Salesforce is buying Fin, formerly Intercom: what CRM consolidation means for your site agent

Published July 31, 2026 · Last reviewed August 6, 2026

On 15 June 2026, Salesforce signed a definitive agreement to acquire Fin, the AI customer agent company most people still know as Intercom, for approximately $3.6 billion.¹ It is one of the larger agent acquisitions of the year, and the trade press covered it the way it covers most deals: valuation, strategic fit, a line about the founders. That coverage answers a shareholder’s question. It does not answer the one a person running an agent on their own website should be asking.

This piece is that second read. Not “was it a good deal for Salesforce,” but “what does agent consolidation mean for me, the buyer, if my support or sales agent lives on a vendor that could be next?” The honest answer has two sides, and the useful part is separating them: what consolidation genuinely improves, and the one thing it quietly puts a question mark over.

The short version, so the rest earns its space: buying into a big suite gets you deeper integration and a cleaner price. The thing to check before you lean in is whether your agent stays reachable by agents that do not live in the same suite, because that reach is the part an acquisition can change.

What Salesforce actually bought

Start with the facts, held neutral. Fin is the current name of the company founded in Dublin in 2011 as Intercom; it rebranded to Fin in May 2026, taking the name of its flagship AI agent.² So Fin and Intercom are the same company, and the product Salesforce is acquiring is the agent that resolves customer queries across live chat, email, WhatsApp, SMS, phone, and Slack. Salesforce says that agent closes out roughly 76% of incoming support requests with no human stepping in.¹ The scale of what changes hands is worth naming too: the acquisition brings more than 30,000 business customers into Agentforce.³

The stated plan is to fold Fin’s technology and team into Agentforce, Salesforce’s own agent platform, with the two positioned as complements rather than duplicates. Agentforce is aimed at large organisations that want deep customisation and will invest the configuration time; Fin arrives pre-trained and quick to deploy, which makes it the more accessible option for smaller and mid-market teams.⁴ The transaction is expected to close in Salesforce’s fiscal Q4 2027, which runs roughly November 2026 to January 2027, after regulatory approvals, and Salesforce noted it does not change its financial guidance.¹

There is also a pricing signal worth naming, because it is part of what buyers get. Agentforce’s Help Agent can be billed at about $2 per autonomous resolution, roughly 400 Flex Credits, charged only when the agent settles an issue with no human involved, no escalation, and no sign the customer left unhappy. Miss any of those and the resolution is free.⁵ For a support team, paying only for clean outcomes is a real and attractive change. Consolidation is not just a logo swap; it comes with integration depth and an economic model that a standalone tool often cannot match.

The consolidation pattern, read honestly

This is the second time this year a well-funded platform has absorbed a leading agent product; the first was Meta’s purchase of the agent-social network Moltbook, which we looked at through an identity lens in a separate piece. Two data points is not a trend, but the shape rhymes. Capable agent products are becoming features inside larger suites, and the larger suite is usually a place where a lot of business software already lives.

Read from the buyer’s chair, that shape has a clear upside. A support agent that sits natively inside your CRM sees your customer records, your case history, and your workflows without a fragile integration layer in between. The data the agent needs is already in the room. Add pay-per-resolution economics on top and the pitch is coherent: better context, better outcomes, and you pay when it works. None of that is spin, and pretending otherwise would be the dishonest read.

The cost is not hidden either, and it is not a scare word. It is a scope question. When your agent becomes part of one vendor’s suite, its centre of gravity moves toward that vendor’s world: its data model, its channels, its other agents. That is exactly what makes the integration good. It is also what narrows the set of things outside that world your agent naturally talks to. Whether that narrowing matters depends entirely on who you need your agent to reach, which is the next question.

The buyer’s real question is reach, not features

Feature comparisons are where most build-versus-buy analysis lives, and for a support agent they are largely settled: the big suites are good, and getting better. So park features and ask the question they do not cover. When another party’s agent, one representing a buyer, a partner, or a job candidate, comes looking for you, can it reach your agent and get a real answer?

An agent that lives inside a CRM suite is built to be excellent at two conversations: with human visitors on your site, and with the systems inside that same suite. Both matter. Neither is the same as being addressable by agents on other stacks. This is the distinction this blog keeps drawing between a site that is agent-readable and one that is agent-addressable: exposing content or tools a machine can parse is not the same as having a representative other agents can find, query, and route to across the open network.⁶ A suite-bound support agent can be superb at the first job and still be hard to reach for the second.

That gap is easy to miss because today most inbound is still human. It gets harder to miss as more of the professional web is mediated by agents doing errands for their people, the shift behind why a buyer’s agent will not wait on a form. The moment a meaningful share of your inbound is other agents, “can they address mine” stops being theoretical. A consolidation is a good prompt to ask it early, while the answer is still cheap to change.

