Salesforce buying agentic AI firm Fin for $3.6 billion

When Salesforce announced on Monday its plan to buy agentic AI firm Fin, which, until a few weeks ago, was known as Intercom, it said that the company “resolves complex customer queries end-to-end, across every channel, including live chat, email, WhatsApp, SMS, phone, and Slack.” But analysts and consultants questioned whether the acquisition, one of more than a dozen Salesforce AI acquisitions since the beginning of last year, was going to end up accelerating the Salesforce functionality timeline or merely confuse matters. Scott Bickley , advisory fellow at Info-Tech Research, found the acquisition baffling. “I can’t figure out their focus. They are all over the place,” Bickley said. Bickley said making so many AI acquisitions in such a short timeframe means that there will be a lot of decisions about what stays and what doesn’t. “They are going to be integrating dozens of code bases in short order. When I zoom out, it is a little bit troubling.” “What they are trying to accomplish is not a bad goal: filling out their agent narrative. But they are trying to do it all at once. That means marketing will precede functionality,” Bickley said. For enterprise CIOs, he noted, “trying to map out the future will force a lot of questions. That tells me that they are trying to build the airplane while still in the air.” He noted that Salesforce would benefit from diversifying its installed base, because “its growth trajectory with enterprise is starting to wane”, and this acquisition would potentially help it increase its SMB market share. Fin claims 30,000 customers, which likely means most of them are SMBs. But that means Salesforce will have to make other changes, he said. “They will have to adjust. They cannot charge enterprise pricing with that model.” The news release about the announcement said that the acquisition “is expected to close in the fourth quarter of Salesforce’s fiscal year 2027” which spans the calendar dates from Nov. 1, 2026, through Jan. 31, 2027. The company declined a request for an interview about the deal. Fin had been known as Intercom for almost its corporate life, but it changed its name to Fin on May 12, just a few weeks before the acquisition announcement. Given that negotiations were almost certainly resolved by mid-May, as lawyers and marketers from both companies fine-tuned language for SEC filings and the news release, it seemed unusual to announce a corporate name change in mid-May, rather than wait for the acquisition to close and to then let the new owners decide. Justin Greis , CEO of consulting firm Acceligence, found the name change “fascinating,” but said his instinct is that the rebrand was driven by long-term positioning rather than short-term transaction considerations. “Founders and leadership teams typically spend months shaping the identity they believe best represents the future of the company,” he pointed out. “If they had already concluded that the market opportunity centered on Fin as a category-defining AI platform, changing course simply because acquisition discussions were underway may have felt shortsighted.” Needs a clear vision “For CIOs, implementation complexity has become one of the biggest barriers to AI adoption. The challenge is rarely access to models. It’s connecting data, workflows, governance, security, and business processes to deliver tangible results,” Greis said. “Enterprise AI is moving from a technology conversation to an execution conversation, and time-to-value is becoming the new battleground for competitive advantage. I think Salesforce is making a strategic bet that enterprises want multiple paths to adoption.” Greis added that enterprise CIOs already assume that product lines, especially those including generative AI and agentic AI, will constantly change. “What matters far more is whether Salesforce provides a clear vision for where customers should place their bets over the next three to five years and incentivize adoption, stickiness, and value to their customers,” he said. “Enterprise buyers can manage product change, but unclear roadmaps create much bigger problems than portfolio consolidation. No CIO would build their foundation on a platform that has a high risk of sunsetting in the next few years.” But, he added, enterprise CIOs today have little choice but to accept a lot of uncertainty from all of the major vendors. Others argued that CIOs still need some firm targets so that they can make concrete decisions for their enterprises. “The execution risk is acquisition indigestion, the same affliction that turns platform breadth into licensing fog,” said Sanchit Vir Gogia , chief analyst at Greyhound Research. “The most important date in this announcement is the one that does not exist: Salesforce has given a transaction close window but it has not given an integration timetable. Acquisitions run on three clocks. Ownership transfers fastest. Commercial alignment follows. Architectural integration, where identity, data and governance actually converge, runs slowest and decides the outcome. The credible path to meaningful convergence is 12 to 24 months beyond close, not a quarter beyond announcement.” Salesforce’s own history illustrates this, he said. A straightforward acquisition has closed inside two months, a complex one took the better part of a year, “and value realization trailed both,” Gogia said. A better way to pay for AI However, one of the potential benefits that Fin would bring to Salesforce is Fin’s differentiated pricing strategy. One of the aspects of AI that delivers the most problems for CIOs is the lack of any direct ROI . Fin has been trying to address that by charging only when a customer experiences an apparent good outcome. Nik Kale , a member of the Coalition for Secure AI (CoSAI) and of ACM’s AI Security (AISec) program committee, pointed out, “what Salesforce paid $3.6 billion for is an outcome-pricing model and a book of proof that the model works. Fin charges about 99 cents per resolved conversation and the customer pays nothing when the agent fails. That pricing structure is the actual asset because it sidesteps the exact thing that’s been gating agentic adoption, which is that enterprises don’t want to pay for agents they haven’t learned to trust yet.” But, Kale said, there is a catch. “For CIOs, the per-outcome model quietly moves the risk of a bad agent decision from the buyer back to the vendor, at least on paper. That sounds great until you ask who writes the definition of ‘resolved.’ In outcome pricing, a customer who gives up and walks away looks identical to a customer who actually got helped because both of them stop replying,” he said. “Silence gets billed as success. So outcome pricing looks like the vendor taking on your risk, but when the vendor also owns the definition of a win, you’ve handed them the scoreboard and the referee. The definition of ‘outcome’ is where the governance lives and, right now, it lives in the vendor’s measurement, not yours.” “Salesforce didn’t buy a better agent. They bought a better way to get paid for one,” he said.