Intercom Customer Service Software Features: The Complete 2026 Guide for Support Teams (Now Fin, Acquired by Salesforce)

Table of Contents

• What you'll achieve with this guide

• What Is Intercom?

• Intercom is now Fin: what the 2026 rebrand and Salesforce acquisition mean for buyers

• Intercom company snapshot

• Is Intercom a CRM, a helpdesk, or an AI agent?

• Prerequisites before you evaluate or deploy

• Step 1: Map your channels to how Intercom works

• Step 2: Evaluate the unified inbox

• Step 3: Evaluate Intercom as a ticketing system

• Key Capabilities of Intercom for Support Agents

• Step 4: Price it out, seats versus per-outcome AI

• Step 5: Test reporting, analytics and predictive features

• Step 6: Run a scoped pilot and migrate knowledge

• Intercom limitations and where teams hit walls in 2026

• Compliance in 2026: SOC 2, GDPR and the revised EU AI Act timeline

• Intercom vs. Alternative Support Platforms

• Common mistakes when evaluating Intercom

• How to measure whether the rollout worked

• Choosing your next step

What you'll achieve with this guide

By the end of this guide you will be able to evaluate Intercom customer service software features the way a buyer does in 2026: channel by channel, feature by feature, and dollar by dollar. You will know what changed with the Fin rebrand, what the pending Salesforce acquisition means for your contract, and how to score the inbox, the ticketing layer and the analytics stack against your own volume.

Intercom, rebranded to Fin in 2026, is an AI-first customer service platform that unifies live chat, email, WhatsApp, SMS, phone and Slack into one workspace where an AI agent and human agents work the same conversation. Its AI agent, Fin, resolves queries end to end using a proprietary model called Apex. Helpdesk seats are still sold as Essential, Advanced and Expert, while AI usage is billed per outcome rather than per seat.

Two things happened in mid-2026 that make every pre-2026 Intercom guide wrong. The company renamed itself after its AI agent, and Salesforce signed a definitive agreement to buy it for roughly $3.6 billion (MarTech). Everything below reflects the state of the product as of 2026-07.

This is a working how-to, not a brochure. Each step names the actual screens, plan tiers and billing units you will meet, and ends with something you can check off.

What Is Intercom?

Intercom is a customer service platform that centralizes support conversations across messaging channels into a single workspace shared by AI and human agents. Since its 2026 rebrand it operates under the name Fin, named after its AI agent, and markets itself AI-first; helpdesk plans are now sold as "Fin and Intercom". The word most people still search for, and the word used throughout this guide, is Intercom.

Its architecture rests on four layers. A messaging layer captures conversations across live chat, email, WhatsApp, SMS, phone and Slack. An automation layer routes and prioritizes by rules you define. A knowledge layer powers self-service. And Fin, the AI agent, resolves queries end to end.

Unlike legacy helpdesks built around email tickets, Intercom's model centers on real-time messaging. A customer who starts in chat and follows up by email stays one continuous thread, not two tickets. That single design decision explains most of the difference in daily agent workflow.

The 2026 version of Fin is materially more capable than the 2025 version. Salesforce describes Fin as resolving, on average, 76% of support volume without a human (The Next Web). The "suggests responses to agents" function is now a separate paid add-on called Copilot.

Intercom meaning, in one line: in customer service software, "Intercom" refers to this specific vendor's messaging-first support platform, not to the building hardware of the same name.

Intercom is now Fin: what the 2026 rebrand and Salesforce acquisition mean for buyers

Intercom adopted the name Fin roughly one month before June 2026, taking the name of its AI agent, and on 15 June 2026 Salesforce signed a definitive agreement to acquire the company for approximately $3.6 billion. The deal is expected to close in Q4 of Salesforce's fiscal year 2027, subject to regulatory clearances. Fin's technology is slated to fold into Agentforce (MarTech).

The deal is signed, not closed. Build change-of-control terms into your contract.

