Six Help Desks, One Fight Over Who Pays for the AI

Buy a help desk in 2026 and you will sit through six demos that look almost identical. A ticket comes in. An AI drafts a reply. An SLA clock turns red. A dashboard fills with green. The features have converged so completely that the software is no longer where these companies compete. The pricing page is. And the single line that decides your bill is one most buyers skim past: when the AI resolves a ticket, does the meter run?

SolarWinds, Supportbench, Intercom, TOPdesk, Vision Helpdesk and BMC all say they sell "help desk software." They are chasing the same ticket from six very different directions. Three of them grew up in IT departments and think in assets, incidents and change requests. Three of them grew up in inboxes and think in conversations. Put them on one comparison spreadsheet and you will compare a facilities work-order tool against an in-app chat widget - and wonder why none of the columns line up.

So before a single feature checkbox, the useful move is to split the six into the two jobs they actually do.

The two planets hiding in one category

On the first planet sits IT service management. SolarWinds Service Desk folds ticketing, IT asset management, a CMDB and a service catalog into one screen, so an agent can close a laptop request and see the laptop in the same click. TOPdesk, out of Delft in the Netherlands since 1997, runs a single service desk across IT, facilities and HR - the same tool that resets your password can log that the third-floor air conditioning is broken. BMC Helix is the heavyweight, a cloud-native, AI-driven platform built for large organizations with complicated estates, competing in the same weight class as ServiceNow.

On the second planet sits customer conversation. Intercom is messaging-first, built to answer your app's users over live chat, now fronted by its Fin AI agent. Supportbench is built for B2B support, where a "customer" is a company with a hierarchy, a contract and a set of SLAs, not a single email address. Vision Helpdesk, founded in Nashik, India in 2007, made its name on multi-brand support - one team quietly running help desks for many companies from a single console.

Pick the wrong planet and no discount saves you. An ITSM platform will fight a scrappy support team with process it does not need; a messaging tool will buckle the first time an enterprise IT department asks it about change management. Get the planet right, shortlist two, and only then argue about features.

Where the real money hides

Now the part the demos rush past. AI turned every one of these tools into an "AI-first" platform, but it did not make any of them cheaper. It made the pricing harder to read. There are now three billing models wearing identical interfaces, and the difference between them is the difference between a flat bill and a variable one.

Supportbench is the clearest example of AI baked into the seat. Its pitch is almost entirely about what it does not charge extra for: AI, dynamic multi-level SLAs, escalations, the knowledge base, reporting and onboarding all sit inside a per-agent price starting at $32 a month, with no per-resolution fee. You can multiply agents by dollars and know your bill.

Intercom is the clearest example of the opposite. A seat starts around $29 a month, but the Fin AI agent bills roughly $0.99 for every resolution on top of that. When Fin works well, it works often - and the invoice climbs with your success. In 2026 Intercom leaned so far into this that it began reshaping its identity around Fin itself, a rare case of a 15-year-old company handing its name to its own AI agent. Fin has been reported north of $100 million in annual recurring revenue. Per-outcome pricing is not a footnote for them. It is the business.

Between those poles sit the ITSM players, where AI and premium modules tend to be add-ons layered onto enterprise contracts, and the price you pay is the price you negotiate. The lesson is the same across all three models: the sticker is the marketing, and the meter is the math.

Read the chart the way a buyer should: not left to right by height, but by color. The two green bars are the only ones where the number in front of you is close to the number you will pay. Everything else is a starting point in a longer negotiation with usage.

The question that ends every demo

There is a single question that cuts through all of it, and you can ask it on any call: show me the invoice at forty agents with AI turned on. Not the starting seat. Not the feature grid. The real number, at your real headcount, with the AI doing the work you are buying it to do. Watch how quickly the room gets specific.

Because the help desk you can afford at five agents is not always the one you can afford at fifty. Metered AI is wonderful when volume is low and punishing when volume is exactly what you hoped for. Included AI costs more on day one and stops surprising you after that. Neither is wrong. But one of them matches how your team will actually grow, and the only way to know which is to model it before you sign, not after.

A field note on age and identity

Two small facts frame how strange this market has become. BMC Software was founded in 1980, in Houston, by three former Shell employees whose surname initials - Boulette, Moores, Cloer - became the name. It has sold service management since before the public internet existed. And in 2026 it competes for attention with a Canadian startup that gives away the AI for free and a fifteen-year-old messaging company that has started answering to the name of its own bot. The category did not get older. It got weirder, and the weirdness lives almost entirely on the pricing page.

If you take one thing into your next eval, make it this: decide your planet, shortlist two, and compare on total cost at your real scale with AI switched on. The flashiest demo rarely wins the year. The most predictable invoice usually does.