The Price Tag Nobody Wants to Talk About

AI support tools love talking about resolution rates and automation percentages. They're a lot quieter about what it costs.

So I did the work. I went through every major AI support tool's pricing page, talked to teams using them, and pulled the actual numbers. Not the "starting at" numbers. The real ones.

The Per-Resolution Breakdown

Per-resolution pricing is becoming the standard for AI support. The idea is simple: you pay when the AI successfully handles a customer question. No resolution, no charge. Sounds fair. The differences are in the details.

Look at that spread. The cheapest major competitor charges $0.90 per resolution (annual billing). The most popular one charges $0.99. That is 4 to 5 times more than a classification-first approach.

Why the Prices Are So Different

The expensive tools are doing more per resolution. Intercom's Fin reads your help center, generates a natural language answer, and carries on a multi-turn conversation. That's a large language model running on every interaction. LLM inference is expensive — OpenAI and Anthropic charge by the token, and a multi-turn support conversation racks up tokens fast.

A classification-first approach works differently. Instead of generating freeform answers, it classifies the customer's intent (billing question, password reset, refund request) and triggers a pre-set action. The heavy lifting is a lightweight ML model, not a billion-parameter LLM. That is why the cost is lower.

The trade-off: classification gives you speed and control but not freeform conversation. LLM-based tools give you flexibility but cost 4 to 5 times more and sometimes hallucinate answers.

For most small teams, classification handles 70% of incoming messages. The remaining 30% that need judgment go to a human anyway.

The Hidden Costs

The per-resolution price is only part of the story.

Intercom: The $0.99/resolution is on top of the base plan (Essential starts at $29/seat/month billed annually, $39 monthly). Each team member is a paid seat. A 3-person team handling 300 AI resolutions/month pays $87 + $297 = $384/month minimum (annual billing).

Zendesk: AI Agent pricing sits on top of the Suite plans ($55 to $115/agent/month). You need the Suite plan to access AI features. A 3-agent team on Suite Team with 300 AI resolutions at $1.50 each pays $165 + $450 = $615/month.

Gorgias: Their AI Agent is available as an add-on on any plan. On the Pro plan at $360/month (or $300/month annual), add 200 AI resolutions at $0.90 (annual billing) and you hit $540/month. It's built for Shopify stores specifically, so if you're not on Shopify, it's not for you.

Supp: No base fee. 300 resolutions with actions at $0.30 each = $90/month. Every team member gets dashboard access. No seat charges.

When Expensive Tools Make Sense

I'm not going to pretend the cheaper option is always better. Here is when paying $0.99/resolution is worth it:

You have a massive knowledge base. If you have 500+ help articles and customers ask complex, varied questions, an LLM-based tool like Fin can generate answers you'd never be able to pre-build rules for.

You need multi-turn conversation. If your support involves back-and-forth clarification ("Which plan are you on?" "When did you see this error?"), an LLM handles that naturally. Classification works best when the first message tells you enough.

Your volume is very low. If you get 20 support messages a month, the per-resolution cost barely matters. The base subscription fee matters more, and Intercom's $29/seat/month gets you a full messenger platform, not just AI.

When Cheaper Tools Win

Your volume is high. At 1,000 resolutions/month, the difference between $0.20 and $0.99 is $790/month. That is $9,480/year. For a startup, that buys a lot of runway.

Your questions are predictable. Password resets, billing inquiries, order status, refund requests — these don't need a conversational AI. They need fast, accurate classification and the right action. Pay $0.20 for that, not $0.99.

You want control over responses. With classification, you write the response templates. You know exactly what customers see. With LLM-generated answers, you're trusting the model to not hallucinate your refund policy.

You're scaling fast and costs compound. At 2,000 resolutions/month, you're saving $1,580/month — that's $19,000/year. At that point the savings fund a part-time hire.

The Math for Your Team

Here is a quick formula:

Monthly AI support cost = (monthly messages × automation rate × per-resolution price) + base fee

Run it for your numbers:

• 200 messages/month, 70% automation, Intercom: (200 × 0.7 × $0.99) + $29 = $167.60/month
• 200 messages/month, 70% automation, Supp: (200 × 0.7 × $0.25) + $0 = $35/month

• 500 messages/month, 70% automation, Intercom: (500 × 0.7 × $0.99) + $29 = $375.50/month
• 500 messages/month, 70% automation, Supp: (500 × 0.7 × $0.25) + $0 = $87.50/month

• 1,000 messages/month, 70% automation, Intercom: (1,000 × 0.7 × $0.99) + $29 = $722/month
• 1,000 messages/month, 70% automation, Supp: (1,000 × 0.7 × $0.25) + $0 = $175/month

The gap widens as volume grows.

Bottom Line

If you need a full-featured messenger platform with AI on top, Intercom or Zendesk will serve you well. You'll pay for it, but you'll get a polished product.

If you need fast, accurate support automation and want to keep costs proportional to usage, a classification-first tool at $0.20 to $0.30/resolution will save you thousands per year.

The right choice depends on your volume, your question complexity, and how much you want to spend. But at least now you have the real numbers to decide.