The distinction that matters to your procurement team: the deal is signed, not closed. Salesforce does not own Intercom today, and no source states whether pricing or plan structure will change after close.

Here is what a buyer should actually do with that information.

The strategic logic Salesforce gave is worth reading as a product statement. Agentforce suits large organizations that need deep customization; Fin arrives pre-trained and fast to deploy for SMB and mid-market teams (MarTech).

If you are a mid-market team signing a three-year deal this quarter, that framing is your negotiating leverage. You are buying the fast-to-deploy half of a merger that has not happened yet.

Intercom company snapshot

Intercom is the customer service software company behind the Intercom helpdesk and the Fin AI agent, now operating under the Fin brand and reachable at intercom.com. It sells to support, sales and product teams, with a helpdesk sold in three seat tiers and an AI agent sold on consumption. Its ownership status is the single most current fact about it: an acquisition agreement with Salesforce, signed but not yet closed.

People searching for "intercom.io" are almost always looking for this company. The canonical product domain is intercom.com, with the AI agent documented separately at fin.ai.

Founding-year claims circulate widely, but this guide does not repeat them because they were not verifiable during research for this update. Where a number is not sourced here, treat competing pages that state it confidently with appropriate suspicion.

Company financials tell you nothing useful either, because ARR, customer count and headcount are not disclosed in the acquisition coverage. The $3.6 billion price tag is the only public valuation signal.

Is Intercom a CRM, a helpdesk, or an AI agent?

Intercom is primarily a helpdesk with an AI agent bolted to the front of it, plus enough customer data and outbound messaging to feel CRM-adjacent without replacing a real CRM. It stores customer attributes, conversation history and product usage signals, and it can trigger proactive campaigns. It does not manage pipeline, forecasting or opportunity stages the way Salesforce or HubSpot do.

The cleanest way to categorize it in 2026 is by which of three jobs you are buying.

• As a helpdesk: a shared inbox, ticket objects, SLAs, macros, routing rules and a help center. This is the paid-seat part.

• As an AI agent: Fin, sold per outcome, capable of running on top of Salesforce, HubSpot, Freshworks or Zoho without any Intercom seats at all.

• As a lightweight CRM layer: customer profiles, segments, tours and outbound messages. Useful, but not a system of record for revenue.

That third bullet is why "is Intercom only for customer support?" keeps appearing as a question. It is not, teams use it for onboarding tours, product announcements and lead qualification, but support is where the seat economics make sense.

If your team already runs Salesforce as the system of record, the interesting question is not helpdesk versus CRM. It is whether you buy Fin as a standalone agent layer over what you have, which is now an officially supported deployment.

Prerequisites before you evaluate or deploy

Before you touch a trial, assemble five things: a channel inventory, a knowledge base audit, admin access to the systems Intercom must integrate with, a monthly conversation count, and a named owner for AI quality. Missing any one of these turns a 30-day evaluation into a 90-day one. The conversation count matters most, because 2026 AI pricing is consumption-based and you cannot model cost without it.

Tools and access you need:

• Admin rights in your current helpdesk to export conversations and articles

• API credentials for your CRM, billing system and product analytics

• A staging or sandbox environment for workflow testing

• Your last 12 months of conversation volume, split by channel and by intent

• Your current CSAT and first-response-time baselines

Plan tiers and what unlocks where:

Seat prices render client-side on Intercom's own pricing page and did not expose themselves to automated verification during research. Third-party GetVoIP, updated 2026-07-01, lists $29 / $85 / $132 per seat per month on annual billing; another tracker lists $39 / $99 / $139. Confirm at checkout before you budget.

Permissions to sort out early: who can publish help center articles, who can edit Fin's answer sources, and who can approve a workflow that touches billing. Support leads routinely discover in week three that only one admin can change routing.

