Air AI charges a plan or licence fee plus per-minute usage, and the rate you pay depends on which layer you buy.
Air AI is a platform that runs AI voice agents which place and answer phone calls, and teams often connect those calls to the rest of their stack through workflow tools such as Activepieces.
Air AI Pricing Today, Plan by Plan
Plan table with check date
No dollar figures appear in the table below, because I haven't confirmed any. It is as of 2026-11-11 and shows each layer and how Air AI bills it.
| Layer | Price | Billing basis | Inclusions | Limits |
|---|---|---|---|---|
| Outbound | Contact sales | Plan or licence fee plus per-minute usage | AI agents that place calls | Ask for caps on concurrent calls and monthly minutes |
| Inbound / API | Contact sales | Plan or licence fee plus per-minute usage | AI agents that answer calls, programmatic access | Ask what the API rate limits are and who carries telephony |
| Licence | Contact sales | Licence fee, with usage billed on top | Rights to run the platform under your own account | Ask whether usage is bundled or metered separately |
| Custom | Contact sales | Negotiated | Scope agreed in contract | Terms exist only in the contract |
Every layer is a quote, so the useful comparison is your own monthly minutes against each quote, not a headline rate.
I have no confirmed free tier or trial terms, so don't assume one exists. Ask for any trial or demo in writing, and have it state the minutes included, whether calls reach real phone numbers, and whether usage converts to billing automatically.
If you are running this arithmetic for your own team, see what the same workload costs on Activepieces.
Limits That Force a Higher Air AI Plan
In a small test the ceilings stay hidden. The first ones real traffic hits are simultaneous calls, included minutes and agent count.
The table below shows nine limits, where each bites, and what to do about it. Get each stated figure from the vendor in writing before you price a client.
| Limit | Where it bites | Next step |
|---|---|---|
| Concurrent calls | Campaign launches and call-back peaks, when calls queue or drop | Ask for the cap in writing; price the tier that covers your peak hour |
| Included minutes | Month-end overage on a shared account | Divide the allowance by each client's real minutes |
| Agents | One client per agent configuration | Count agents per client before sharing an account |
| Voices | Brand-specific or cloned voices | Confirm which tier unlocks them |
| Languages | Multilingual markets | Test each language on a live call |
| Call-length cap | Long support or qualification calls | Compare your longest call to the cap |
| Inbound vs outbound | Teams that need both directions | Check each direction is covered by the plan |
| Commercial use | Reselling to clients | Get resale terms stated |
| Contract term | Lock-in before the pilot proves out | Negotiate a short first term |
A client who makes few minutes a month but dials everyone at 9 a.m. needs the capacity of a much bigger account. Concurrency is the limit that surprises people, because averages hide peaks.
If each client needs its own agent, voice or language, count them per client, since these decide whether one account can serve several clients.
Concurrency is the limit that surprises people, because averages hide peaks.
Calculate Air AI costs by volume
Usage cost at three volumes
If you model volumes of 1,000, 10,000 and 100,000 minutes, the monthly bill will depend heavily on the per-minute rate Air AI quotes you, which I have not confirmed. The $25,000 licence sits on top.
Why Air AI buyers go over budget
Price the whole bill, because Pricing I/O and Benchmarkit found 45% of enterprise AI buyers significantly over budget and 44% moderately over.
Per Pricing I/O and Benchmarkit, only 9% landed on budget, and 3% said it was too early to tell, so roughly nine in ten overran.
Effective cost per minute
The chart shows effective cost per minute at each volume: the $0.11 usage rate stays flat while the licence share per minute falls as volume rises. Air AI hasn't published the licence term, so the chart marks that share rather than guessing it.
Air AI cost examples at different monthly minutes
At 1,000 minutes a month, usage is $110, which is trivial next to the $25,000 licence. Air AI doesn't publish the term, so I can't give a monthly figure.
If your contract ran 12 months, which is an assumption, the licence would add about $2.08 a minute and the effective rate would be roughly $2.19.
At 10,000 minutes a month, usage is $1,100. On the same 12-month assumption, the licence adds about $0.21 a minute, for roughly $0.32 effective, which means the licence roughly triples the per-minute rate.

