Flat-rate vs per-resolution AI support pricing: the real math

The two models, plainly

Per-resolution (also called outcome-based): you're charged a fixed fee each time the AI resolves a ticket without a human. Typical published rates in 2026: Intercom Fin $0.99, Gorgias Automate $0.90–$1.00 (with a $1.50 overage), and Zendesk AI around $1.50 (analyst estimate — Zendesk doesn't publish it). The appeal is obvious: no resolution, no charge.

Flat-rate: one monthly price for a volume band — say, up to 1,000 resolutions for $99. Go over the band and you either upgrade a tier or pay a small, disclosed overage. The appeal is a bill that doesn't move.

Worked example: a store at 1,000 resolutions/month

Say your AI handles 1,000 Tier-1 tickets in a month. Here's the AI fee each model charges, on top of your helpdesk subscription:

At this volume, flat-rate saves roughly $800/month versus $0.90/resolution — and more against Fin or Zendesk. Over a year that's about $9,600 back in your pocket.

Where's the break-even?

Per-resolution is genuinely cheaper at low volume, because you're not paying for a band you don't use. Against a $99 flat plan at $0.90/resolution, the crossover is around 110 resolutions a month ($99 ÷ $0.90 ≈ 110). Below that, the meter wins; above it, flat wins — and the further above, the wider the gap.

Flat figures use Glassdesk's published tiers ($49 / $99 / $149) as a concrete example; volume bands mean the "cheaper" column can shift a little at the edges of each band. Per-resolution column uses $0.90 for illustration — Fin and Zendesk are higher.

Cost isn't the only axis — watch for these

• Double-billing. Some helpdesks charge a ticket fee and an AI automation fee on the same resolved ticket, so the effective per-resolution cost is higher than the sticker rate. Read the billing docs.
• What counts as a "resolution." Definitions vary — a deflected chat, a sent email, an escalation. Make sure escalations to a human don't count against you.
• Overage rates. A friendly flat plan should disclose its overage (Glassdesk's is $0.15/resolution — a hard, visible cap, not a surprise). A per-resolution plan's overage can be higher than its base rate.
• Accuracy and auditability. A cheap wrong answer is expensive. Prefer tools that show citations and a confidence score, and that publish an accuracy eval, so you're not paying per resolution for guesses.

So which should you choose?

• Under ~100 resolutions a month, and you want the lowest possible floor → a per-resolution plan may edge it.
• Over ~150 a month, or you value a predictable bill during peak season → flat-rate wins on cost, usually by a wide margin.
• Either way → weight accuracy and transparency, not just the rate. The point of automating support is fewer bad answers, not just cheaper ones.

Glassdesk is flat by design — $49 / $99 / $149, no per-resolution meter — and every reply is cited and confidence-scored so you can audit it. See the full comparison against per-resolution agents, or try the savings calculator on the pricing page with your own numbers.