First-person operator playbook

Migrate from Intercom to Fin (2026): operator playbook

Intercom rebranded the company to Fin in 2025-2026. If you're already on Intercom Inbox + Help Center, you're not migrating to a different vendor — you're enabling the AI agent (Fin) on top of your existing setup. This is the first-person operator playbook for that upgrade. I run Intercom 2 + Fin daily at my head-of-growth job, so the framing here is what I'd tell a friend evaluating the upgrade: pre-migration Help Center prep, per-conversation pricing math, the rollout phases, common pitfalls, and when to skip Fin for Typewise / Ada / Forethought.

By Nick French · Founder, StackSwap · 10yrs B2B SaaS GTM (BDR → AE → Head of Revenue) · Methodology →

Phase 1: Pre-migration prep (2-4 weeks)

Help Center hygiene

Fin is a RAG system pulling from your existing Help Center content. Resolution quality is directly tied to content quality. Audit checklist: (a) verify every article is current (no stale screenshots, broken UI references, retired feature mentions), (b) consolidate duplicate articles into single sources of truth, (c) add explicit "NOT this" framing to articles where Fin commonly confuses with adjacent topics, (d) write articles for your top 20 ticket categories specifically. Most teams spend 2-4 weeks on Help Center hygiene before enabling Fin — skipping this step makes Fin look worse than it is.

Macro / saved-reply consolidation

Audit your existing Intercom Macros + saved replies. Tag each by ticket category. Map macros to Help Center articles where there's overlap — Fin should pull from one canonical source, not multiple variants. Where macros contradict Help Center articles, fix the contradiction (don't let Fin see conflicting source data).

Per-conversation pricing math

Run the math before signing. Three inputs: (1) monthly conversation volume — count past 90 days ÷ 3, adjust for trend. (2) Fin per-conversation rate — varies by contract, ~$0.99 published typical, volume discounts apply. (3) Projected AI resolution rate — most production teams land 30-50%, model 40% as baseline. Formula: monthly conversations × Fin rate = monthly Fin cost. Then monthly conversations × resolution rate × fully-loaded agent cost per ticket = avoided cost. If avoided cost > Fin cost by 2-3×, math works. If close to 1× or below, Fin's per-conversation pricing is wrong shape — Typewise's outcome pricing wins.

Routing rules plan

Don't let Fin handle everything by default. Plan explicit routing for: ticket categories Fin handles first (tier-1: password reset, billing question, basic how-to, feature lookup), categories that route straight to human (tier-2/tier-3: anything needing judgment, edge cases, account-specific data), escalation triggers (Fin fails to resolve → human, customer expresses frustration → human, ticket includes specific keywords → human). Run these rules through QA before going live.

Phase 2: Enabling Fin (1 week)

The enable step

Inside Intercom: Settings → AI Agent → Enable Fin. Connect Help Center as primary knowledge source. Connect Macros as secondary source. Configure routing rules from Phase 1. Set conversation cap (start with 25% of inbound to limit blast radius). Enable feedback loop so Fin learns from human agent corrections.

Week 1 monitoring

Watch three metrics daily: (1) Fin resolution rate (target: above 30% in week 1, climbing to 40-50% by week 4 with tuning). (2) Customer satisfaction (CSAT) on Fin-handled tickets — should match or exceed human-handled CSAT for tier-1 categories. (3) Escalation rate — what % of Fin tickets escalate to human. Track which categories escalate most and add Help Center articles for those.

Phase 3: Tuning + scale-up (4-8 weeks)

Resolution-rate tuning

Most teams see resolution rate climb from 25-35% in week 1 to 40-50% by week 4 with active tuning. The tuning loop: (a) identify ticket categories where Fin's resolution rate is low, (b) audit Help Center content for those categories — add missing articles, fix conflicting articles, (c) review Fin's actual responses for those categories — flag wrong answers for retraining, (d) repeat weekly until resolution rates plateau at your category-specific ceiling.

Scale-up routing

Once Fin's resolution rate on the initial 25% conversation cap holds at target level for 2 consecutive weeks, scale routing to 50%, then 75%, then 100% over 4-6 weeks. Each scale step is a checkpoint — if resolution rates drop materially at higher volume, hold the scale and tune before continuing.

Tone + brand-voice tuning

Fin pulls tone from Help Center content. If your support team has a distinct reply voice (warmer, more technical, more terse), your Help Center may not match. Tone audit: pull 50 of Fin's recent replies, compare against 50 human-agent replies on similar categories, flag tone mismatches. Update Help Center articles to align with target tone. This is ongoing — tone drift is real as Fin learns from corrections.

Common pitfalls (first-person observations)

  1. Enabling Fin without Help Center hygiene first. Fin looks bad because the source data is bad. Pre-Fin Help Center audit is the highest-ROI work — most teams skip it and blame Fin for the resulting low resolution rates.
  2. Setting unrealistic resolution-rate targets. Marketing claims 60-70% resolution. Reality is 30-50% on well-tuned tier-1 categories. Plan economics at 40%; treat 50%+ as upside.
  3. No routing rules → Fin tries to handle everything. Fin will attempt tier-3 tickets it can't resolve and burn per-conversation spend. Explicit routing rules are non-negotiable.
  4. Skipping the tone audit. Fin replies that don't match your support team's voice damage brand consistency. Tone audit + Help Center tone refinement is part of Phase 3.
  5. Signing annual without quarterly minimum review: Negotiate quarterly minimum-spend review during contract signing so you can pause if resolution rates miss projections.

