A workload-first pricing guide
AI Customer Service Pricing: Compare Units Before Prices
AI customer service prices use different denominators. A seat, conversation, automated outcome, action, credit, and voice minute do not buy the same work. Normalize each quote against one workload and its required support stack before comparing totals.
What the terms mean
Build a comparable quote by recording the work, the billing definition, and the required support stack beside each total. The worksheet below does the arithmetic without inventing a conversion between different vendors' units.
- Per seat
- A recurring charge for each human teammate or administrator. AI usage may be included or added separately.
- Per conversation or session
- A charge for an interaction window. Ask what starts, closes, reopens, or combines a conversation.
- Per resolution or outcome
- A charge when the vendor's rules classify an issue as resolved. Ask how repeat contact, handoff, exclusion, and reversal affect billing.
- Per action or credit
- A charge for a tool call or metered unit. Ask how many actions a typical customer job consumes.
- Per minute
- A voice usage charge that may sit beside telephony, model, number, transfer, and platform fees.
- Platform fee or minimum
- A base commitment before variable usage. Include required help-desk plans, implementation, integrations, and support.
Normalize the quote
An evaluation framework, not a vendor score or capability certification.
| Criterion | Input or need | Evaluation scope | Evidence to collect |
|---|---|---|---|
| Demand | Inbound requests by channel | Use the same monthly workload | Ticket, conversation, or call export |
| Success | Vendor-defined billable result | Apply one accepted-outcome definition | Outcome policy and repeat window |
| Human work | Seats or escalations omitted | Include required people and suite plans | Staffing and license inventory |
| Variable usage | Headline unit rate | Include overages, actions, messages, and minutes | Scenario calculation |
| Change cost | Software only | Add implementation, migration, integration, and administration | Year-one cost record |
Reusable decision tool
Create a comparable pricing brief
Check every cost driver that belongs in your quote request.
Bring your own quote · no signup
Compare two quotes against one workload
Enter the billable units each vendor says the same workload produces. This tool does not turn conversations into resolutions or predict an automation rate. Your inputs stay in this page and are cleared on reload.
Monthly = other monthly costs + max(0, units − recurring included units) × unit rate.
Year one = 12 × monthly + one-time costs. Usage rounds to cents each month.
Planning estimate only, not a vendor quote or savings forecast. Assumes identical monthly usage and a flat unit rate. No automatic conversion between conversations, outcomes, credits, actions or minutes. Enter taxes and all other charges yourself. Tiered rates, minimum commitments, prepaid packs, ramp-up and one-time promotional credits require a separate model. Two totals are comparable only if scope, unit definitions and cost coverage have been checked.
Complete the workload, both vendor names, unit definitions, all numeric inputs and assumptions to export.
Give every vendor the same workload
Provide monthly requests by channel, languages, average voice minutes, peak concurrency, number of human agents, expected automation scope, and the ticket sample used for the pilot. State which requests are excluded because of risk, policy, or missing data.
Ask the vendor to return the billable units produced by that workload. Do not convert conversations into resolutions with a generic percentage; run the sample and preserve each vendor's classification rules.
Write the billing definition beside the rate
For an outcome or resolution, record the closure rule, time window, repeat-contact treatment, customer confirmation, handoff treatment, excluded topics, and dispute process. For a conversation, record the session window and channel rules. For credits or actions, record a representative task's unit consumption.
A public starting price is useful for discovery, but it is not a quote for your workload. Profiles on this site preserve vendor units instead of presenting a false normalized cheapest ranking.
Separate year-one and steady-state cost
Year one can include implementation, knowledge cleanup, migration, integration, testing, training, and parallel operation. Steady-state cost can include platform minimums, seats, usage, telephony, premium support, monitoring, evaluation, and ongoing content work.
Model at least three workloads and record assumptions. The goal is not a precise five-year forecast; it is to expose which inputs can move the invoice and which operating costs are absent from the rate card.
