Per-Resolution AI Pricing: Run This Math Before You Sign

Intercom Fin currently charges $0.99 per resolved conversation (as of this writing). That sounds reasonable - you only pay when the AI actually closes a ticket. But run the math during a product launch or a bad bug week and that number looks very different. I built Trigli, which prices flat, so I have a bias here. I am going to name it and then show you the actual numbers anyway. You can decide what to do with them.

The short version

Per-resolution pricing is cheap on a quiet month and expensive exactly when you need AI the most - during a spike. When your product ships a bug, when you run a Black Friday sale, when you do a major launch, your support volume goes up. Under per-resolution pricing, your AI bill goes up with it. That is the trap. Not a scam. Just a model that punishes you on your worst days.

This post is not a Fin teardown. Fin is genuinely capable and I will say so clearly in a later section. This is a math exercise. Every founder should run it before signing up for any per-resolution tool, including ones that are not Fin. If you want to skip ahead to the numbers, the next two sections are the ones that matter.

What per-resolution pricing actually means

A resolution, in the context of Fin and similar tools, is a conversation the AI closes without a human touching it. The customer asks something, the AI answers, the ticket closes. That is a billable event at $0.99.

On the surface this sounds fair. You only pay when the AI succeeds. The vendor has skin in the game. If the AI is useless, you pay nothing. That alignment of incentives is the pitch, and it is not wrong.

The hidden assumption is that your volume is predictable and roughly flat month to month. Most SaaS and e-commerce teams I have talked to do not have flat volume. They have a baseline with spikes. That is where the model breaks down.

The math at three volume levels

I am going to use Fin's published rate of $0.99 per resolution, as of September 2026. Some Intercom-bundled plans include an initial monthly resolution allowance before per-resolution billing starts, so confirm your exact plan terms before running this math. Vendor pricing shifts, so check their pricing page before you sign anything. I am also going to assume a 65% AI resolution rate as a rough working estimate - though in practice I have seen teams report anywhere from 45% to 70% depending on how repetitive their volume is. Use your own number if you have data. The math structure is the same either way.

Scenario one: a quiet SaaS team or small e-commerce shop at 300 tickets a month. At 65% resolution, that is 195 resolutions billed, coming to roughly $193. That looks fine.

Scenario two: a mid-size team at normal growth, 1,500 tickets a month. At 65%, that is 975 resolutions, coming to roughly $965. That is a real line item.

Scenario three: a product launch, Black Friday, or a bug week at 5,500 tickets. At 65%, that is 3,575 resolutions, coming to roughly $3,540. That is close to what you would pay a part-time contractor for a month - and that bill arrives during the month your product is already under the most stress.

That last scenario is the one that should make you stop. The invoice is highest exactly when your team is most stretched. A bug ships on a Tuesday, tickets flood in, the AI handles 65% of them, and you get a $3,500 bill for the privilege. That is a bad time to be surprised.

Why spikes are the real test

Most teams I have talked to have two or three volume spikes per year. A product launch. A seasonal sale. An outage. Sometimes all three in the same quarter. Per-resolution pricing means your AI bill is correlated with your worst days, not your average days.

A flat-rate tool costs the same whether you get 200 tickets or 2,000 tickets that month. Trigli's Growth plan is $149/month. If a bug ships and you get 2,000 tickets in a week, the bill is still $149. That same workload under per-resolution pricing at a 65% resolution rate would run you roughly $1,287. Same ticket volume. Same AI resolution rate. Very different invoice.

I am not saying flat pricing is always better. I will get to when per-resolution actually makes sense. But if you have any history of volume spikes, you should run the spike math before you sign, not after.

The resolution definition problem

Here is a question most people do not ask: what exactly triggers a billable resolution? The answer varies by vendor and it changes your cost estimate in ways that add up fast. A 10% swing in what counts as a resolution on 1,500 tickets is roughly $150 a month. Over a year that is $1,800.

Some tools count a resolution when the ticket status changes to closed. Some require the customer to explicitly confirm satisfaction. Some use a timeout - if the customer does not reply within 24 hours, the conversation is marked resolved. The timeout model is the one to watch. A customer who did not get a satisfying answer and just gave up can still generate a billable resolution event.

This is not a gotcha. It is a legitimate question that any vendor should be able to answer clearly. Ask them: show me the exact trigger that creates a billable resolution event. If the answer is vague, that is a signal.

Where per-resolution pricing is actually fair

I want to be honest here because the model is not evil. If your volume is very low and genuinely unpredictable - say, under 100 tickets a month - paying per resolution can be cheaper than a flat monthly plan. At $0.99 per resolution and a 65% rate, 100 tickets costs you about $64. Trigli's Starter plan is $49/month. Those numbers are close enough that the model is not obviously wrong.

Early-stage teams who are not sure whether AI will work for their use case might also prefer to pay per success rather than commit to a flat fee. That logic is reasonable. You are hedging against the AI being useless.

That pricing structure also aligns vendor incentives with outcomes in theory. If the AI resolves more, the vendor earns more. That is not nothing. The failure mode is not the model itself - it is the combination of the model with unpredictable volume and a loose resolution definition. Those three things together are where teams get hurt.

