Close didn’t “add AI credits” on June 3, 2026. Close admitted what every buyer learns the hard way: an AI agent inside a CRM is not a feature. It’s a meter.
As of June 3, 2026, Close formalized three separate billable surfaces: your CRM plan, telephony usage (calls, SMS, numbers), and AI credits. Three invoices. Three ways to blow your budget while your dashboard still looks “efficient.” (AI Credits doc)
TL;DR
- Close AI credits pricing is the real story, not “Chloe is included.”
- Voice Agents burn AI credits for processing and also burn telephony minutes for the call itself. Two meters. Same call. (AI Credits doc)
- Close telephony rides on Twilio usage-based rates, billed per-minute and rounded up. Short calls cost more than you think. (Calling and SMS usage, Twilio rounding)
- Call transcription via Close’s Call Assistant is $0.02 per minute transcribed (rounded up) and enrichment is $0.05 per enriched field. That’s before you buy data elsewhere. (Calling and SMS usage)
- There is no global credit cap or per-agent limit for Voice Agents. If workflows go wild, your bill follows. (AI Credits doc)
What actually changed on June 3, 2026
Close’s own docs spell it out. If you were in the Chloe beta, Close waived some costs. That party ended.
Starting June 3, 2026:
- Telephony charges now apply to Voice Agent calls. During beta, they were waived. Now Voice Agent calls get billed at standard telephony rates, the same as human calls. (AI Credits doc)
- AI credits cover AI processing costs for Voice Agent, Enrichment, Auto Updates. (AI Credits doc)
- Close frames this as normal: seats + usage + credits. Their pricing page says so, in plain text. (Close pricing)
This is not a scandal. It’s maturity. And it’s exactly where buyers get wrecked.
Close AI credits pricing, defined in one sentence
Close AI credits pricing = a monthly pool of credits (account-level), reset monthly, non-rollover, consumed by specific AI features like Voice Agent processing, enrichment, and auto-updates. (AI Credits doc)
Here’s the part buyers miss: credits are only one of the meters.
The three-meter reality: seats, telephony, credits
Close explicitly separates billing into:
- CRM subscription (seats)
- Telephony usage: calls, SMS, phone numbers
- AI credits (AI Credits doc)
So the real question is not “What does Close cost per seat?”
The real question is: What does Close cost per booked meeting when an agent is driving touches?
That number is always higher than the seat price. Sometimes 3x. Sometimes worse.
What burns AI credits in Close (and what doesn’t)
Close draws a clean line.
Free, no credits required (even if your credit balance hits zero):
- Notetaker call transcripts
- Chloe Chat
- AI Search
- AI Knowledge Sources
- Lead summaries
- Email thread summaries (AI Credits doc)
Credit-based:
- Chloe Voice Agents (AI processing for voice calls)
- Enrichment (per-field)
- Auto Updates (post-call field updates/actions) (AI Credits doc)
And Close adds the gotcha in bold letters, basically:
- All Voice Agent calls get charged telephony usage AND AI credits. (AI Credits doc)
One call. Two meters.
Credit pricing tiers: the math Close published
Close lists AI credit subscription tiers with explicit $/credit.
A few anchor points from their table:
- Tier 1: $20 for 1,200 credits (~$0.017/credit)
- Tier 6: $300 for 25,200 credits (~$0.012/credit)
- Tier 9+: $0.010/credit at scale (example: $1,000 for 100,000 credits) (AI Credits doc)
Also:
- Credits reset monthly.
- Credits do not roll over.
- Billing cycle aligns with CRM billing. (AI Credits doc)
So if you underuse credits, you eat waste. If you overuse credits, you buy another tier. No proration on upgrades, per Close. (AI Credits doc)
Telephony costs: the other meter you can’t ignore
Close telephony is usage-based and passed through at cost, tied to Twilio “pay-as-you-go” voice pricing. (Calling and SMS usage, Close pricing)
Key billing mechanics:
- Charged per minute
- Rounded up to the next minute (70 seconds becomes 2 minutes) (Calling and SMS usage, Twilio rounding)
Close’s pricing page gives the headline most people remember:
- “Most outbound calls cost around $0.02/min”
- “Phone numbers start at about $1/month” (Close pricing)
Even if that’s directionally right, rounding means your average cost per connected conversation climbs fast when you run high volume and lots of short calls.
The hidden telephony multipliers: transfers and forwarding
Close documents how charges stack:
- Forwarded incoming calls can bill as two calls (incoming + outgoing forward leg). (Calling and SMS usage)
- Transferred calls can bill as three calls (original + transfer leg + incoming to the recipient). Transfer twice, pay again. (Calling and SMS usage)
If your AI agent hands off to humans a lot, this matters. “Handoff” is not free. It’s literally extra call legs.
