Intercom Fin vs UptoNova — Why Paying Per "Resolution" Hurts a Small Business
Intercom Fin is a strong AI customer support agent that charges roughly $0.99 every time it closes a chat. The model works well for big enterprise help desks where every "resolution" is a clear ticket. It quietly breaks when your business actually sells over WhatsApp, Instagram, and Telegram — because "resolution" becomes squishy, the same customer counts twice across channels, and every reply takes a bite out of your margin.
- Per-resolution pricing — Fin charges around $0.99 each time it closes a chat — incentivizes the vendor to define "resolution" generously and discourages you from putting AI on sales chats at all.
- For a small store doing 8,000 conversations a month, Fin can land in the $3,000-$5,000 range. A flat-priced AI sales agent runs a fraction of that.
- Intercom is the right answer for enterprise software companies with big help centers and ticketing workflows. UptoNova is the right answer for small and growing businesses selling on messaging channels.
What Fin actually is
Intercom Fin is an AI customer support agent that sits on top of Intercom's web messenger, email, and (since late 2024) WhatsApp and Instagram. It pulls answers from Intercom's help center articles and connected knowledge sources. Pricing is per "resolution" — roughly $0.99 each time Fin closes a conversation without handing off to a human. Check intercom.com/fin for current numbers.
Fin is excellent at what it was built for: deflecting support tickets for a software company with thousands of help center articles.
Why "resolution" pricing breaks for sales chats
A "resolution" in Fin's world is a conversation that ended without the customer reopening it for a while. That definition is reasonable for a traditional support inbox. For a sales chat it gets uncomfortable:
- A customer asks about a product, gets pricing, doesn't reply. Was that a resolution? Yes — you pay. Did the customer actually buy? Probably not.
- The same customer DMs you on Instagram three days later. New resolution. You pay again, even though it's the same person and the same question.
- A customer says "thanks", you say "you're welcome". Two resolutions? One? Whatever the vendor decides.
The deeper issue: per-resolution pricing turns every single reply into a margin decision. You start hesitating to engage. That is the opposite of what a sales tool should do.
This is why transparent AI pricing matters: with a flat fee you can read your AI cost per conversation off a calendar, not a meter. UptoNova is built as an AI sales agent for WhatsApp and the rest of your channels — a grounded AI chatbot that quotes prices and stock only from your live catalog, and a one-click human handover for anything it shouldn't answer alone.
The math at scale
| Chats per month | Fin estimated bill (at ~$0.99 each, 70% resolution rate) | UptoNova flat fee (approx.) |
|---|---|---|
| 1,000 | ~$700 | $49–$149 (Starter / Growth) |
| 5,000 | ~$3,500 | $149–$399 (Growth / Scale) |
| 10,000 | ~$7,000 | $399 (Scale) |
| 25,000 | ~$17,500 | $399 + volume add-on |
These numbers use Fin's published per-resolution price as a starting point. Real Intercom bills also include seat fees and channel add-ons. The exact dollar isn't the point — the slope is. Per-resolution scales with conversation volume. Flat pricing scales with how many businesses you run.
Head-to-head capability comparison
| What you need | Intercom Fin | UptoNova |
|---|---|---|
| Pricing model | Per resolution + seat fees | Flat monthly fee per business |
| Built for | Software-company support deflection | Multi-channel SMB sales |
| Channels covered | Web, email, WhatsApp, Instagram, Messenger (add-ons) | WhatsApp, Messenger, Instagram, Telegram, WhatsApp groups, web widget |
| Live product catalog | Limited — Fin is help-desk-shaped | Native product catalog and price list, with image search |
| Payment links inside the chat | Via external integration | Native — send a Stripe payment link in one tap |
| Customer sends photo "do you have this?" | No | Yes, via AI image search across your catalog |
| Help center as a CMS | Best in class | We're not trying to be a help center CMS |
| White-label for agencies | No | Native — branded workspaces per client |
Where Fin is genuinely better
To be balanced:
- Help center as a product. Intercom's authoring tools — article versioning, audience targeting, multi-language — are mature. If you maintain 500+ articles, that matters.
- Routing and workflow builder. The visual workflow tool for routing, escalation, and CSAT collection is more granular than UptoNova's automation engine.
- Enterprise procurement. SOC 2 Type 2, HIPAA, EU data residency — all checked. UptoNova is SOC 2 Type 1 with Type 2 in progress as of 2026.
- Ticketing. If your team thinks in tickets and SLAs, Intercom's whole data model fits naturally.
Where UptoNova wins for small businesses
- Selling is the main job. UptoNova's tools are sales tools — recommend a product, send a payment link, book an appointment, save a lead. Fin's tools are deflection tools.
- Predictable monthly cost. You can plan growth without modelling "what if our resolution rate drops?".
- One contact across channels. A person who messages you on WhatsApp, Instagram, and your website is one contact in UptoNova — Fin treats each channel as a separate record unless you stitch them together.
- AI sales coach for your team. Your human agents see suggested replies grounded in your catalog in real time. Not just AI-drafted ticket replies — actual next-best-action whispers. See the sales coach post.
- Generous engagement. You never have to think about whether a reply is "worth it" — flat pricing means engage every customer.
The honest decision table
| Your situation | Pick |
|---|---|
| B2B software company, web-first support, 500+ help center articles | Intercom + Fin |
| Online store, multi-channel inbound, conversational selling | UptoNova |
| Clinic / salon / service business with bookings and WhatsApp customers | UptoNova |
| Restaurant or café taking orders and reservations across DMs and WhatsApp | UptoNova |
| Enterprise support team with SLA and ticketing requirements | Intercom + Fin |
| Agency reselling AI chat to several SMB clients | UptoNova white-label |
What "resolution" should really mean for your business
If you want to actually measure how well your AI is doing, the right numbers are:
- Did it answer correctly? Was the price right, the stock right, the policy right? (You can review this in UptoNova's reply traces.)
- Did the chat end in a sale? Payment, booking, or qualified lead saved to the CRM.
- Did the AI hand off when it should? Anger, complaints, refunds, and edge cases get a human, not a bot.
Notice what's missing: "resolution count". That's a billing metric, not a business metric. See what a real AI sales agent looks like.
Frequently asked questions
Can I use Intercom Fin with WhatsApp?
Yes — since late 2024 Intercom supports WhatsApp as a channel. You still pay per Fin resolution on top of the conversation-based WhatsApp charges from Meta.
Is Fin grounded against my catalog?
It's grounded against your help center articles and connected sources. The grounding quality is good. The difference is what the AI can do: Fin is built around answering questions, UptoNova is built around taking action — recommending products, sending payment links, booking appointments, and saving leads.
How does Fin handle a customer asking for a price?
Fin needs a help center article that contains the price, or a data source you've connected. When you change a price, you also change the article. UptoNova reads price straight from your live product list, so there's no second source to keep in sync.
What about Zendesk Answer Bot or Salesforce Einstein?
Both have similar AI agents now, and similar per-resolution or seat-bundled pricing. The argument in this post applies to all of them equally.
What does moving from Intercom to UptoNova look like?
Export your contacts and conversation history (Intercom supports both), import to UptoNova, then reconnect WhatsApp through UptoNova. The bigger lift is mentally — your help center articles become "knowledge base entries" that the AI pulls from as needed, not a separate published site.