Short answer: Salesforce completed its acquisition of Fin, formerly Intercom, on September 10, 2026. Salesforce says the transaction brings Fin’s customer-agent platform, technical AI team and customer base of more than 30,000 companies into Salesforce. Fin’s AI customer agent is described as handling complex queries across live chat, email, WhatsApp, SMS, voice and Slack. For customer-service and marketing-automation teams, the immediate task is not to assume a product merger; it is to verify the current channel, data, handoff, governance and commercial terms for the workflow you want to operate.
What changed on September 10
In its closing announcement, Salesforce says Fin’s customer-agent platform and technical AI team now join Salesforce, and that Fin will continue serving customers and advancing its AI-agent capabilities as part of Salesforce AI Labs. The release describes the transaction as expanding Salesforce’s customer-service AI portfolio, not as a statement that every Fin and Agentforce surface has already been consolidated.
Salesforce attributes an average 76% resolution rate to Fin’s AI customer agent. Keep that number in its proper category: a vendor-reported claim about Fin, not an independently audited benchmark, a forecast for a new deployment or a guarantee for a marketing funnel.
How Fin and Agentforce fit in the decision
Fin’s value proposition is a customer-agent experience that can be deployed across channels and existing systems. Agentforce is Salesforce’s deeply customizable platform for building and governing enterprise agents. The acquisition may broaden Salesforce’s choice between rapid customer-agent deployment and tailored enterprise transformation, but the close announcement does not define a universal replacement, migration or product-consolidation path.
| Question | Verify before making a decision |
|---|---|
| Customer context | Which Salesforce objects, knowledge sources and external systems can Fin read, and under which identity? |
| Channel coverage | Which of live chat, email, WhatsApp, SMS, voice and Slack are enabled for the current org and region? |
| Handoff | How does an autonomous conversation transfer to a human, queue or specialist workflow? |
| Action scope | Can the agent update records, trigger workflows or send messages? Which actions require approval? |
| Measurement | How is resolution defined, deduplicated and reconciled against cases, qualified leads, revenue and refunds? |
| Commercials | What are the current package, outcome, seat, usage, minimum-commitment and implementation terms? |
What current Fin documentation shows
Fin’s Salesforce integration documentation describes a setup sequence that includes syncing knowledge articles, connecting to Salesforce with OAuth, customizing the greeting and handoff, and testing before going live. It also describes cases, email, web forms and live chat, with other channels and availability subject to the current product surface.
Documentation is useful for the present integration path, but it is not a post-close roadmap. Preserve the exact documentation date, account entitlement and region in any implementation plan. Do not turn “works with Salesforce” into an assumption that all data, actions or channels are available in every org.
Implications for marketing and customer operations
- Map the customer journey. Separate a marketing lead, a support case, a product question and a sales conversation. An agent resolution is not automatically a qualified lead.
- Define the source of truth. Decide whether customer identity, consent, case status, conversation outcome and revenue live in Salesforce, Fin, another CRM or a controlled data layer.
- Design handoffs. Record when the agent asks for a person, why it escalates and how context follows the conversation.
- Measure business outcomes. Track resolution, deflection, qualified opportunities, retention or revenue with explicit definitions; keep vendor resolution claims separate from your account results.
- Review risk. Test incorrect context, overconfident answers, prompt injection, unauthorized actions, duplicate messages, sensitive data and failure recovery.
Packaging and migration are still open questions
The completed acquisition announcement does not specify a complete post-close packaging matrix, a mandatory migration deadline, a universal price, a product sunset or a region-by-region rollout. Teams should ask Salesforce for written answers tied to the current contract:
- Is the existing Fin deployment supported unchanged during the transition?
- Which Salesforce products or Data Cloud dependencies are required?
- How are existing knowledge, routing, channels, procedures and analytics migrated?
- Which data-processing, retention and residency terms apply?
- How are outcome charges, usage, seats, minimum commitments and overages calculated?