Take this to every vendor
Your evaluation checklist
- Export one representative monthly workload by channel
- Define a successful outcome and repeat-contact window
- Record included units, minimums, tiers, overages, and exclusions
- Include every required seat, platform plan, integration, and telephony charge
- Separate implementation, migration, training, and parallel operation
- Model low, expected, and high scenarios
- Keep the vendor's unit definitions beside every calculated total
The decision tool above includes this checklist in its agent-ready copy and styled PDF. Continue with the full pilot worksheet to record results against each shortlisted product.
Primary-source evidence
Short excerpts establish definitions and product boundaries. Vendor claims are not independent performance testing. Sources reviewed September 3, 2026.
“Usage-based pricing applies to Fin and some messaging channels.”
“Pay per Action as you scale Agentforce.”
Language and multilingual support checks
A multilingual claim alone does not establish quality for your language. Test real questions, source material, channels, handoff, and the management interface.
DocsBot documents multilingual answers, including Japanese. Verify quality with your own content and workflow before buying.
For other providers, the precise Japanese and multilingual coverage shown here is not independently verified. Confirm the plan, model, channel, and handoff with each vendor.
Examples to investigate
See all 30 products →Selected examples illustrate operating models, not an exhaustive market ranking or capability certification. This site is published by DocsBot, which is included in the directory; see the editorial policy. Verify the exact plan and workflow. Corrections policy
| product | What to investigate | Catalog caution |
|---|---|---|
| Fin AI Agent | Outcome usage; the Intercom-seat path and standalone minimum-commitment path have different prerequisites. | Outcome-based AI usage plus paid seats when using Intercom. Standalone Fin has no seat, platform or setup fee, but requires a minimum monthly commitment. For chat/email on Intercom, billable service outcomes can include a configured Procedure handoff; sales qualification has a different rate. Confirm channel and contracted terms. Profile sources · checked 2026-09-02 |
| Agentforce for Service | Conversations or action credits. A customer job can consume multiple actions and require additional Salesforce licenses. | Conversation or action-credit billing. A standard action consumes 20 Flex Credits, or $0.10 at the published pack rate; a conversation can contain multiple actions. Required Service platform licenses use annual contracts. Confirm the license bundle, action mix and additional services before estimating total cost. Profile sources · checked 2026-09-02 |
| Retell AI Voice and Chat Agents | Voice minutes and chat messages. Model, telephony, phone numbers and selected features affect the bill. | Voice cost combines infrastructure, speech, model, telephony and optional per-minute features. Chat is charged per AI message. Phone numbers cost $2/month; 20 concurrent calls are included, with additional concurrency $8/month each. Use the selected configuration to estimate the monthly bill. Profile sources · checked 2026-09-02 |
| Freshdesk Freddy AI | Seats plus session packs and optional copilot seats. The introductory session allowance is one-time, not monthly. | Paid Freshdesk seats plus email AI Agent session packs. The 500 included sessions are a one-time allowance, not a monthly quota. Chat AI uses Freshdesk Omni with separate pricing. Optional Freddy Copilot costs $35/agent/month on Pro or Enterprise. Profile sources · checked 2026-09-02 |
Keep building your shortlist
Browse public pricing fields →Compare two vendors →Download the broader pilot worksheet →All buying guides →Before buying
Frequently asked questions
How is AI customer service software priced?
Common units include human seats, conversations or sessions, automated resolutions or outcomes, actions or credits, and voice minutes. Products may also require a platform fee, minimum commitment, help-desk plan, implementation, integrations, telephony, or support. Confirm the current definition and rate directly with each vendor.
Can I compare per-resolution and per-conversation prices directly?
No. First define a common customer workload and a successful outcome. Then ask each vendor how many billable conversations, resolutions, actions, or minutes that workload produces, including repeat contacts, handoffs, exclusions, and overages. Only compare totals after the units and prerequisites are explicit.
What belongs in an AI support total-cost estimate?
Include fixed platform fees, required human seats, variable AI usage, messaging or telephony, implementation, migration, integrations, training, monitoring, evaluation, content maintenance, premium support, and internal administration. Separate one-time year-one work from recurring cost.
Inclusion is not an endorsement or performance guarantee. DocsBot publishes this directory. Corrections policy