By the Onsites AI team · Last updated · 5-minute read
Intercom built a genuinely good product, and this guide will not pretend otherwise: the messenger defined the category, the automation is real, and the enterprise features are enterprise. The honest case for an alternative is not "Intercom is bad" — it is arithmetic plus fit: per-seat plans plus per-resolution fees price small teams out at volume, AI is metered at resolutions rather than published credits, and the free trial ends where a small business actually lives. Onsites AI offers the opposite deal — the full product free forever on three seats, $15 per seat beyond, published credit prices, no per-resolution meter — and this guide walks the comparison the way a buyer would: what you pay, what moves, what breaks, and when Intercom deserves to win anyway.
The migration is smaller than the fear of it. Data: export conversations, contacts and documents — email history moves by IMAP import (the migration guide's email-first cutover), CRM records map across, the widget snippet is one tag swap. Workflows: tags and views rebuild in an afternoon; SLAs re-date in one screen; the routing rules re-express in the new queue. The team's learning curve is the honest cost — a week of slower replies while habits re-form — and the honest mitigation is the parallel run: both desks live, mail lands in both, the old one goes read-only when the new one stops embarrassing itself. The AI layer transfers as configuration (drafts, translations, summaries — priced per published credit, not per resolution), and the handbook rides over as documents. Teams report the whole move inside a fortnight, mostly because nothing in it is exotic — the tools' concepts (queue, thread, record, document) are the same; the meters are not.
Migration is smaller than its fear, and it sequences in four moves. Data first: export conversations, contacts and documents; email history rides IMAP import (the migration guide's email-first cutover); CRM fields map; the widget snippet is one tag swap on the site. Workflows second: tags and views rebuild in an afternoon; SLAs re-date in one screen; the routing rules re-express in the new queue's vocabulary. Parallel run third — the honest mitigation for the learning curve: both desks live, mail lands in both, the team answers wherever the context is richest, and the old desk goes read-only the day the new one answers faster than the old one ever did. The AI layer last: drafts, translations and summaries transfer as configuration at published credit prices (the copilot guide carries the unit costs), so the only real re-learning is habit — a week of slower replies while hands re-form, which is the entire honest cost of the move. Teams that price it against the meters' difference find the fortnight pays for itself the first month, which is why the calculators — not the changelogs — decide this comparison.
Compress the whole guide into one afternoon with four lookups. Price your real shape: seats, monthly conversation volume, AI appetite — run both meters on those numbers (the pricing page's calculator takes them; the AI cost guide works the credit arithmetic). Inventory your dependencies: marketplace apps, platform bets, procurement logos — anything migrating costs more than re-earning. Run the parallel fortnight: both desks live, both meters running, the honest cost of re-learning visible before any contract turns. Decide on the annual number: the ten-person worked example — ≈$255 monthly metered against ≈$3,100 on per-seat-plus-per-resolution — is the shape of the gap most teams find when they finally price both meters on the same workload. Feature debates age; meters compound; the two-hour migration is small enough to try and honest enough to finish, which is why this decision belongs to the calculator, not the changelog.
The ten-person worked example is the whole guide's honest summary, and it deserves its last line: ≈$255 monthly metered against ≈$3,100 monthly on per-seat-plus-per-resolution pricing — an annual gap near $34,000 that neither invoice hides, and the shape of the difference most teams find when they price both meters on the same workload for the first time. Feature debates age; meters compound; the fortnight's migration is small enough to try and honest enough to finish.
Walk the meters and the difference compounds. Seats: Intercom's published seat pricing sits in the tens per seat monthly; Onsites' first three seats are free forever and every seat beyond is $15 — at ten seats that is roughly $105 monthly versus hundreds-plus (the models guide carries the full table). The resolution meter: Intercom's Fin-era pricing asks ≈$0.99 per AI resolution — a meter that scales with your success; Onsites has no per-resolution fee at all, and prices AI as published prepaid credits (a draft at 3, a Harness question at 15 — under a cent each; the cost guide works the month). Storage: per-GB packs at a dollar versus plan tiers that bundle it invisibly. The trial: fourteen days versus never expiring. The ten-person worked example — the same busy desk on both meters — lands near $255 here and around $3,100 there, and neither number hides a line: that gap is the honest headline of this comparison, and the part of the decision that outlasts every feature debate.
An honest alternative guide names the cases it loses. Intercom deserves the seat when: your product genuinely rides its ecosystem (the app marketplace, the Messenger-as-platform bets you have already built on), when enterprise procurement requires a logo they already trust and its compliance paperwork is the door to a contract that dwarfs the meter, or when your volume is small enough that seat pricing never compounds and its automation genuinely resolves your queue. The honest version of switching away is the reverse list: small teams that price their future (a resolution meter that scales with growth is the anti-deal), businesses whose support volume is their growth story, and anyone who wants the desk's tools — CRM, quotes, invoices, AI — under one published meter rather than inside a plan's packaging. Run both meters on your real numbers (the pricing calculator takes a seat count and a volume), and let the arithmetic do the arguing; the products differ less than the invoices do, and the invoice is the part that compounds.
What are the main Intercom alternatives for small teams?
Onsites AI (free forever tier with the full product — inbox, CRM, quotes, invoices — then $15/seat beyond three seats and published AI credits, no per-resolution fees), plus the per-seat platforms scaled down: Freshdesk and Zoho Desk at the low tiers, Crisp and Tidio for chat-first teams. The honest comparison runs both meters on your own seat count and volume.
Why do small teams leave Intercom?
The meters: per-seat pricing plus the ≈$0.99 AI-resolution fee scale with your success, the free trial expires before the product starts paying for itself, and AI features meter at resolutions rather than published credit prices. Onsites' structure is the inverse — permanent free tier, published credits, no resolution meter.
How hard is migrating from Intercom?
A fortnight's honest move: export conversations and records, IMAP-import email history, swap the widget tag, rebuild tags and views in an afternoon, re-date SLAs in one screen, and run both desks in parallel until the new one stops being new. The concepts transfer; the meters do not — which is usually the point.
When should a team stay on Intercom?
When its marketplace ecosystem is load-bearing for you, when enterprise procurement requires a logo they already trust, or when volume is light enough that seat pricing never compounds — the only cases where the switch costs more than the meters save. Run both invoices on your real numbers before concluding either way.
How much does the difference actually cost per year?
The worked ten-person example: ≈$255/month metered versus ≈$3,100/month on per-seat-plus-per-resolution pricing — an annual gap near $34,000, which is the honest headline of any such comparison and the reason the meters, not the feature lists, decide it.