Compare AI agent prices by their billable unit, minimum commitment and required base plan. A seat, a conversation, a resolved outcome and a credit measure different things. Quickchat AI Enterprise starts at $0.50 per resolution with a 2,000-resolution monthly minimum. Fin’s external-helpdesk plan combines a base charge with outcome pricing. Salesforce offers separate conversation, resolution and credit-based options. Those headline rates alone do not establish which deployment costs less.
This guide compares the published terms below, checked on November 3, 2026, and works through 2,000, 10,000 and 50,000 monthly conversations. All dollar amounts are USD. The examples are planning calculations, not vendor quotes or measured resolution results.
Optional setup tools
- Compare supplied pricing quotes. Normalize declared assumptions and flag missing inputs. This tool does not fetch vendor prices or provide a commercial quote.
- Compare resolution billing scenarios. Vary supplied volume and resolution-rate assumptions. This tool does not predict resolution quality or fetch prices.
- Estimate usage costs. Calculate Quickchat AI model credits and a separate total for supplied fee lines. This tool does not look up provider rates, combine credits with money or check your account.
TL;DR: 2026 AI agent pricing at a glance
| Billing model | What is counted | Inputs needed for a comparison |
|---|---|---|
| Per-seat | Licensed human users | Seat count, included features, base platform fees and usage limits |
| Per-ticket or conversation | Eligible tickets or sessions | Session boundaries, channels, repeated contacts and minimums |
| Per-resolution or outcome | Events meeting the vendor’s definition | Confirmed versus assumed outcomes, handoffs, reopening rules and included usage |
| Credits or actions | Metered replies, actions or other operations | Credits per operation, operations per conversation and allowance or purchase terms |
Subscriptions can combine a seat fee with usage charges. A credit has no common value across vendors, and a conversation can consume several replies or actions. Measure your workload before converting either into a cost per customer issue.
AI agent pricing models
Per-seat
A seat fee buys access for a named or eligible human user. It may cover a helpdesk, a copilot or employee-facing automation. Check whether customer-facing AI usage is included or charged separately. A small seat bill says little about the amount of autonomous work included.
Per-ticket
A ticket or conversation fee follows the vendor’s event definition. A reopened issue or a new session can create another billable unit. Conversation pricing does not inherently require a successful resolution. Count the sessions the selected product would bill, rather than applying your resolution rate to all conversation-priced products.
Per-resolution
The charge follows a qualifying outcome. Some vendors accept inferred success after inactivity; some distinguish a routine escalation from a configured, billable handoff. The written definition, included allowance and minimum commitment determine the invoice. A published resolution rate alone does not establish customer satisfaction or verified task completion.
Credits and actions
Credits meter usage under a product-specific rate card. Quickchat AI self-service credits measure AI replies and vary with model choice; Salesforce Flex Credits meter operations such as actions. Estimate the number and type of operations in a representative conversation. Include testing and additional platform consumption where the applicable terms charge for them.
Published prices in 2026
The following table records the selected product and its public terms as of the source-check date. Confirm the account’s pricing version, region and contract before purchasing. Required helpdesk subscriptions, taxes, channel fees and implementation can sit outside the headline unit price.
| Product or pricing option | Published unit or starting price | Allowance, minimum and scope |
|---|---|---|
| Quickchat AI Enterprise | From $0.50 per resolution | Starts at 2,000 resolutions/month; custom agreement and seat count. This is separate from self-service reply-credit plans. |
| Fin with an external helpdesk | $49/month including 50 resolutions; $0.99 per additional outcome | Unused monthly resolutions do not roll over. Outcomes can include configured Procedure handoffs. Existing helpdesk costs remain separate. |
| HubSpot Customer Agent | 50 HubSpot Credits per resolved text conversation | $0.50 at the standard USD $0.01/credit rate. Requires an eligible Professional or Enterprise subscription; included credits and shared credit consumption affect additional spend. |
| Zendesk AI agents | Advertised $1.50 committed / $2 pay-as-you-go automated resolution | Base plan and account pricing model determine allowances and eligibility. Legacy automated-resolution allowances and newer resolution tiers are different arrangements. |
| Salesforce Agentforce Conversations | $2 per conversation | Session-based consumption; do not multiply by a resolution rate. Confirm platform, channel and commitment requirements. |
| Salesforce Help Agent Resolutions | $2 per resolution at list rates | Separate offering, metered at 400 Flex Credits per production resolution. Requires an eligible Salesforce edition and entitlement. |
| Salesforce Agentforce Flex Credits | $500 per 100,000 credits | Standard Agentforce actions use 20 credits ($0.10 at this rate). Several actions can occur in one conversation; other operation types have different rates. |
Fin can run on an existing external helpdesk without Intercom seats. Fin on Intercom instead combines the selected Intercom plan with its applicable outcome charges. Do not apply the external-helpdesk base bundle to every Fin deployment.
