Build vs Buy AI Agent: What 6 Vendors Actually Charge (2026-07-26 Capture)

01 The Answer

Buy an AI agent vendor's plan while your monthly interaction volume sits below the point where your own build-side cost, divided by that vendor's captured per-interaction rate, exceeds what buying costs — build only past that crossover. There is no single brand that wins outright, because six vendors' own pricing pages, captured 2026-07-26, show a unit price spread 18x wide for one automated interaction: from Crisp's $0.056 per automated conversation up to Intercom Fin's $0.99 per outcome. Which side of the threshold a reader lands on depends entirely on which vendor's rate they're pricing against, not on a universal number this page can hand over. Unless you're already below a vendor's stated monthly floor — Intercom Fin states a 50 outcomes per month minimum, captured 2026-07-26 — then buying's fixed floor cost can exceed what its per-outcome rate alone implies, and the formula still points to buying at low volume regardless.

02 What Crisp, Botpress, Chatbase, Gorgias, Intercom Fin And Vapi Actually Charge

All figures below are read from each vendor's own pricing page, captured 2026-07-26. Crisp's Mini plan is $45/mo per workspace and bundles $5 of included AI credits — its own page states that $5 covers "around 90 automated conversations." Essentials runs $95/mo with $25 of included AI credits, and Plus runs $295/mo with $75 included, plus $10/mo for each additional agent seat.

Botpress bundles a conversation allowance into each tier rather than charging per credit: Plus is $150/mo for 250 conversations/mo included, with overage sold in packs of 100 for $65. Team is $750/mo for 1,500 conversations/mo included, with overage packs of 100 for $50.

Chatbase sells message-credit tiers: Free is $0/mo for 50 message credits/month, Hobby is $32/mo for 500 credits/month, Standard is $120/mo for 4,000 credits/month, and Pro is $400/mo for 15,000 credits/month. Extra credits beyond any tier cost $40 per 1,000.

Gorgias states plainly that it is "never priced per agent," and bills two separate meters — a Helpdesk ticket meter and an AI Agent interaction meter. Its entry tier is $10/mo for Helpdesk plus $30/mo for AI Agent at a 50-ticket/month volume. Helpdesk overage runs $0.40 per ticket at lower tiers and $0.36 per ticket at higher tiers, and AI Agent overage is $1.50 per automated interaction once the plan's included batch is used. The higher published tiers run $471/mo (2,000 tickets/mo, billed annually) and $1,227/mo (5,000 tickets/mo, billed annually).

Intercom Fin is priced at $0.99 per outcome, and its page states a customer is "only charged once per conversation, even if multiple questions are answered." Fin's own pricing page adds a 50 outcomes per month minimum. Intercom's seat plan runs $29/seat/mo list, $19/seat/mo promoted; a Copilot add-on is $29 per agent/month billed annually; and a Pro add-on is $99/month, including analysis of 1,000 conversations/month.

Vapi charges a flat $0.05 per minute platform fee for calls and $0.005 per message for SMS/chat, and passes model and voice costs through "at cost ($0 if you bring your own API key)." Ten concurrent call lines are included, then $10 per line per month. A HIPAA add-on is $2,000/mo and a Zero Data Retention add-on is $1,000/mo.

03 The Finding: A Unit Price Spread 18x Wide

Put the six vendors' own captured rates side by side and the published price of one automated interaction shows a unit price spread 18x wide, all captured from each vendor's own pricing page on 2026-07-26. Crisp equates $5 of included AI credits to "around 90 automated conversations" — $5 ÷ 90 = $0.056 per automated conversation, the cheapest rate in the set. Chatbase sells extra message credits at $40 per 1,000, or $0.040 each, versus $0.064 for the credits already bundled into its $32/mo Hobby plan ($32 ÷ 500 = $0.064). Botpress includes conversations at $0.60 per conversation on Plus ($150 ÷ 250 = $0.60) and $0.50 per conversation on Team ($750 ÷ 1,500 = $0.50). Gorgias charges $1.50 per automated interaction once a plan's included batch is used. And Intercom Fin charges $0.99 per outcome, the most expensive rate captured. $0.99 ÷ $0.056 ≈ 17.7, which this capture states as a unit price spread 18x between the cheapest and most expensive published rate in the cohort.

No page ranking for "build vs buy ai agent" in the 2026-07-26 SERP capture states a single per-interaction price at all — the incumbent content in this category sells seats or credit bundles and leaves the division above to the reader. That division is the entire load-bearing fact of this page: which vendor a reader is comparing against changes the answer by more than an order of magnitude before build cost even enters the picture.

04 The Break-Even

As of September 2026, INSO does not publish its own build-delivery cost on this page, so the build side of this comparison is a variable the reader supplies, not a number this page can hand over. The formula is: your own monthly build-side cost, divided by a vendor's captured per-interaction rate, gives the monthly interaction volume at which building and buying cost the same; above that volume, building costs less per interaction, and below it, buying does.