Where Intercom Fin wins

I am not going to pretend Fin is bad. It handles multi-turn conversations well, its AI handoff logic is mature, and if you are already inside the Intercom ecosystem for sales and marketing, Fin fits without extra setup. No new tools, no new logins, no migration headache.

Intercom is a large, well-funded company with more than a decade of history. If you are an enterprise buyer who needs to know the vendor will be around in five years, that matters.

Fin covers channels that Trigli does not. Phone, SMS, Instagram DMs, Facebook Messenger - Trigli is email, chat widget, and built-in ticketing only. If your customers reach you on social or by phone, Fin is a more complete answer. Trigli does not have those channels and I am not going to pretend otherwise.

Fin also has per-agent productivity dashboards and skill-based routing that Trigli does not have yet. If you need to pull a report showing tickets resolved per agent per week, or if you need tickets to auto-assign to specific agents based on tagged expertise, Fin is ahead of us there. You can read a fuller breakdown at /alternatives/intercom and at /blog/intercom-fin-vs-zendesk-ai-vs-trigli if you want the side-by-side.

What Trigli does not do

Since I am asking you to consider Trigli as an alternative, you deserve the honest list of what it does not do.

• No phone or SMS support. Email (Gmail and Outlook via OAuth 2.0), chat widget, and built-in ticketing only.
• No per-agent productivity dashboards. You cannot pull a tickets-resolved-per-agent-per-week report right now.
• No skill-based agent routing. Trigli routes by ticket category to an action (auto-send, draft, human review, or skip). It does not auto-assign tickets to specific agents based on tagged expertise.
• No native Shopify order actions. The AI can reference your refund policy but cannot process a refund inside Shopify.
• No multi-brand support inside one account. One tenant, one brand.

If any of those are hard requirements, be honest with yourself - Trigli is probably not the right fit yet. I would rather you know that now than after you have spent two weeks setting things up.

How to run the math for your own team before you sign

This is a 20-minute spreadsheet exercise. Do it before you sign anything.

1. Pull your last 12 months of support ticket volume. Write down your average month and your peak month. Both numbers matter.
2. Estimate your AI resolution rate. Most teams I have talked to see 60-70% of volume as repetitive and AI-handleable, but be conservative - use 50-60% if you are not sure.
3. Multiply peak-month volume by your estimated resolution rate by the per-resolution fee. That is your worst-case monthly bill under per-resolution pricing.
4. Compare that number to a flat-rate plan at the same volume tier. The difference is what you are paying for predictability.
5. Ask the vendor for their exact resolution definition and whether there is a monthly cap on fees.

If your peak-month number is close to your average-month number, per-resolution might be fine. If your peak is two or three times your average, flat pricing is probably worth it. The math will tell you.

Questions to ask any AI support vendor before you sign

These are not trick questions. Any vendor worth working with should answer them clearly and quickly.

• What is the exact event that triggers a billable resolution? Is it a status change, a customer confirmation, or a timeout?
• Is there a monthly cap on resolution fees, or is it truly unbounded?
• What happens to my bill during an outage or a product launch spike?
• Can I set a resolution budget limit that pauses AI auto-sends if I hit it?
• Are there additional per-seat or per-agent fees on top of the per-resolution charge?

If a vendor gets defensive about any of these, that is useful information. The pricing model should be explainable in plain language.

How Trigli prices instead

Flat monthly plans. The Growth plan is $149/month. Pro is around $349/month. The bill does not change whether the AI resolves 50 conversations or 500 that month. You can see the full breakdown at /pricing.

I built it so you get email (Gmail and Outlook, both OAuth 2.0), the chat widget, and a real ticketing system - priority, assignment, internal notes, SLA tracking - all in one plan. No separate line items per product. No per-resolution charges.

Draft-and-approve is the default mode. The AI drafts, you approve before it sends. Auto-send is opt-in. If you are nervous about AI sending things without a human seeing them first, that is the right starting point. You can read more about how to set up automation without losing control at /blog/automate-customer-support-without-losing-human-touch.

There is a free tier at $0/month - 50 emails, 25 chats, 10 tickets. Paid plans come with a 14-day free trial. Card is collected at signup, charged on day 15 only if you have not cancelled. If you want to test whether AI support actually works for your volume before committing, that is the no-risk way to do it. More detail on what each tier includes is at /blog/cost-of-support-rep-vs-ai if you want to compare the cost against keeping a human rep.

The honest summary

Per-resolution pricing is not a scam. It is a business model with a specific failure mode: it costs the most when your team is under the most pressure. If your volume is low and flat, it might genuinely be cheaper than a flat plan. Run the math and you will know.

If you have seasonal spikes, launch cycles, or any history of outage-driven ticket surges, a flat plan removes the variable. You know what you are paying in January and you know what you are paying in November when the holiday rush hits.

The goal of this post is not to sell you Trigli. It is to make sure you have done the 20-minute math exercise before you sign anything. Whether you end up on Fin, Trigli, or something else entirely, you should know what your worst-case monthly bill looks like before you find out the hard way.