Custom AI Chatbot vs Platform (Intercom, HubSpot): Real Pricing

Every chatbot platform prices differently on purpose. Per resolution, per seat, per conversation, plus AI credits on top: the model is built so the headline number undersells your real bill. A custom build has no per-conversation tax at all, but it charges a five-figure entry fee instead. Which one wins comes down to one number: how many conversations you actually handle.

TL;DR

• Platforms price three ways in 2026: Intercom Fin charges ~$0.99 per resolution on top of seats; HubSpot charges ~$90/seat plus per-resolution AI fees; older flat-seat tools bundle a usage cap you'll outgrow.
• Drift, the classic sales chatbot, was sunset by Salesloft on March 6, 2026. If you're comparing it, you're comparing a dead product.
• Per-resolution pricing looks fair and scales against you: the better your bot gets at resolving, the more you pay.
• A custom chatbot costs $15K–$40K up front and $200–$800/month to run, with no per-conversation fee, so the 3-year math flips toward custom somewhere around mid-to-high conversation volume.
• Volume isn't the only axis. If your qualification logic is specific, platforms cap you on capability long before they cap you on cost.

The three pricing models, decoded

Platforms don't compete on a single price you can line up side by side. They compete on pricing models, and each one hides the real cost somewhere different.

Per-resolution (Intercom Fin)

Intercom's Fin agent charges roughly $0.99 per resolution (a conversation it handles without a human) on top of seat costs (published seat tiers run about $29 / $85 / $132 per seat per month). The pitch is that you only pay for outcomes. The catch is in the incentive: as Fin gets better and resolves more conversations, your bill goes up, not down. At 3,000 resolutions a month, that's ~$3,000 on top of seats, for a bot that's succeeding.

Per-seat plus AI credits (HubSpot)

HubSpot layers AI onto its existing seat model: roughly $90/seat/month on Professional, plus per-resolution or per-credit AI charges (around $1 per AI conversation, depending on the current packaging). The seat price is the number on the page; the AI usage is the number on the invoice. For teams already living in HubSpot, the integration is the draw, but the headline seat price meaningfully understates the all-in cost once AI usage scales.

Flat-seat with a usage cap (older tools)

The traditional model is a fixed monthly fee per seat with a conversation or contact cap baked in. Predictable until you hit the ceiling, at which point you're pushed to the next tier or hit with overages. Fine at low volume; punishing right at the moment your bot starts earning its keep.

A note on Drift: it was the reference sales chatbot for years, but Salesloft officially sunset Drift on March 6, 2026. If a comparison or quote still lists it, that content is out of date, so plan around its replacements rather than Drift itself.

Custom build: no per-conversation tax

A custom chatbot inverts the cost structure. Custom AI chatbot development costs $15K–$40K up front (a lead-qualifying bot with CRM integration) and roughly $200–$800/month to run at moderate volume. Crucially, there's no per-resolution or per-seat fee. Ten thousand conversations cost the same to price as one thousand; only the underlying token cost moves, and that's yours to optimize.

That's the whole argument: platforms tax you per conversation, a custom build doesn't. The trade is a real upfront cost and the fact that you own the maintenance. (What that ongoing cost looks like is its own topic; see what it costs to run a chatbot per month.)

Where the math flips

A custom chatbot becomes cheaper than a platform somewhere in the mid-thousands of conversations per month. Below that, a platform's $500–$1,500/month beats amortizing a $15K–$40K build. Above it, per-resolution pricing compounds past $3,000–$10,000+/month while a custom build stays near $400–$1,500. Run a 3-year projection at your real volume to find the exact point.

Put the two side by side over three years and the crossover is clear:

At low volume, the platform wins easily: no upfront cost, nothing to maintain. At high volume, per-resolution pricing compounds into six figures while the custom build's cost barely moves, because you're only paying tokens and hosting, not a per-conversation margin. The crossover for most businesses lands somewhere in the mid-thousands of conversations per month.

But cost isn't the only crossover, and it's usually not the first one you'll hit.

What most people get wrong about the comparison

They treat it as purely a cost question. It's at least as much a fit question, and fit usually breaks before cost does.

Off-the-shelf platforms ship with qualification logic designed for the median company. If your definition of "qualified" is specific (a particular company size, a specific tool in their stack, a signal that actually predicts a good customer), you'll spend month one bending the platform's rules engine to approximate your logic, and month six discovering it can't quite get there. That's the real ceiling. Most teams that move to a custom build don't do it because the platform got expensive; they do it because the platform couldn't express what they needed.

So run the decision in two passes. First: can a platform actually model your qualification and routing logic? If your criteria fit a standard lead form, the answer is yes, and a platform is the right call until volume makes it expensive. If your criteria don't fit a form, a platform will cap you on capability no matter what you pay, and the cost comparison is moot.

How to actually choose

A short decision path:

• Low volume + standard logic → platform. Don't build what you can rent for $500/month.
• High volume + standard logic → run the 3-year math; the custom build likely wins on cost alone.
• Specific qualification logic, any volume → custom, because platforms cap you on capability. Cost just decides how soon.
• Not sure which bucket you're in → scope it before you commit to either. A wrong platform choice is a migration; a wrong build is a rewrite.

If you're genuinely unsure, that uncertainty is the thing to resolve first, not by signing a platform contract or a build SOW, but by mapping your actual use case. That's what the AI Profit Leak Audit does: it tells you whether your highest-ROI move is a platform, a custom build, or something else entirely, before you've spent on either.

The bottom line

Platforms price per conversation so the cost grows exactly as the bot succeeds; a custom build trades a five-figure entry fee for no per-conversation tax and code you own. Volume decides the cost crossover, but fit usually decides first. If your qualification logic is specific, a platform caps your capability long before it strains your budget. Pick the platform when your logic is standard and your volume is modest; build custom when either of those stops being true.

Next step: Size a custom build with the full AI chatbot cost guide, or get a $497 AI Profit Leak Audit that tells you whether to rent or build before you commit to either.

Written by

Pankaj Kumar

Founder · Metageeks Technologies

Metageeks builds production-ready AI products for $1M–$15M companies — shipped in fixed-price sprints, not open-ended retainers. We write about what actually works in the field.

Connect on LinkedIn