Call recording, transcription, and the “AI tool” fees people confuse with credits
Close splits “AI credits” from “usage-based AI tools” in telephony billing.
From Close’s usage doc:
- Call Assistant transcription costs $0.02 per minute transcribed, rounded up, and only calls longer than 30 seconds are transcribed. (Calling and SMS usage)
- Enrichment costs $0.05 per field each time you enrich that field. Bulk actions can get expensive. (Calling and SMS usage)
So you can pay for:
- Telephony minutes
- Call Assistant transcription minutes
- AI credits for Voice Agent processing
- Enrichment per-field costs
- Plus whatever you pay outside Close for lead data
“AI inside the CRM” starts looking like a receipt printer.
The Frankenstein stack problem Close doesn’t solve
Close is a CRM. It has calling, SMS, workflows, and now an AI agent. Respect.
But “booked meetings” still depends on four other cost centers buyers pretend don’t exist:
-
Data acquisition
- You still need prospect lists.
- If your ICP is narrow, you need fresh data and intent signals.
-
Enrichment
- Close charges enrichment per field. Many teams still run Apollo, Clay, ZoomInfo, Clearbit substitutes, or internal data pipelines.
-
Sequencing and deliverability
- Email volume means domain rotation, inbox placement work, warming, suppression, bounce controls.
- If your outbound program is serious, you have infrastructure costs. No way around it.
-
Human exception handling
- AI fails in predictable ways: wrong persona, wrong timing, wrong objection handling, compliance risk.
- Humans still clean it up.
So even if Close bundles a lot, the real cost model includes:
- Parallel tools
- Ops time
- The cost of mistakes
If you want the deeper deliverability mechanics, pair this with Chronic’s deliverability post: Cold Email Deliverability in 2026 Is a Relevance Problem: The Reply-Rate-First Playbook and infra math: Cold Email Domain Rotation in 2026.
Model the true cost per booked meeting (simple, CFO-proof)
Here’s a clean model you can drop into a sheet. No fantasies. Just meters.
Step 1: Define your unit economics inputs
Track per month:
- AI credits used (Voice Agent + Auto Updates + enrichment credits)
- AI credit $ cost (tier price or blended)
- Telephony minutes (outbound + transfer legs + forwarded legs)
- Telephony $ cost
- Transcription minutes (if using Call Assistant)
- Transcription $ cost ($0.02/min transcribed) (Calling and SMS usage)
- Phone numbers (count, $/month)
- Seat costs (Close subscription)
- External tools (data providers, enrichment vendors, email sending tools)
- Human hours on handoff + cleanup
- Booked meetings count
- Show rate (optional, but you should)
Step 2: Calculate Total Outbound Cost
Total Outbound Cost =
- Close seats
-
- AI credit subscription spend
-
- telephony usage
-
- transcription spend
-
- phone number rentals
-
- external tools
-
- human time cost (handoff + QA + ops)
Step 3: Cost per booked meeting
Cost per booked meeting = Total Outbound Cost / Meetings Booked
Also track:
- Cost per held meeting
- Cost per qualified meeting Because AI can “book” junk. Calendars accept anything.
Example (intentionally conservative)
Let’s say in one month:
- $900 Close seats
- $400 AI credit plan
- $600 calling minutes (rounded up stings)
- $250 transcription
- $60 numbers
- $1,200 data + enrichment outside Close
- $1,500 human time cleanup
- 40 meetings booked
Total = $4,910
Cost per booked meeting = $122.75
That’s not insane. It’s just not the seat price. And this is where Close’s June 3 change matters: it forces you to stop pretending the agent is “included.”
Buyer checklist: what you must verify before you sign anything
Print this. Use it in procurement. Save yourself from the “wait, why is there another invoice?” moment.
1) What counts as an AI action?
Ask Close (or any vendor):
- Which features burn credits vs which are free?
- Does post-call summarization burn credits?
- Do auto-updates burn credits every time they run? (AI Credits doc)
In Close specifically:
- Voice Agent processing burns credits.
- Enrichment burns credits or per-field cost depending on how it’s applied in your setup.
- Auto Updates burns credits. (AI Credits doc)
2) What triggers credit burn?
You want exact triggers:
- Per call attempt?
- Per connected call?
- Per minute?
- Per transcript?
- Per tool invocation?
Close states there is no global credit cap and no per-agent limit. Your workflow design is the governor. (AI Credits doc)
3) Call recording and transcription fees
Separate:
- Call minutes (carrier)
- Transcription minutes (AI tool)
Close’s Call Assistant: $0.02 per minute transcribed, rounded up, calls over 30 seconds. (Calling and SMS usage)
4) Carrier fees and rounding
Confirm:
- Per-minute rates by geography
- Rounding rules
Twilio rounds partial minutes up. Close reiterates this and links to Twilio. (Calling and SMS usage, Twilio rounding)
5) Local presence and number provisioning
If you run local presence:
- How many numbers do you need for coverage?
