How much does an AI customer service agent cost? Less than the quote, and more than the price

The quote and the invoice are two different documents

Every buyer in this category has the same experience. You are quoted a friendly monthly number. Four months later the invoice is meaningfully larger, and nobody has done anything wrong — the extra is channel fees, overage, an AI charge sitting alongside the seat charge, and a helpdesk module that turned out to be a separate product. Nothing was hidden, exactly. It simply was not on the page you made the decision from.

This happens because software pricing pages are written to be compared favourably, and total cost of ownership cannot be. So the industry standardised on publishing the part that looks best: the subscription. Everything else is discoverable, but only if you know to ask, which is a peculiar standard to hold buyers to.

There is a more useful frame for all of this. Before comparing any price, price the thing you already own: an inbox that answers late. That is what the buyer’s guide to AI agents for business opens with, and it changes the question from “is this expensive?” to “is this cheaper than what I am already losing?” — which is the only version of the question with a defensible answer.

The four pricing models, and what each one does to your behaviour

These are not four prices for one product. They are four different bets about how your business will grow, and you are taking one of them whether or not you notice. Read the last column of each — the question is worth more than the price.

Per-seat pricing deserves one extra sentence, because it is the model most small teams end up in by default and the one that fits them worst. You are buying software whose entire purpose is to stop your headcount scaling with your volume — and being charged by headcount. When vendors then layer an AI resolution charge on top of the seat charge, you have both problems at once: the model punishes you for growing and for succeeding. If you are weighing an agent against the helpdesk you already pay for, the fuller comparison is in AI agent vs live chat vs helpdesk.

What vendors actually charge

Publicly reported prices, gathered so you can see the shape of the market in one place. Every figure needs confirming on the vendor’s own page before you budget from it — this category re-prices constantly, and several of these vendors have changed model, not just number, within a single year.

Figures as publicly reported at the time of writing, presented as directional rather than as quotes. Verify on vendor pricing pages before purchase. Jugl figures are published list tiers.

Two patterns are worth naming. First, the enterprise platforms charge twice — a seat price and a resolution price — which is how a $79 quote becomes a four-figure invoice at volume. Second, every WhatsApp platform on the market has a second bill behind it that has nothing to do with the vendor, because Meta charges per conversation independently. If that is your main channel, read the WhatsApp AI platform comparison before you compare subscriptions, and ask every vendor the same blunt question: are Meta’s fees passed through at cost, or marked up?

We have modelled several of these head to head at real volumes: Jugl vs Intercom Fin on per-resolution billing, Jugl vs Zendesk on seats plus resolutions, Jugl vs Freshdesk Freddy, Jugl vs Gorgias for Shopify stores, Jugl vs Decagon, and Jugl vs Meta Business Agent on what a token meter really costs when the software is nominally free.

How each model behaves as you grow

Here is the part that decides whether you are happy in month twelve. Two of the three plottable models are a function of how well the software works — so the better your agent gets, the more you pay for it. Drag the resolution rate and watch.

The trap is subtle and almost nobody sees it coming, because it does not arrive as a bill shock. It arrives as a series of sensible-sounding decisions. Let’s not switch on email yet. Let’s keep the agent to FAQs rather than sales conversations. Let’s cap it at 60% and review next quarter. Each one protects the invoice. Together they turn an AI agent back into a chatbot, at agent prices — and eighteen months later the company concludes the technology did not work for them.

The costs that are not on the pricing page

None of these are anybody’s dirty secret. They are simply the parts of the purchase that live outside the subscription, and they are where first-year budgets break — not because the numbers are enormous, but because they were never numbers at all until the invoice arrived.

Turn that into something you can act on. Take the quote you are closest to accepting, and tick only what you have in writing. What is left is not a scare number — it is your homework, and an hour of emails usually resolves most of it.

