Why does per-resolution pricing bother small teams so much?
A per-resolution charge puts a meter on the one thing you were trying to increase. Every question the AI handles on its own becomes a line item, so the better the automation works, the larger the invoice. Owners describe the same reflex: they start narrowing where the assistant appears, pulling it off busy pages, or switching it off during a promotion, because a good week on the marketing side turns into an unplanned bill on the support side. That is the opposite of what you bought it for.
The second problem is that you cannot budget it. Support volume for a small business is not a smooth line. A shipping delay, a product recall, a viral post, a holiday rush, a billing run that confuses people: any of these can multiply conversations for a few days. With seat pricing you know what you owe before the month starts. With resolution pricing you find out afterwards, and the spike arrives at exactly the moment your team is least able to think about procurement.
The third problem is definitional. Vendors count resolutions differently. Some count a conversation where the customer never asked for a human. Some count any conversation the AI participated in. Some count a session that ended without an agent reply inside a time window, which means a customer who wandered off mid-chat can register as a billable success. None of this is dishonest, but it does mean you are now auditing somebody else's counter, and reconciling their definition against your own sense of what actually got solved. That is unpaid work for a support lead who already has plenty.
What should you compare besides the per-resolution line?
The resolution meter is the headline, but it is rarely the only variable cost. Work through this list for every vendor on your shortlist, because two tools with similar sticker prices can land far apart once you add the pieces you actually need.
Seats. Count everyone who will ever need to read a ticket, not just full-time agents. Store managers, warehouse staff, a bookkeeper who answers invoice questions, a founder who likes to reply personally at night. Per-seat pricing quietly discourages you from giving those people access, which pushes work back through one overloaded inbox.
Contacts or monthly active users. Several helpdesks have moved away from pure seat pricing toward billing on the number of people you talk to. That can be kinder for a large team with modest volume and harsher for a small team with a big audience, so check which side of that trade you are on.
Channels. Ask which channels are included and which are add-ons. Voice, WhatsApp, and SMS are the usual extras, and they are often the ones that matter most to a local service business. Corebee includes web chat, email, voice, WhatsApp, Slack, Facebook, Instagram, SMS, and Telegram in the same flat $99 per month, with no per-seat and no per-resolution component, which is the shape worth comparing the rest against on pricing.
Overage behavior. When you hit a cap, does the AI stop answering, or does it keep answering and keep charging? Both are defensible. Only one of them is a surprise.
Knowledge sources. Some plans limit how many articles, URLs, or documents the AI can learn from, which is a soft cap on quality rather than cost. If your answers live in a messy mix of site pages and internal notes, test whether the tool can read all of it before you commit.
Which Zendesk alternatives actually avoid the resolution meter?
There are three honest answers, and they suit different businesses.
Flat-rate platforms that bundle AI. Here the AI conversations are simply part of the subscription. The appeal is arithmetic you can do in your head: one number, regardless of volume and regardless of how many teammates you add. This is the category Corebee sits in, with unlimited seats and unlimited AI conversations, answers drawn from your website and knowledge base, handoff to a human whenever the customer asks or the AI is unsure, meeting booking through Cal.com or Calendly, and sync to HubSpot, Salesforce, Pipedrive, monday, and Zoho. Flat rate is a bet by the vendor that average usage stays reasonable, which is why you should still read the fair-use language rather than assuming the word unlimited settles everything.
Open source you host yourself. Chatwoot is the best-known option. The self-hosted edition is free to run, so there is no resolution meter by construction, and you keep the data on your own infrastructure. The cost moves rather than disappears: you are paying for servers, upgrades, monitoring, and somebody's attention when it breaks at nine on a Friday night. Its AI assistant features are a paid layer in the hosted product, so check how those are metered before you plan around them. If you have an engineer who likes this kind of thing, Chatwoot is a genuinely good answer. If you do not, it is a hobby you did not ask for.
Seat or contact pricing with a flat AI add-on. Help Scout, Freshdesk, and Tidio all sit in this general territory, though each has moved its AI billing around, and the details differ by plan and by region. Help Scout bills on the people you help rather than the seats you fill. Freshdesk packages its AI features into tiers and add-ons. Tidio's AI agent is tied to the number of conversations it handles, which is closer to a meter than a flat fee. Intercom's Fin is the clearest example of outcome pricing, charging per resolution, and Gorgias prices its automation in tiers of automated interactions, which is a volume meter with a different label. None of these is a bad product. They are priced for different buyers, and some of those buyers genuinely prefer paying only when the AI works. Pricing pages change more often than blog posts do, so confirm the current model on the vendor's own page before you decide, and treat any comparison table, including ours at Zendesk alternatives, as a starting point for that check.
Is a flat fee ever the wrong choice?
Yes, and it is worth being clear about when. If you are a one-person business with a handful of conversations a week, a metered plan can cost less than any flat subscription, because you are paying for a trickle instead of a capacity. The same is true if your support is highly seasonal and genuinely quiet for most of the year.
The way to settle it without guesswork is to do the division yourself. Take the flat monthly fee, divide it by the number of conversations you honestly expect the AI to handle in a month, and compare that figure to the per-resolution price on the metered vendor's pricing page. Then run the same sum again with your worst month rather than your average month, since the worst month is when you will care. Flat pricing tends to win on three specific patterns: volume that spikes unpredictably, volume that is growing, and teams where a lot of people need a login even though only a few of them work tickets all day.
There is also a softer factor that does not show up in a spreadsheet. With a flat fee, nobody on your team has to think about cost while deciding where to put the assistant or how widely to let it answer. That removes a small, constant tax on experimentation, and in practice it is often the reason flat-rate customers end up automating more than metered ones.
How do you move off Zendesk without dropping tickets?
Do not cut over everything at once. Start by exporting what you already have: help center articles, macros, canned replies, and a list of your most common questions pulled from the last few months of tickets. That list is the real asset. It is the test set you will use to judge whether any new tool actually answers your customers correctly.
Then pick one channel and run it in parallel. Web chat is the usual first choice, because it is new traffic rather than migrated traffic, so a mistake does not strand an existing thread. Point the AI at your site and your knowledge base, feed it your question list, and read the answers critically. Pay particular attention to the handoff: what the AI does when it does not know something matters more than what it does when it does. A clean escalation into a shared inbox, with the full conversation attached, is the difference between automation that helps your team and automation that creates a second queue of cleanup work.
Move email last, since changing forwarding rules is the step that can actually lose messages. Keep both systems live until a full cycle of your busiest week has passed on the new one. Setup time varies by tool and by how tidy your content is; Corebee's own onboarding takes roughly ten minutes and needs no developer, which makes the parallel-running approach cheap enough that there is little reason to skip it. If a vendor's setup requires engineering time you do not have, that is a cost worth counting alongside the subscription.
What should you ask a vendor before you sign anything?
Send these in writing and keep the reply. A clear answer is informative, and a vague answer is informative too.
What exactly counts as a billable resolution, in your definition, and can I see the raw log behind the count? What happens when I reach a cap: does the AI stop, queue, or keep billing? Are seats unlimited, and does that include read-only and occasional users? Which channels are included at my plan level and which are priced separately? If I add voice, how is that billed, and does the assistant tell callers it is an AI? Can I export my conversations, contacts, and articles, in a usable format, without a support ticket? What is the notice period, and does the price hold if my volume doubles?
That last question is the one that separates pricing models in practice. A vendor whose revenue grows with your automation has every reason to be delighted when your volume doubles. A vendor on a flat fee has to have built for it. Both can serve you well, but you should know which conversation you are having before you put a card down.