What stays portable across any acquisition

Here is the structural point, and it is the reassuring one. You do not have to bet against consolidation to protect yourself from it. You just have to keep two layers out of any single suite: your agent’s identity and its reachability.

The conversation engine is fungible. Whether Fin, Agentforce, or the next product runs the actual back-and-forth is a decision you can revisit, and buying it from a big vendor is often the right call. What should not be fungible is who your agent represents and how it is found. If the identity of the human or company behind the agent lives on an open, readable standard, and if discovery of that agent is open rather than internal to one product, then an acquisition changes the engine without changing your reach. The agent that answers can be swapped; the address other agents use to find you stays put.

This is the layer Tobira works on, and it is complementary to a CRM agent, not a competitor to it. Tobira is the trust layer for the agentic web: a readable @handle for the person or company an agent represents, a credibility signal built from real conversation history on a 0-5 scale shown as four plain levels, and mutual reveal, where contact details change hands only after both sides agree. As of the June 2026 founder update, the Tobira network listed 648 public agents, including 102 business agents.⁷ A Salesforce-hosted agent, a Fin agent, or one you build yourself can each carry that same handle and be addressable on the same open network. The point is not to replace the engine you buy; it is to keep the address you own.

A short checklist for buyers weighing the consolidation

If you are choosing or renewing a site agent while the market consolidates, a handful of questions keep the decision clear.

• Separate the two purchases in your head. The conversation engine and the agent’s identity and discovery are different layers; you can buy the first from a big suite and still keep the second on open standards.
• Ask the reach question directly. Can an agent that does not use your vendor’s suite find and address your agent, or only human visitors and suite-native systems?
• Look at where identity lives. Is the human or company behind your agent represented on a readable, portable standard, or only inside the vendor’s account model?
• Price the good part honestly. Pay-per-resolution and native CRM context are real gains; a portability hedge does not require giving them up.
• Treat an acquisition as a review trigger, not an alarm. It is a normal moment to confirm your reach and identity are not things a future deal can quietly move.

None of this argues against buying from Salesforce, Fin, or anyone else. It argues for knowing which layer you are buying, and keeping the one that has to outlast the vendor in your own hands.

What to remember

• Salesforce signed a definitive agreement on 15 June 2026 to acquire Fin, formerly Intercom, for about $3.6 billion, folding it into Agentforce; the deal is expected to close in Salesforce’s fiscal Q4 2027 (roughly November 2026 to January 2027).
• Consolidation brings genuine buyer gains: native CRM context and a pay-per-resolution model that charges only for clean, human-free outcomes.
• The trade-off is not lock-in as a slogan; it is a scope shift. Your agent’s centre of gravity moves toward one vendor’s world, which is what makes integration good and reach narrower.
• For a site agent, the durable question is reach, not features: can agents on other stacks address yours, or only human visitors and suite-native systems?
• Two layers stay portable across any acquisition if you keep them on open standards: the identity of who the agent represents, and how the agent is discovered and reached.
• A readable @handle and open discovery are complementary to a CRM agent, not a replacement. Buy the engine from whom you like; keep the address you own.

FAQ

What did Salesforce buy, and for how much? On 15 June 2026 Salesforce signed a definitive agreement to acquire Fin, the AI customer agent company formerly known as Intercom, for approximately $3.6 billion, subject to customary adjustments. Fin’s technology and team are set to power Salesforce’s Agentforce. The deal is expected to close in Salesforce’s fiscal Q4 2027, which runs roughly November 2026 to January 2027, after regulatory approvals.

Why did Intercom become Fin? Intercom, founded in Dublin in 2011, renamed itself Fin in May 2026, taking the name of its flagship AI agent. So Fin and Intercom refer to the same company; Fin is the current name and the product Salesforce is acquiring.

What does the acquisition mean for my own site agent? If you run an agent on your website through one of these vendors, consolidation tends to bring deeper CRM-native integration and a clearer economic model, such as Agentforce’s pay-per-resolution billing. The trade-off to weigh is reach: an agent that lives inside one vendor suite is good at talking to human visitors and to that suite, but whether agents on other stacks can address it is a separate question worth asking before you commit.

How does Agentforce pay-per-resolution pricing work? Agentforce’s Help Agent can be billed at about $2 per autonomous resolution, roughly 400 Flex Credits, charged only when the agent resolves an issue with no human involved, no escalation request, and no sign the customer was unhappy. If any of those conditions is missed, that resolution is not charged. Flex Credits, priced around $0.10 per standard action, are the alternative model for higher-volume, simpler tasks.

What keeps a site agent portable across an acquisition? The conversation engine can come from any vendor. What stays portable is the agent’s identity and its reachability when both live on open standards: a readable handle for the human or company behind the agent, and open discovery so other agents can find and address it. Those layers do not change when the product running the conversation is acquired, which is why keeping them outside any single suite is the hedge.

Sources