Step 1: Map your channels to how Intercom works

Intercom captures customer input across live chat, email, WhatsApp, SMS, phone and Slack, then threads them into a single conversation rather than separate tickets. Real-time messaging is the architectural center: a chat that goes quiet and returns as an email stays one thread with one history. Before evaluating anything else, list every channel your customers actually use and check it against that inventory.

Customer Input: Omnichannel Message Capture works like this in practice. A message arrives on any channel and Intercom enriches it before an agent ever sees it, attaching previous conversations, product usage data, subscription status and any custom attributes your team configured.

That enrichment step is the thing legacy ticketing cannot replicate cheaply. It is also the thing that breaks if your customer attributes are not synced, which is why the prerequisites section put API credentials before trial signup.

Do this now:

1. List your channels in order of volume. If phone is 40% of your contacts, Intercom's voice support becomes a primary evaluation criterion, not a footnote.

2. For each channel, note whether conversations currently carry customer context or arrive blind.

3. Flag any channel where you require strict SLA timers, since real-time channels and email SLAs behave differently.

4. Check whether the channels you need are covered on your candidate seat tier or require an add-on.

5. Test one cross-channel handoff during the trial: start in chat, reply by email, confirm it stays a single thread.

Fin now operates across all of those channels, not just chat. Exact multilingual coverage was not verifiable during this research, so ask for language-specific resolution rates in writing rather than accepting a general claim.

If a large share of your volume is voice, price it separately. Intercom's pricing page showed no distinct Fin Voice SKU during verification, so treat voice economics as an open question to resolve with sales.

Step 2: Evaluate the unified inbox

Unified Inbox: Centralized Conversation Management

The Intercom inbox consolidates every conversation regardless of channel into one queue, with filters for unassigned messages, your own conversations, and tag-based views like "billing" or "urgent." Keyboard shortcuts (J/K to navigate, R to reply, E to resolve) let agents move through volume without a mouse. Status is visible at a glance: open, pending customer response, snoozed, or resolved.

The interface is split-screen. Conversation thread on the left, context panel on the right showing customer details, history, linked accounts and relevant knowledge articles. Bulk actions let you assign, tag or close multiple threads at once.

That is the marketing description. Here is the rubric that separates a good inbox from a demo-friendly one.

Inbox evaluation checklist. Score each 0-2, target 18+ out of 24:

Criterion 6 is the one most teams skip and most regret. When an AI agent handles the first four turns of a conversation, the human who inherits it needs the full transcript plus the reason for escalation, not a summary.

Criterion 7 separates a messaging tool from a support platform. Ask specifically: can an agent process a refund from inside the conversation, or does the integration merely link to an external page? We wrote about where that boundary sits in a piece on when an AI should act versus hand off.

Run this rubric during a trial with real conversations, not a sandbox. Sandbox inboxes are always fast because they are empty.

Step 3: Evaluate Intercom as a ticketing system

Intercom does have a ticketing system, but it is a messaging-first one: conversations are the primary object and tickets are a structured layer applied on top, rather than the reverse. That suits teams whose work arrives as chat and gets resolved in minutes. It fits less naturally where work arrives as long-lived cases with multi-day SLAs, approvals and hierarchical queues.

To evaluate the customer service company Intercom on its ticketing system specifically, score it on the seven dimensions below rather than asking whether it "has tickets."

The escalation dimension is where Intercom does well. Internal notes, screenshots and reproduction steps live inside the same thread, so an engineer sees the full history without asking the customer to repeat anything.

The audit dimension is where AI-heavy setups get complicated. When Fin resolves a conversation autonomously, your audit trail needs to show which knowledge sources it used and why it decided the issue was closed.

A practical rule: if more than 20% of your volume needs a case to stay open for over 72 hours with multiple internal owners, test Intercom's ticketing hard before committing. If most of your volume closes same-day, the conversation-first model is an advantage.

Key Capabilities of Intercom for Support Agents

The Intercom customer service software features that matter day to day cluster into six areas: the unified inbox, the Fin AI agent, the help center, workflows, the customer context panel, and reporting. Five of the six are unchanged in shape since 2025; the AI agent is the one that was rebuilt. Below is what each does now, in 2026 terms.

