting it already has the relationships and the trust to be that marketplace. They may be right. Either way, the question they are forcing the industry to answer is the right one.


The most interesting thing about AgentExchange is not what it sells today. It’s the procurement behavior it’s training enterprise buyers to expect tomorrow.