When to skip Fin and pick an alternative instead

FAQ

Intercom rebranded the company to Fin in 2025-2026 — the platform you used to call "Intercom" is now branded "Fin" with the AI agent as the headline product. If you're already on Intercom Inbox + Help Center, you're not migrating to a different vendor; you're upgrading within the same platform to add the AI agent (Fin) on top of your existing setup. The "migration" is enabling Fin, configuring it against your Help Center content, and routing tickets through Fin before they reach human agents. No data migration, no contact import, no help-desk swap.

Two phases. Phase 1 (1-2 weeks): enable Fin, validate Help Center content is current + comprehensive, tune Fin's resolution rate on your historical ticket categories. Phase 2 (4-8 weeks): monitor Fin's actual resolution rate against projections, refine knowledge base + macros where Fin misses, adjust routing rules to send the right tickets to Fin first. Total time-to-stable: 2-3 months. The pace depends on (a) how clean your Help Center already is, (b) how many ticket categories you want Fin to handle, (c) per-conversation pricing tolerance — slow rollout protects spend.

Three numbers. (1) Your monthly conversation volume (count the past 90 days, divide by 3 — be honest about whether trend is up or down). (2) Fin's per-conversation rate (varies by contract; typical published is ~$0.99 but volume-discounted). (3) Your projected AI resolution rate (most production teams land 30-50%, so model 40% as baseline). Math: monthly conversations × Fin per-conversation rate = monthly Fin cost (regardless of resolution). Then: monthly conversations × resolution rate × your fully-loaded agent cost per ticket = avoided cost. If avoided cost > Fin cost by 2-3×, the math works. If close to 1× or below, Fin's per-conversation pricing is wrong shape for your volume — Typewise's outcome pricing wins economically.

Three things to watch. (1) Resolution-rate gap vs marketing claims — Fin's marketed resolution rates assume well-tuned Help Center + clean ticket categorization. Your actual rate in week 1 is usually 40-60% lower than projected until you tune. (2) Tone consistency — Fin pulls from your Help Center voice, which may not match your support team's actual reply tone. Run a 2-week tone audit before letting Fin handle high-visibility ticket types. (3) Routing complexity — Fin tries to handle everything by default. You need explicit rules for what Fin handles vs what routes straight to a human. Without rules, Fin attempts low-success ticket types and burns per-conversation spend.

Yes. Fin's resolution quality is directly tied to Help Center quality — it's a RAG system pulling from your existing content. Pre-Fin Help Center audit: (a) verify every article is current (no stale screenshots, broken UI references, retired feature mentions), (b) consolidate duplicate articles into single sources of truth, (c) add explicit "NOT this" framing to articles where Fin commonly confuses with adjacent topics, (d) write articles for your top 20 ticket categories specifically — Fin needs source content to deflect tier-1 tickets. Most teams spend 2-4 weeks on Help Center hygiene before enabling Fin — that's the highest-ROI prep work.

Four cases. (1) Outcome-pricing math wins: at projected resolution rates below 50%, Typewise's $1/resolution beats Fin's per-conversation pricing. (2) EU compliance required: Typewise / Ultimate.ai have explicit ISO 27001 + EU AI Act posture that Intercom doesn't match. (3) Multi-help-desk environment: Fin is Intercom-only; if you run Zendesk + Salesforce Service Cloud too, Typewise integrates across all 3. (4) Brand-led conversational AI motion: Ada wins for top-of-funnel marketing-led chat that Fin isn't shaped for. For most Intercom-anchored teams without those constraints, Fin's native UI integration wins on friction.

Rolling back from Fin is straightforward: disable Fin, revert to human-agent routing, keep Intercom Inbox + Help Center as before. The platform isn't different — you're just turning off the AI agent. The harder rollback is the cost — if you signed a Fin contract with per-conversation commits, you may have minimum spend obligations. Negotiate quarterly review with Fin minimums during contract signing so you can pause if resolution rates don't hit projections within 60-90 days.

I run Intercom 2 + Fin daily at my head-of-growth job, so this is first-person. The honest read: Fin works on well-defined tier-1 ticket categories (password reset, billing question, basic how-to, feature lookup). Resolution rates 40-50% on those categories after tuning. Fin struggles on tier-2/tier-3 tickets needing judgment, edge cases, or context Fin doesn't have. The economic question depends on your ticket mix — if 70%+ of your tickets are tier-1 and Fin handles 50% of those, the math works. If most of your tickets are tier-2/tier-3, Fin's per-conversation cost burns spend without proportional resolution. Audit your ticket distribution before signing.

Related reading

Canonical URL: https://stackswap.ai/migrate-from-intercom-to-fin