Fin AI Agent: End-to-End Resolution

Fin is no longer a deflection bot that answers FAQs. It is positioned as an AI agent that resolves complex customer queries end to end across live chat, email, WhatsApp, SMS, phone and Slack, powered by the proprietary Apex model (MarTech). Salesforce states Fin resolves an average of 76% of support volume without a human (The Next Web).

Fin can also run standalone on other helpdesks, including Salesforce, HubSpot, Freshworks and Zoho, with no Intercom seats at all. That changes the buying decision: you can adopt the AI layer without migrating your helpdesk.

The agent-assist function that used to be bundled is now Copilot, a separate add-on. If your evaluation assumed suggested replies came free with a seat, re-check your quote.

Help Center and Articles: Self-Service Knowledge Base

Your help center lives at a custom URL and embeds into your product, with Fin interpreting customer questions and surfacing relevant articles rather than keyword-matching. Agents can create articles directly from a conversation, so a repeatedly asked question becomes documentation in a few clicks. Articles support rich media, code snippets and multiple language versions.

The quality of this layer determines the quality of everything above it. An AI agent trained on stale articles produces confident, stale answers.

Audit your articles before you turn Fin on, not after. Kill duplicates, date-stamp policy pages, and assign an owner per category.

Workflows and Automation: Rule-Based Routing and Task Automation

Workflows are if-then rules built in a visual editor with no code required. A typical rule: if a conversation mentions "refund" and the customer is on the Pro plan, assign to billing and tag as high priority. Another: if a conversation sits unassigned for 10 minutes, notify the team lead.

Common patterns include after-hours auto-responses, VIP escalation rules and follow-up sequences on unresolved threads. Most teams end up with 15 to 40 active workflows within six months.

The maintenance cost is real. Every workflow is a small piece of logic someone has to remember, and undocumented rules are the leading cause of "why did this ticket go there?" Slack threads. We have written about why support automation that maintains itself changes that math.

Customer Context Panel: Real-Time User Data

The context panel surfaces customer details, conversation history and integration data in the right-hand rail so agents stop tab-switching. Custom attributes like subscription tier or account value appear alongside recent purchases, tickets in other systems and product usage metrics. It is the single feature agents mention most in reviews.

Its usefulness is entirely a function of your integrations. A context panel wired to nothing is an email address and a timezone.

Ticketing and Escalation

Covered in depth above. The short version: conversations are primary, tickets are structured on top, and engineering handoffs keep full context inside the thread.

Reporting and Analytics

Covered in depth in Step 5, since it is where evaluations most often go shallow.

Step 4: Price it out, seats versus per-outcome AI

Intercom pricing in 2026 has two independent meters: paid seats for human agents, and per-outcome consumption for Fin. Fin AI Agent is priced at $0.99 per outcome, where an outcome is a resolution, a procedure handoff or a disqualification; lead qualifications bill at $9.99 each, and there is a 50-outcome monthly minimum when Fin runs on a non-Intercom helpdesk (Fin). Copilot is a separate add-on.

Dual-meter consumption pricing versus a bundled monthly resolution allowance.

Note the billing-unit change carefully. Pre-2026 guides say "per resolution." The current unit is "per outcome," defined as: a customer confirms their issue is resolved, or they do not ask for more help after Fin responds, or Fin completes a workflow.

That definition includes conversations where the customer simply went away. It is a broader unit than a verified resolution, which matters when you compare against vendors billing on confirmed outcomes only.

Worked example, 1,000 conversations per month, 8 agents:

Assume Fin handles 76% of volume to an outcome, in line with the vendor's stated average. That is 760 outcomes at $0.99, or $752.40. Eight Advanced seats at the third-party annual figure of $85 add $680. Copilot for all eight agents adds $232.