Because the cohort's rates span a unit price spread 18x wide, the same build-side cost crosses that line at very different volumes depending on the vendor. Divide your monthly build cost by Crisp's captured $0.056 per automated conversation and you need roughly 17.7 times as many monthly interactions to reach the same crossover you'd reach dividing that identical cost by Intercom Fin's captured $0.99 per outcome. A team pricing itself against Crisp's $0.056 rate needs a proportionally larger volume before building pays off than a team pricing itself against Intercom Fin's $0.99 rate, purely because of where each vendor's own page sets its per-interaction rate.

05 When Buying Wins Anyway

The formula in section 04 is a volume threshold, not a blanket case for building. Several published conditions push toward buying even for teams that clear the volume threshold on paper:

• Stated monthly minimums. Intercom Fin's own pricing page states a 50 outcomes per month minimum, captured 2026-07-26 — a team below that volume is billed for the floor regardless of actual usage, which changes the effective per-outcome rate at low volume even though the headline $0.99 figure doesn't move.
• Compliance add-ons a small team would otherwise have to build from scratch. Vapi's HIPAA add-on is $2,000/mo and its Zero Data Retention add-on is $1,000/mo, captured 2026-07-26 — both are single line items on a vendor's bill, versus an unbounded internal certification project for a team building its own.
• Bring-your-own-model pricing that removes a build's main variable cost. Vapi passes model and voice costs through "at cost ($0 if you bring your own API key)," captured 2026-07-26 — a team that would otherwise have to manage its own model routing and failover gets that infrastructure bundled into the $0.05 per minute platform fee instead.
• Low or spiky volume in general. Below whatever crossover the formula in section 04 produces for a reader's own build cost, buying is cheaper by definition of the formula itself — that includes seasonal or early-stage products where interaction volume hasn't yet reached a steady state worth optimizing.

06 What To Check On Your Own Bill

Before running the formula in section 04 against any vendor, confirm on that vendor's own pricing page, with a capture date: the per-interaction, per-credit or per-minute rate itself and its unit basis (a per-minute rate and a per-outcome rate are not interchangeable, as Vapi's $0.05 per minute and Intercom Fin's $0.99 per outcome show); any stated monthly minimum, like Fin's 50 outcomes per month; the overage rate once an included allowance runs out, like Botpress's $65 per 100-conversation pack on Plus or Gorgias's $1.50 per automated interaction; and any per-seat or add-on costs layered on top, like Crisp's $10/mo per additional agent or Intercom's $29/$19 per-seat plan.

One limitation of this capture is worth naming plainly: it reads six vendors' list prices on one day, 2026-07-26, and none of the six vendors publish what their AI agent actually costs at your specific conversation length, resolution rate or industry vertical — those inputs change the effective per-interaction cost within a single vendor's own pricing structure, and this page's captured figures are the sticker price, not a personalized quote.

The overall answer holds regardless: buy while your interaction volume sits below the crossover your own build-side cost produces against a given vendor's captured rate, and build only past it — unless a stated minimum, compliance add-on, or bring-your-own-model term changes the math at your specific volume, in which case buying is the better call even above the raw crossover line.

07 Frequently asked questions

Is there one vendor that's cheapest for every AI agent use case?

No — Crisp's captured $0.056 per automated conversation (2026-07-26) is the cheapest rate in this cohort, but Crisp prices per workspace with bundled AI credits, while Intercom Fin's $0.99 per outcome (2026-07-26) is priced per resolution on top of a paid seat plan. The two aren't interchangeable without matching the unit basis to your own usage pattern.

How do I calculate my own build-vs-buy break-even?

Divide your own monthly build-side cost by a vendor's captured per-interaction rate — for example Crisp's $0.056 per automated conversation or Intercom Fin's $0.99 per outcome, both captured 2026-07-26. The result is the monthly interaction volume above which building costs less per interaction than buying at that vendor's rate.

Does INSO publish what it charges to build a custom AI agent?

As of September 2026, INSO does not publish its own build-delivery cost on this page, so the build side of any comparison here is a variable the reader supplies rather than a quoted bracket.

Why don't Vapi's per-minute rates compare directly to the other five vendors?

Vapi charges $0.05 per minute of connected call, captured 2026-07-26, which prices a unit of time rather than a unit of interaction. Comparing it directly to per-outcome or per-conversation rates like Intercom Fin's $0.99 per outcome or Crisp's $0.056 per automated conversation without accounting for average call length is the unit mismatch this page's pricing capture is built to avoid.

What should I check before trusting a vendor's headline per-interaction rate?

Check the unit basis, the overage rate once an included allowance runs out (Gorgias charges $1.50 per automated interaction past plan), and any per-seat costs layered on top (Crisp charges $10/mo per additional agent).