Why AI customer service pricing is hidden

If you've shopped for an AI customer service platform, you've hit the wall: you search for pricing, find a range like "$55–$115 per agent per month," and click to the add-on section—only to see "AI features: Contact sales." This isn't laziness. It's deliberate opacity, and it maps directly to how AI vendors bill.

The core reason: per-resolution and per-interaction pricing can't be fixed

Traditional customer service software—Zendesk, Freshdesk, Help Scout—charges per seat or per ticket. The cost per agent doesn't change when your team gets better at resolving tickets. Your bill stays predictable.

AI platforms introduced a new model: pay per resolution or per interaction. HubSpot Breeze charges $0.50 per resolved ticket. Freshdesk AI costs $49 per 100 sessions (~$0.49 each). Kustomer bots run about $0.60 per conversation. Intercom Fin is $0.99 per resolution. Gorgias hovers around $1.00 per resolution.

Here's the trap: if your AI gets better at resolving tickets, your costs go up. A team that answers 100 tickets per day at 40% AI resolution pays $20. If your AI improves to 60% resolution, you pay $30—and the vendor is happy. This creates a perverse incentive: the vendor can't publish a price because the price depends on something you don't control at purchase time: your AI's performance.

Some platforms bill per interaction or per conversation instead, charging whether the AI succeeds or fails. Ada charges per interaction, even on failure. Salesforce charges $2 for a 24-hour session. This is even harder to quote because you don't know upfront how many interactions you'll have.

How the big platforms hide pricing

Salesforce (Service Cloud) doesn't publish AI agent pricing at all. It publishes seats at $165–$330/month per agent, but bots are gated behind enterprise contracts. You'll need to call sales and sign a multi-year deal; minimum contracts start around $30k–$100k per year.

Zendesk publishes seat pricing ($55–$115 per agent per month) and mentions "AI resolution rate" in product docs, but the cost per resolution stays hidden. Zendesk publishes expected resolution rates (how many tickets their AI solves), but not the price you pay per ticket resolved. Zendesk charges on a committed model (you guarantee X resolutions per month) or a pay-as-you-go model where overages spike—often doubling your bill. A mid-market customer reported paying $2.00 per resolution on overage vs. $1.50 on commitment.

Ada, Sierra, Decagon, PolyAI, Air AI publish no pricing at all, not even ranges. They position themselves as "enterprise only" and require a demo call before they'll even discuss cost. Ada charges per interaction; Sierra, per resolution (but won't disclose the rate). Decagon hides everything. PolyAI, likewise. Air AI bills on ring time and custom deals.

Intercom (Fin) publishes $0.99 per resolution upfront, making it one of the few transparent AI vendors. Likewise, Help Scout publishes $0.75 per resolution. Gorgias publishes around $1.00. These are exceptions.

What hidden pricing signals (and what to demand)

When a vendor won't publish AI pricing, look for these red flags:

• Annual minimums or seat floors. Kustomer requires an 8-seat annual minimum. Salesforce and Zendesk for enterprise can lock in $30k–$400k per year. Ask upfront: what's the smallest contract you'll offer?
• Committed vs. overage pricing. Zendesk's overage model can double your bill. Ask for a pilot period with a fixed monthly cap—e.g., "I'll pay $X per month for the first 3 months, no overage charges."
• AI resolution rate assumptions. Vendors won't tell you their average resolution rate because it varies wildly by use case. A billing chatbot resolves 80% of tickets; a claims inquiry bot resolves 20%. Before you sign, ask: "What resolution rate do you assume for my use case, and what would my monthly bill be at 30%, 50%, and 70% resolution?"
• Hidden per-interaction or per-session fees. Ada bills every interaction, even failures. Air AI bills ring time, not just completed calls. Ask: "What happens if my AI gets a question it can't answer? Do I still get charged?"

How to get a real quote without a sales call

Use our AI support cost calculator to estimate your monthly bill based on ticket volume and assumed AI resolution rate. Plug in your numbers (e.g., 500 tickets/month, 50% AI resolution) and see what each vendor would charge. This gives you a baseline to ask for in a sales call.

When you do talk to sales, ask for three quotes:

• Pilot pricing (3-month fixed monthly cost, no overage, covers up to X interactions).
• Standard pricing (annual commitment, what's the all-in cost at 30%, 50%, 70% resolution?).
• A cap on monthly overage (e.g., "Bill me for the first $X in overages per month, then it stops").

Transparent vendors like Intercom, Help Scout, and Crisp publish flat rates or per-seat costs, so you can skip this dance—but you'll still want to model ticket volume to estimate your actual spend.

Which platforms publish pricing upfront

If opacity bothers you, these vendors make it easy:

• Help Scout: $25–$75/user or $0.75 per resolution for AI.
• Intercom: $29–$132/seat or $0.99 per resolution (Fin).
• Gorgias: ~$1.00 per resolution; watch for double-billing (ticket fees + automation fees).
• Crisp: $45–$295/workspace, flat monthly, includes basic AI.
• Kommunicate: $40–$200/month, all-in, no hidden AI charges.
• Tidio: free tier + Lyro AI from $32.50/month flat.

These don't necessarily cost less—but you can budget them without a sales call.

How to get a real number

Hidden pricing isn't a bug in the AI customer service market; it's baked into per-resolution and per-interaction models. When your bill scales with your AI's performance, vendors can't publish a number—and they benefit from keeping you guessing until you've invested time in a demo.

To stay in control: use our cost calculator to model your own assumptions, ask sales for pilot pricing with a monthly cap, and gravitate toward transparent vendors when your use case fits. Check our vendor comparison or use the vendor finder to filter by pricing transparency if that matters to your team.