Total: roughly $1,664 per month, before the Pro analysis add-on ($99) and before any lead qualifications at $9.99 each. Add 100 qualifications and you are at $2,763.

Now stress-test it. At 5,000 conversations a month with the same 76% outcome rate, Fin alone costs $3,762, and your seat count probably rises too. Consumption pricing is linear in volume, which is exactly what a growth-stage team should model before signing.

For contrast, Fini bundles the platform, implementation and a monthly resolution allowance into one plan: Growth at $3,600/mo ($3,000/mo billed yearly) with 2,000 resolutions included and $0.89 per resolution beyond that; Scale at $9,000/mo ($7,500/mo billed yearly) with 8,000 resolutions plus 500 answered voice calls and $0.69 per resolution beyond. There are no per-seat fees, annual billing gives two months free, and unused allowance rolls forward one month. Enterprise is custom pricing, contact Fini.

The structural difference is worth naming: Intercom charges for seats and AI separately, so cost scales with both headcount and volume. Allowance-based pricing scales with volume only.

Step 5: Test reporting, analytics and predictive features

Intercom's reporting covers real-time operational dashboards (first response time, resolution time, CSAT, conversation volume), per-agent performance views, and conversation analysis that clusters topics across your volume. Its predictive and data-analysis features sit mostly in that conversation-analysis layer, which surfaces recurring intents, escalation patterns and knowledge gaps. Treat the word "predictive" carefully during a demo and ask what is forecast versus what is merely aggregated.

Intercom's predictive analytics and real data analysis features are best evaluated with a specific list of questions. Here is the one to bring.

Reporting evaluation questions:

1. Can I segment every metric by channel, plan tier, and AI-handled versus human-handled?

2. Does the AI resolution number count verified resolutions or all outcomes, including customers who stopped replying?

3. Can I see the specific knowledge sources Fin used on any given conversation?

4. Does topic clustering run on all conversations or only a sampled subset?

5. Is conversation analysis included, or is it the $99-per-1,000-conversations Pro add-on?

6. Can I export raw conversation data to my warehouse, and on what schedule?

7. Can I forecast staffing from volume trends inside the tool, or does that need a separate WFM product?

8. Are agent-level metrics visible to agents themselves, or only to managers?

Question 2 is the one that reshapes budgets. Deflection counts any conversation that never reached a human, including customers who gave up; resolution counts only conversations genuinely solved. Those two numbers diverge, and the 2026 industry-wide move to outcome and verified-resolution billing exists precisely because they do. We unpacked that gap in detail in a piece comparing deflection rate against true resolution rate.

Maturity is measurable and it correlates with reporting discipline. Intercom's own 2026 Customer Service Transformation Report finds 87% of teams at the mature deployment stage report improved metrics since implementing AI, compared to 62% overall (Intercom).

The same report finds 82% of senior leaders say their teams invested in AI for customer service over the last 12 months, with 87% planning to invest in 2026 (Intercom). Investment is no longer the differentiator. Measurement is.

Set your baselines before go-live: current FRT, current CSAT, current cost per contact, current escalation rate, and current backlog age. If you do not capture these in the two weeks before launch, you will spend the next quarter arguing about whether anything improved.

Step 6: Run a scoped pilot and migrate knowledge

Run the pilot with three to five agents on one conversation type for two to four weeks before touching the rest of your operation. Then migrate knowledge, configure workflows, train agents, and cut over gradually with the old system in read-only mode for 30 to 60 days. The knowledge migration is the long pole: budget four to six weeks and involve the agents who know which questions customers actually ask.

Knowledge migration is the long pole: budget four to six weeks.

Phase-by-phase plan:

Do not copy-paste articles during migration. Rewrite them in the plain, direct language an AI agent can parse, one question per article, with the answer in the first paragraph.

Set a firm cutoff date for the old system. Open-ended "we'll switch when everyone's comfortable" timelines run for months and stop teams from committing to either platform.