- What’s the monthly rental per number?
- Do you pay for “lines” or “numbers” differently?
Close says numbers start around $1/month. That’s cheap until you need dozens. (Close pricing)
6) Concurrency limits and throughput
For voice agents:
- How many calls can run at once?
- Is concurrency tied to seats, plan, or telephony configuration?
- What happens when you hit the ceiling?
Even if a vendor doesn’t publish this clearly, you need the operational answer. Concurrency is how “AI agent” becomes “pipeline.”
7) Failure mode: handoff + human time
This is the tax nobody budgets.
- What’s the handoff workflow?
- Does handoff create extra call legs (transfer fees)?
- Who updates fields when the agent messes up?
Close documents transfers can create multiple billable calls. If your AI hands off often, your cost curve changes. (Calling and SMS usage)
8) Parallel tool costs you still pay
Ask your team, not the vendor:
- Are we still paying for Apollo, Clay, Instantly, dialer tools, enrichment tools?
- If yes, what are we actually buying Close for?
If your stack looks like a crime scene, your “AI agent inside CRM” ROI will too.
If you want a structured way to evaluate signals and prioritize who gets the agent’s attention, use: The 2026 Outbound Scorecard: 25 Signals That Predict Meetings. That’s how you stop wasting call minutes on dead leads.
The real lesson from Close’s June 3 pricing change
Close didn’t get greedier. Close got honest.
AI agents create usage. Usage creates billing surfaces. Billing surfaces create variance. Variance creates budget fights.
So buyers need one thing: a cost model tied to outcomes. Not credits. Not minutes. Not seats. Outcomes.
If you want a framework for usage-based AI pricing that doesn’t collapse in week two, pair this with: Usage-Based AI Pricing Calculator for Sales Teams (Credits, Actions, Outcomes).
One clean contrast: Chronic vs the Frankenstack
Close is a strong CRM with native comms and an agent. If you already live in Close, Chloe can be a real throughput unlock. Just don’t pretend it’s free.
Chronic takes a different stance: $99. Unlimited seats. End-to-end, till the meeting is booked. No Frankenstack. Pipeline on autopilot.
Chronic runs the full loop:
- Build your ICP with the ICP Builder
- Pull and enrich leads with Lead Enrichment
- Write and send personalization with the AI Email Writer
- Prioritize with AI Lead Scoring
- Track everything in your Sales Pipeline
If you’re comparing stacks head-to-head, start here: Chronic vs Close.
FAQ
What is “Close AI credits pricing” in plain English?
It’s a monthly pool of credits used for specific AI features, separate from your Close seat subscription and separate from telephony usage charges. Credits reset monthly and do not roll over. See Close’s AI Credits documentation for the exact list of what burns credits. (AI Credits doc)
Do Close Voice Agent calls cost both credits and per-minute calling fees?
Yes. Close states that all Voice Agent calls are charged standard telephony usage rates and also incur AI credit charges for AI processing. Same call, two meters. (AI Credits doc)
Are call transcripts part of AI credits in Close?
Not necessarily. Close’s usage-based AI tool pricing lists Call Assistant transcription at $0.02 per minute transcribed, rounded up, and only for calls longer than 30 seconds. That is billed as usage, not AI credits. (Calling and SMS usage)
Why do outbound call costs feel higher than the listed per-minute rate?
Because calls are billed per minute and partial minutes are rounded up. A 70-second call bills as 2 minutes. Twilio documents this rounding, and Close references Twilio pricing mechanics. (Calling and SMS usage, Twilio rounding)
What’s the fastest way to estimate true cost per booked meeting with Close?
Add up monthly: seat fees + AI credit spend + telephony minutes + transcription + phone number rentals + external data/enrichment tools + human cleanup time. Then divide by meetings booked, and separately by meetings held. If you only compute “per seat,” you’re lying to yourself.
What should I ask Close about before scaling Chloe Voice Agents?
Ask for: exact credit triggers, any concurrency limits, how transfers and forwarding impact billing, how to control credit usage, and what happens operationally when the agent fails and hands off to humans. Close notes there is no global credit cap, so workflow design becomes your cost control. (AI Credits doc, Calling and SMS usage)
Run the numbers before you scale the agent
Do this in a spreadsheet today:
- Pull last month’s call minutes.
- Apply rounding reality and transfer/forwarding multipliers where relevant.
- Estimate AI credit burn per workflow.
- Add transcription, enrichment, number rentals.
- Add your parallel tool stack.
- Divide by booked meetings and held meetings.
If the number still works, scale it. If it doesn’t, fix the system, not the vibes.