- What support and rollback path exists if resolution quality falls?
For separate commercial context, see the site’s Salesforce pricing and buying checklist. For a broader enterprise-agent architecture discussion, see AI agent harnesses in 2026. Neither page supplies Fin-specific pricing or migration terms.
Readiness sequence for an existing customer operation
Teams that already use Salesforce or Fin can make the acquisition decision concrete with a staged review:
- Inventory: list the help desk, knowledge base, CRM objects, routing rules, channels, bots and human queues currently in production.
- Define outcomes: separate self-service resolution, assisted resolution, qualified sales conversation, renewal, escalation and revenue. Give each outcome an owner and timestamp.
- Choose a low-risk cohort: start with a bounded intent, language, product or support queue where a human can review results.
- Test context and handoff: confirm that the agent receives the right account, case and knowledge context and that escalation preserves the conversation history.
- Validate controls: attempt unauthorized data access, sensitive-data disclosure, duplicate actions and unsafe instructions before broadening access.
- Reconcile: compare agent outcomes to case records, CRM stages and customer feedback; record false resolutions and avoidable escalations.
This sequence does not assume that Salesforce has published a required migration plan. It is an operator checklist for reducing uncertainty while current product and contract information is gathered.
Channel-specific questions
Channel coverage is not the same as channel readiness. Live chat may need an authenticated account context and a clear handoff queue. Email may need threading, attachment and reply-to rules. WhatsApp and SMS require consent, template and session controls, retention rules and a plan for message delivery failures. Voice requires transcripts, recording permissions, interruption handling and a safe path when the agent cannot verify identity. Slack or other internal surfaces need workspace permissions and a policy for sharing customer data with employees.
For each channel, document the source of identity, the allowed data fields, the approved actions, the escalation destination, the user-visible disclosure and the retention period. Keep channel-level outcomes separate when resolution definitions differ. A conversation that is closed by automation is not necessarily a customer issue solved, and a sales handoff is not necessarily a qualified opportunity.
Measurement, governance and customer trust
Use a measurement contract that names the unit of analysis and prevents double counting. One customer may open multiple conversations, switch channels or return after a human handoff. Define whether resolution is counted per case, conversation, customer or issue; state how reopened cases are handled; and preserve the timestamps needed to distinguish an agent response from a business outcome.
Review quality with both quantitative and qualitative evidence. Resolution rate, escalation rate, first-response time and cost can be useful, but a lower escalation rate is not automatically better if the agent is failing to recognize uncertainty. Sample transcripts under an approved privacy process, review unresolved and reopened cases, and give support teams a way to report harmful or misleading behavior.
The acquisition also makes governance a continuity question. Preserve access logs, knowledge versions, routing rules, consent records and rollback instructions while commercial or product changes are clarified. Ask who owns incident response, model evaluation, security review and customer communication during the transition. These controls protect the customer journey regardless of how Salesforce packages Fin in the future.
Frequently asked questions
Does Fin replace Agentforce?
The acquisition announcement does not say that Fin replaces Agentforce. It says Fin broadens Salesforce’s customer-service AI portfolio. Treat product overlap and future consolidation as questions to verify.
Can Fin handle WhatsApp and voice?
Salesforce’s announcement names WhatsApp and voice among Fin’s channels. Actual access, configuration, consent and regional availability still depend on the current product documentation and customer agreement.
Does a 76% resolution rate predict my results?
No. It is an attributed Salesforce/Fin average claim. Your resolution definition, customer mix, knowledge quality, routing, escalation policy and measurement window can produce a different result.
Should a team migrate immediately?
Not from the acquisition announcement alone. Inventory the current workflow, preserve a rollback path, test a bounded use case and obtain written commercial and technical answers before changing a production customer journey.
Bottom line: the close makes Fin part of Salesforce’s customer-agent strategy. The useful response for operators is a documented capability and migration review, not an assumption that products, contracts, prices or roadmaps have already become one.