For teams weighing whether this whole sequence is worth it versus a faster deployment path, our comparison of AI support chatbots and true AI agents covers what the six phases look like when the AI layer arrives pre-trained.

Intercom limitations and where teams hit walls in 2026

Intercom's four recurring limitations in 2026 are dual-meter cost growth, an acquisition that has not closed, a conversation-first data model that strains under long-lived cases, and an outcome billing definition broader than a verified resolution. None of these disqualify the product. All of them belong in your risk register.

Where teams actually hit walls:

• Cost scales on two axes. Seats grow with headcount, outcomes grow with volume. A team that doubles volume and adds four agents pays more on both meters at once.

• The Salesforce deal is unresolved. Signed 15 June 2026, expected to close in Q4 of Salesforce FY27, subject to regulatory clearances (MarTech). No public source states what happens to pricing or plan structure after close.

• Outcome definitions favor the vendor. An outcome includes a customer who simply stopped replying. Ask for a breakdown of outcomes by type before signing.

• Copilot is an extra line item. Agent-facing suggestions cost $29 to $35 per agent per month on top of seats.

• Long-lived cases fight the model. Multi-week cases with approval chains fit a case-management tool better than a messaging-first inbox.

• Language coverage is unverified. Fin's exact multilingual capability was not confirmed during this research. Request per-language resolution data.

• Voice pricing is opaque. No distinct Fin Voice SKU or price surfaced on the pricing page during verification.

Two more caveats belong here for completeness. Integration counts widely quoted as "300+" were not verifiable, so check the current app directory for the specific systems you need. And customer-story ROI figures that circulated pre-rebrand may no longer be live, so ask for a reference call instead of a case study PDF.

None of that is unique to Intercom. Every vendor in this category has a version of the same list, which is why the evaluation rubrics in this guide matter more than any single feature claim.

Compliance in 2026: SOC 2, GDPR and the revised EU AI Act timeline

SOC 2, ISO 27001 and GDPR remain the baseline for any support platform touching customer data, and HIPAA plus a BAA is the additional bar for healthcare. What changed in 2026 is the EU AI Act calendar: the compliance cliff many 2025 articles placed in August 2026 has moved. Buyers should plan for AI Act readiness on the revised 2027-2028 timeline, not an imminent 2026 deadline.

Under the AI/Digital Omnibus, stand-alone Annex III high-risk obligations move from 2 August 2026 to 2 December 2027, and Annex I embedded systems move to 2 August 2028; watermarking and transparency marking obligations shift to 2 December 2026. Provisional agreement was reached 7 May 2026, Parliament endorsed it on 16 June 2026, and the Council gave final approval on 29 June 2026 (Council of the EU).

Annex III high-risk use cases cover recruitment, credit scoring, education, law enforcement and border control. Most customer support AI does not land there, but support agents that make eligibility or credit decisions might, which is a question for your legal team rather than your vendor's sales engineer.

Compliance checklist to run on any vendor:

Intercom publicly positions on SOC 2 and GDPR; verify current certificate scope directly with the vendor rather than relying on any third-party guide, including this one. Zendesk does not publicly state a per-automated-resolution rate on its pricing page, so compliance and pricing questions alike belong in a written RFP.

Fini's own posture, stated as a vendor claim: SOC 2 Type II, ISO 27001, HIPAA-compliant and BAA-eligible, GDPR and CCPA.

Intercom vs. Alternative Support Platforms

The 2026 comparison set is no longer Intercom versus Zendesk. It is Intercom/Fin, Zendesk's Resolution Platform, Salesforce Agentforce (which is also Intercom's acquirer), and outcome-priced AI agent specialists like Fini. All four have converged on billing for results rather than seats, and they diverge on how honestly a "result" is defined.

Intercom is the strongest fit for messaging-led teams that want a pre-trained AI agent live quickly and are comfortable with a pending acquisition. Its channel coverage is broad and Fin's ability to run on rival helpdesks removes the usual migration blocker.