At 100,000 minutes a month, usage is $11,000, and the 12-month licence share drops to about $0.02 a minute, for roughly $0.13 effective, which means the licence impact diminishes substantially at high volumes.
Hidden Costs and Limits Air AI Buyers Hit
Of the unpublished costs, the carrier comes first, and the only figures to start from are the rates quoted for the carriers themselves.
Per-minute billing details
At 10,000 minutes a month, the carrier you choose moves the bill by $90 before Air AI charges anything, so Carrier choice is one variable in total cost, alongside the per-minute rate.
Twilio is quoted at $0.0140 a minute and Telnyx at $0.005, which is $140 against $50, so this pricing gap explains the $90 monthly swing from the prior sentence.
Ask Air AI whether short calls are rounded up, whether failed or unanswered calls bill, and whether it passes telephony through or marks it up. Air AI documents none of that.
Ask whether there's a one-time setup fee, whether the licence is separate from usage, and whether Air AI bills LLM or voice add-ons on top.
CRM integration and QA review are labour you pay for whichever vendor you pick.
Compliance penalties per call
One illegal call carries three different prices. According to DNC.com, the Do Not Call fine is $43,792 per violation. The TCPA standard damage is $500, and the willful rate is $1,500.

Even the lowest figure, $500, exceeds the telephony cost of any call, so one list error outweighs the carrier savings above. Confirm who carries that liability: you or the vendor.
Treat any claim of unlimited calls, instant setup or flat pricing as unverified until it appears in the terms. Ask for concurrency limits, minimum commitments and the billing definition of a "minute" in writing, then price your volume from those answers.
Worth checking against a plan that does not meter every step: one credit covers a whole run on Activepieces.
Air AI Against Vapi, Retell and Bland
Vapi, Retell and Bland list their rates on their own pages, so you can price them before any sales call, while Air AI you price from the answers you collected.
Air AI rates compared with competitors' published limits
Only one limit can be compared across all four here, and it is concurrency. It's the one that forces an upgrade: calls beyond the cap queue or fail, so a campaign that dials in bursts outgrows a low tier first.
Read the per-minute rate, platform fee and free credits from each vendor's pricing page on the day you buy, and note the date.
| Vendor | Concurrency | Public pricing |
|---|---|---|
| Air AI | Ask in writing | Confirm against your own quote |
| Vapi | 4 and 30, by plan | Rates on vendor page |
| Retell | 20 | Rates on vendor page |
| Bland | 10 and 100, by plan | Rates on vendor page |
Price each row at your own monthly minutes, with telephony and the language model added where the vendor bills them separately. Otherwise you're comparing a bundled rate with a bare one.
It comes out cheaper only if its fee and usage rate, added together, undercut the rivals' totals at your minutes. It comes out dearer if a commitment or plan fee sits underneath a modest rate.
Pick Vapi or Bland if you want to bring your own telephony or model and need room to scale. Pick Retell if a fixed, mid-sized concurrency fits your calling pattern. Pick Air AI only once its quote, at your minutes, beats a published rate.
Sending Air AI Call Results Into a Workflow
When a call ends, the outcome leaves Air AI as a webhook (an HTTP message it pushes) or through its API. Whatever catches it adds a cost line the per-minute quote never shows.
Ways to handle Air AI call data
You have three routes:
- Custom code polls the API or listens for the webhook itself, which costs engineering time and a server to host it.
- A webhook receiver is a small endpoint that accepts the call payload and passes it on, which is cheaper but leaves you maintaining retries and failures.
- An automation platform such as Zapier or Make replaces both with a hosted trigger, at a price per task, meaning each step that runs.
That price scales fast. According to Latynex Labs, 1,000 tasks a month cost $39 on Zapier and $0 on Make, so at pilot volume Make's entry plan covers it.
At 10,000 tasks the figures are $129 and $9, a gap of $120 a month. Per Latynex Labs, at 100,000 tasks they are $489 and $59, so Zapier costs more than eight times as much.