One item on that list deserves singling out, because it is the one buyers most often skip and most often regret: knowledge base preparation. It is ten to forty hours of a real person’s time, it never appears on a pricing page, and it is the single biggest determinant of whether the agent turns out to be excellent or embarrassing. Spend the hours deliberately — our guide to AI agents for customer support covers what to write and in what order.

A worked example, with both sides of the ledger

An e-commerce business handling 5,000 customer conversations a month, with three support staff at a $4,000 loaded monthly cost each, and a $60 average order value. Here is what the same year looks like under each model.

Now the side of the ledger almost nobody models

Deflection savings. 3,000 conversations resolved without a human, at six minutes each, at a $25 loaded hourly cost, is $7,500 a month of recovered capacity. That is the number every vendor puts in a deck, and it is the smaller one.

After-hours capture. If 30% of those conversations arrive outside business hours — a conservative figure for most consumer businesses — that is 1,500 conversations. If just 5% of them would convert when answered immediately, at a $60 average order value, that is another $4,500 a month of revenue that currently goes to whoever replied first.

Note what the second term implies. If you are a consumer business, this is not a support purchase with a cost-saving justification. It is a revenue purchase that happens to also save support cost — which is why the concierge framing tends to produce better buying decisions than the ticket-deflection framing.

The ROI term everybody forgets

Almost every ROI model in this category counts one thing: tickets deflected × time saved × hourly cost. It is a real number and it is the wrong headline, because it measures the value of not doing work rather than the value of work that now gets done.

The honest formula has two terms:

The second term is larger than the first for most consumer businesses, and it is invisible in every support-centric ROI calculator on the internet — including most of the ones built by vendors, which is a strange piece of self-harm on their part. If you sell anything at all through conversations, model it before you decide this is expensive.

Seven questions to ask before you sign

Send these as one email. How fast each answer comes back tells you as much as the answers do — and every one of them is worth more than a discount, because discounts expire on a schedule the vendor controls while definitions govern every invoice you will ever receive.

What Jugl costs, and why it is shaped that way

We have spent this page telling you how to interrogate anyone’s pricing, including ours. So here is ours, plainly, and the reasoning behind the shape rather than the number.

Free, $31, $119 and $390 a month. Published, on a public page, with the AI included at every tier and nothing metered per message, per token or per resolution. That last clause is the whole design, and it exists because of the behaviour described further up this page: the moment improvement costs money, teams stop improving. We would rather your agent handled more of your business next quarter than fewer conversations to protect a meter.

The honest limits

A flat tier is not automatically the cheapest thing you can buy. If your volume is tiny, the free tier is the right answer and we would rather you stayed on it than bought something you do not need. If your volume is violently seasonal, per-resolution genuinely can beat paying for peak capacity all year. And above the Business tier we quote, like everybody else — the difference is that we say so here rather than letting you find out at the end of a sales process.

What we will claim is this: for a business doing somewhere between a few hundred and a few thousand conversations a month across more than one channel, a flat tier with the AI included is both cheaper and — more importantly — forecastable. You can put it in a spreadsheet in January and it is still true in November, which is not a small feature when the alternative is a line item that moves every time your agent has a good month.

The third number is the difficult one. Deflection is cheap to buy at the cost of satisfaction — a bot that irritates people into closing the tab looks magnificent on a dashboard. Holding 94% satisfaction while removing 73% of ticket load is the combination that takes engineering. On revenue, customers report +31% checkout recovery, $16,000 recovered in a single quarter and 2.6× ROI in 60 days — the detail is in customer stories, and the full product picture is in what is Jugl, including who should walk away.

What another quarter of deliberating actually costs

We are not going to invent scarcity. There is no expiring discount at the bottom of this page. But there are three real reasons that waiting is more expensive than it looks, and all three are arithmetic rather than persuasion.

The third one is why we would rather you started on a free tier this week than ran a procurement process this quarter. Not because the paperwork is wrong, but because an agent that has read a month of your real conversations is a materially different product from one that has read none — and the month you spend deciding is a month of that data you do not get to keep.

Pricing questions buyers ask