Zendesk rebuilt its story at Relate 2026 on 19 May, launching the Autonomous Service Workforce and pricing AI agents on independently verified resolutions rather than seats. CEO Tom Eggemeier put it bluntly: "The era of the chatbot, the era of frustration and deflection, is over" (CMSWire). Its verified-resolution definition, which excludes spam and routine exchanges, is stricter than a broad outcome count.

Salesforce Agentforce makes sense if Salesforce is already your system of record and you need deep customization. Its Flex Credits model prices individual actions rather than conversations, which suits workflow-heavy automation. Note that Flex Credits and Conversations cannot be used in the same org.

Fini takes the agentic path: the AI does not just answer, it acts, with vendor-stated 99% accuracy and a 90% resolution rate, deployed live in 30 days on top of your existing stack. Our head-to-head with Decagon covers how that architecture compares within the AI-agent category specifically.

Gartner's forecast frames where all four are heading: agentic AI will autonomously resolve 80% of common customer service issues without human intervention by 2029, cutting operational costs by 30% (Gartner). The vendors differ mainly on how much of that they can do today.

Common mistakes when evaluating Intercom

Most failed Intercom evaluations share five errors: modelling cost on seats alone, accepting a deflection number as a resolution number, skipping the knowledge audit, ignoring the acquisition in contract terms, and demoing on clean sandbox data. Each is avoidable in an afternoon of preparation. Each costs six figures if you skip it.

The knowledge-audit mistake deserves extra weight. AI quality is downstream of content quality, and no model fixes a contradictory refund policy documented in three places. Our write-up on how a structured knowledge layer differs from retrieval explains why that failure mode is architectural, not a tuning problem.

The demo-data mistake is the most common. Ask for a trial on your own conversation history, and if a vendor cannot support that, treat it as a data point.

How to measure whether the rollout worked

Six metrics prove an Intercom rollout worked: verified resolution rate (not deflection), cost per resolved contact across both meters, CSAT split by AI-handled versus human-handled, first response time, escalation precision, and backlog age. Capture all six for two weeks before go-live, then compare at day 30 and day 90. Anything measured only after launch is a story, not a result.

Two reporting habits separate teams that improve from teams that plateau. First, always report resolution and deflection as two separate numbers so nobody quietly conflates them. Second, review a random sample of 20 AI transcripts every week with a named owner.

Intercom's own data supports the maturity link: 87% of teams at the mature deployment stage report improved metrics since implementing AI, versus 62% overall (Intercom). Mature deployments are the ones that measure weekly.

Tie the numbers to money in the board deck. Cost per resolved contact, multiplied by volume, is the only figure a CFO will use to renew.

Choosing your next step

The right next step depends on which problem you actually have: if you need a messaging-led helpdesk with a fast-to-deploy AI agent, trial Intercom with real volume and the rubrics above; if you already have a helpdesk you like, evaluate AI agent layers that sit on top of it; if your priority is predictable cost, compare allowance-based pricing against dual-meter consumption. Run the same eight-question reporting test and the same eight-point inbox rubric on every vendor. Score them on paper before anyone takes a sales call.

Do three things this week. Pull your last 12 months of conversation volume by channel and intent. Audit your top 50 intents against your current help center articles. Then build a 12-month cost model on both a seat-plus-consumption structure and an allowance structure, at today's volume and at double.

If you keep your current helpdesk and want the AI layer to do the work, that is exactly the shape Fini is built for: an agentic AI that resolves and acts on top of your existing stack, with vendor-stated 99% accuracy, a 90% resolution rate, and bundled implementation. Teams like Wefunder went that route rather than replatforming.

If you are weighing Intercom against an agentic AI layer for a support queue that is growing faster than your headcount, book a 30-minute working session with Fini and bring your volume numbers; the team will model your resolution allowance against your actual channel mix and show what going live in 30 days looks like for your queue.