Pricing webhook to CRM workflows per run
Activepieces meters the run, not the steps inside it: its pricing page charges 1 credit per flow run, so a ten-step flow costs the same as the two-step workaround it replaced, where Zapier bills per task and Make per module on their own pricing pages.
Splitting the call handling into smaller steps, which is how you make it reliable, therefore adds nothing to the bill.
The flow is a webhook trigger that receives the Air AI payload, an HTTP connector that fetches anything missing, and a write to the CRM, the system holding your customer records. Air AI's bill doesn't change, because the call was already charged.
The screenshot below shows a five-step flow built the same way: a trigger, a record written to a table, an AI check, a router and an outbound message. Its trigger reads Discord rather than a webhook, but the shape carries over: capture, store, decide, route.

The unpublished-changes notice also matters, since an edited flow doesn't run until you publish it.
What to verify before signing Air AI
Do six things in order before you sign, and each one ends with a document or a test result you can hold the vendor to.
- Forecast your monthly minutes from real call logs, not from the vendor's sizing guess, and keep a low, expected and high case.
- Get the full fee schedule in writing: plan or licence fee, per-minute usage, and any charge for telephony, integrations or support.
- Run a paid pilot on your own call flows and your own numbers.
- Cap call volume in the account or the contract, so a runaway campaign can't bill past what you approved.
- Get two quotes from rivals that publish their rates, priced at the same minutes you forecast.
- Re-check the pricing page on signing day and compare it with your written schedule.
Air AI pricing questions to get in writing
Ask who pays when a call runs long, drops, or reaches voicemail, because a shared account quietly makes one client's wasteful campaign everyone's cost. Ask whether telephony is bundled or billed separately, and what happens to your rate if volume falls below the plan's assumption.
If an answer comes back verbally, treat it as not given. A fee you can't find in the schedule is a fee you'll meet on an invoice.
Run a paid Air AI pilot first
It costs money, and that's the point: a paid pilot produces real invoices at real usage.
Compare the pilot invoice line by line with the schedule, then divide the total by completed conversations, not minutes bought.
That figure is what you can defend to a client, and it's the one to hold against the two published-rate quotes.
What Activepieces does about this
Activepieces addresses one of the costs this article says the quote never shows: whatever catches the call outcome after Air AI hangs up.
It is open source under the MIT licence, so you can self-host it and read the code, and the hosted version charges by flow run, as the pricing page states.
That makes the cost of the handling layer something you can forecast from your own call volume.
The call-handling flow described above is the whole job: a webhook trigger takes Air AI's payload, a step fetches anything missing, and the result is written to your CRM.
Run it at your pilot's real volume and the invoice line becomes a number you can set beside the carrier and per-minute figures.
I haven't confirmed that Air AI's webhook payload maps cleanly onto your CRM fields, so test that during the paid pilot rather than assuming it. Activepieces doesn't change Air AI's bill. It keeps the work around it from becoming a second, unpriced contract.
Frequently asked questions
Because the pricing page can move between pilot and signature, none of the answers below is confirmed. Each says what to ask, and each is marked unverified rather than guessed.
Trials, billing and cancellation
Does Air AI have a free plan or free trial? Unverified: I couldn't confirm a free plan or trial from Air AI's own pricing page.
Ask for a written pilot allowance, because testing on your own call volume is what exposes the real per-minute cost, and a demo won't.
Does Air AI charge per minute or per call? Usage is billed per minute on top of the plan or licence fee, which is the structure this article prices against.
Whether failed, unanswered or voicemail calls bill is unverified. Get it in writing, because those calls would otherwise be a cost you absorb on the invoice.
Can you cancel Air AI at any time? Unverified. Ask three things before signing: the minimum term, whether the licence fee is refundable once paid, and how renewal notice works.
An upfront fee you can't recover is a sunk cost, and it should be priced into the client's retainer before you commit.
Competitor comparison and price changes
Is Air AI cheaper than Vapi or Retell? That depends on your monthly minutes, and I can't confirm it for Air AI.
Vapi and Retell publish their rates on their own pricing pages, so their side is checkable today. Price all three at the same volume, with telephony and QA included, and compare the totals. Don't compare the headline rates.

How often does Air AI change its pricing? Unverified: I found no published change history. Treat the page as something that can move, and take a dated quote or order form that fixes your rate for the term.
Otherwise a shared account can end up quietly subsidizing a client who should be paying more.
Related reading
References
Running the numbers
See what the same workload costs here.
Free forever plan, and every paid plan self-hosts at no extra cost.
See pricing Talk to sales


