AI customer support with no per-resolution fee

The short answer

A per-resolution fee is a charge that fires every time the AI closes a conversation on its own. It sits on top of whatever you already pay per seat, so your software bill tracks your ticket volume rather than your headcount. CustomerEagle does not work that way. Every paid seat carries a monthly allowance of AI resolutions, and that allowance is part of the seat price: 50 resolutions per agent per month on Standard, 150 on Professional, 350 on Premium. A three-agent team on Professional therefore starts each month with 450 AI resolutions already paid for.

This is not a claim that AI volume is free. Resolutions beyond the allowance are metered at €0.99, €0.56 or €0.49 each, depending on the plan. The difference is where the number lives: the first several hundred resolutions of the month are not a separate line item, and the rate for everything after them is published on the pricing page instead of appearing on an invoice you have not seen yet.

The Free plan is the exception worth naming: it has no recurring allowance at all. It ships with a one-time grant of 25 AI resolutions so you can test the assistant, and the definition behind it, before you pay for anything.

What a per-resolution fee is, and who charges one

The rate is the part everyone compares and the part that tells you least. What decides the invoice is the unit — which event trips the counter. Some meters fire on a resolution, some on any conversation the AI touched, some on a fixed-length session that can contain several exchanges. A lower headline rate on a counter that trips more often produces a bigger bill at identical performance.

Here is what the vendors we track publish, in their own units and their own currency:

Why vendors moved to it

The move is easy to explain, and worth understanding before you judge it. Seat-based pricing charges for capacity: configured users, whether the queue is busy or empty. An AI agent that closes a large share of that queue reduces the capacity a team needs — which makes seat count a shrinking base to price against. Putting the AI on its own meter restores the link between what the vendor charges and what the software actually did.

There is a genuine argument for that. You pay for outcomes rather than for logins, and a quiet quarter costs less. The price of the symmetry is that the bill also rises exactly when you are busiest. In e-commerce, November and December are when a variable line item does the most damage, and they are the months you can least afford a surprise on.

The second cost is definitional. A per-resolution meter is only ever as fair as the definition of a resolution — and that definition is written by the party sending the invoice. This is why the units in the table above matter more than the rates. A session counter, a conversation counter and a resolution counter turn the same month of work into three different invoices.

What it costs at real volume

The arithmetic is simple enough to do on the back of an envelope, and most buyers never do it, because the fee is quoted as a rate rather than as a monthly figure.

An automated resolution rate is the share of conversations the AI closes without a human agent replying — the number some vendors call deflection. The 60% used below is a plausible figure for a team with a well-documented knowledge base, not a promise; substitute your own, because it moves the answer more than the rate does.

The included-allowance model instead

The alternative is to attach the AI allowance to the thing you are already buying. CustomerEagle prices per agent per month, and each paid seat carries its own monthly allowance of AI resolutions. The effective allowance is the per-seat number multiplied by your paid seats, so it grows when you hire instead of being renegotiated.

Two things follow from that shape. First, the number of resolutions you get is tied to team size — which is the number you can actually forecast. You know roughly how many people you will hire next quarter; you do not know how many tickets November will bring. Second, the overage rate falls as the plan rises, so choosing the plan that matches your normal month is a real decision rather than a formality. The worked example below shows how large that difference gets.

How we count a resolution, and the settling window

None of the above is worth anything if the definition of a resolution is loose — an allowance is only generous relative to what depletes it. Ours is deliberately strict, and deliberately biased against us:

• A conversation only counts as AI-resolved when no human agent sent a single reply in it. We verify that against the actual author of every message in the conversation; if your team touched it, it is a human resolution and is never billed as an AI one.
• A conversation the customer simply abandoned is closed by timeout and never counted. Silence is not success.
• Every resolution carries a 3-day settling window. If the customer reopens the conversation, a human agent takes over, or the customer leaves a negative rating within those 3 days, the resolution is automatically reversed — removed from your usage and, if it was already invoiced, credited back.

Every counted resolution is visible in your dashboard with its full conversation history, so you can audit any of them against that definition. Read the full definition: how we measure AI resolutions.

Cost comparison for a 3-agent team

One set of assumptions, held constant: three agents, 2,000 conversations in a normal month, and a 60% automated resolution rate — 1,200 AI resolutions.

Premium costs least at this volume, and that is worth stating plainly because it is the opposite of an upsell being disguised: its higher seat price buys 1,050 included resolutions and a €0.49 rate, which at 1,200 resolutions beats the cheaper Professional seat outright. Standard — the lowest seat price of the three — produces the largest bill, because its €0.99 rate applies to most of the 1,200.

For the same month, a per-outcome meter at $0.99 charges $1,188 for the AI alone, before a single seat is paid for. We are not converting that into euros and not declaring a winner: the currencies differ, the products differ, and a comparison that ignored both would be exactly the kind of marketing this page argues against. The point is the shape of the number. One of them is bounded by a plan you chose; the other is bounded by how busy November turns out to be.

When per-resolution pricing is actually cheaper

There are real cases where the meter wins. Pretending otherwise would undercut everything above it.

• Very low volume. If the AI closes a handful of conversations a month, a pure meter costs a handful of times the rate, while a seat-plus-allowance model still bills the seat. At the very bottom end, Tidio’s free plan is hard to beat, and we say so on our own comparison page.
• Sharply seasonal volume with a long quiet season. A meter falls to almost nothing in a dead quarter; a seat price does not move.
• Many seats, almost no AI. An allowance you never touch is capacity you paid for. If your queue is mostly complex human work, the allowance is not the thing you are buying.
• A genuinely lower entry seat. Freshdesk’s Growth tier at $19 per agent per month billed annually is lower than our cheapest paid seat. Whether it wins overall depends on your AI volume, because Freddy sessions are metered separately once the first 500 per account are used.

The honest test is a single number: resolutions per agent per month. Below the included allowance, an allowance model is strictly cheaper than paying per unit, because the marginal resolution costs nothing. Well above it, the comparison collapses into a rate comparison, and which plan you pick matters more than which vendor you pick.

We modelled twelve months of both shapes, peak season included: Per-agent vs per-resolution pricing, modelled over 12 months.