HubSpot’s credit rules specify eligibility, included allowances and additional usage. Included credits are shared across eligible functions, so multiplying all resolutions by $0.50 is a consumption-value illustration, not necessarily the additional invoice amount.
Zendesk documents both the legacy automated-resolution platform and resolution tiers introduced in May 2026. Establish which model the account uses before applying allowances or the advertised rate.
Salesforce’s Help Agent billing documentation lists eligible editions and add-on requirements. Its August 31, 2026 rate card distinguishes Help Agent resolutions from individual actions. A $2 conversation and a $2 Help Agent resolution remain different billing options even though their list prices match.
For vendors supplying custom quotes, including Decagon, Sierra, Ada and Cognigy, collect the exact proposal rather than estimating a missing public rate. The absence of a number does not prove a particular minimum, outcome model or NDA requirement.
For product comparisons, see Quickchat AI and Intercom Fin, Salesforce Agentforce and HubSpot AI Breeze. Use each vendor’s current pricing documentation for the commercial terms.
What “resolution” actually means in a vendor contract
Ask which observed events become billable. Review a sample containing successful answers, inactivity, clarification requests, human handoffs, failed actions and reopened issues.
- Quickchat AI’s published definition includes confirmed and assumed resolutions, excluding greeting-only exchanges. An assumed resolution does not require explicit positive feedback. Confirm the contract’s treatment of disputed or reopened cases.
- Fin’s outcome documentation distinguishes confirmed and assumed resolutions from Procedure handoffs. A configured handoff can be a charged outcome even though a human continues the work; ordinary escalation is treated differently. Its resolution-rate metric and total billed outcomes can therefore differ.
- HubSpot’s resolution rules include a 72-hour evaluation window. Qualifying visitor-requested handoffs prevent a resolution charge; manual or workflow reassignment alone does not. Lead qualification can also count as a resolution.
- Zendesk’s newer tiers distinguish contained and verified resolutions, with an LLM verification step for the latter. Apply the account’s actual allowance and tier rules rather than treating every closed chat as one universal paid resolution.
Ask for a way to reconcile billed events with conversation records, the relevant timestamps and any adjustments. Quickchat AI’s Inbox documentation describes answer explanations, source material and Actions. These support reviewing responses; they do not by themselves establish an invoice reconciliation view or an audit-speed guarantee.
Worked total cost of ownership
Start with the software charge, then add the same categories of operating cost to every option. The tables below isolate selected AI charges; they are not complete TCO or a ranking of equivalent capabilities. A licensed copilot and an autonomous customer-service agent can provide different functions.
Assumptions for these illustrations:
- Monthly inbound conversations: 2,000, 10,000 or 50,000. For conversation-priced rows, assume one billable session per inbound conversation.
- Resolution/outcome volume: 75% of inbound conversations. This is a chosen scenario, not a market benchmark or predicted performance. Assume Fin has no additional billable Procedure handoffs in this example.
- Hypothetical seat-based AI add-on: $50 per seat/month, with 5, 15 or 40 licensed users. This is an illustrative rate, not a quote for a named product.
- Hypothetical ticket-based offer: $0.50 per billable inbound ticket.
- Quickchat AI illustration: $0.50 starting rate and a 2,000-resolution minimum commitment. Confirm the actual rate, commitment and other charges in the Enterprise quote.
- Fin illustration: the external-helpdesk base plan, $49 + $0.99 × outcomes above the included 50 resolutions.
- Salesforce illustration: the Conversations option at $2 per billable session. Help Agent Resolutions and Flex Credits require separate calculations and may produce different costs.
- Human salaries, base helpdesk subscriptions, implementation, channel charges and taxes are excluded from every software row. Included allowances and discounts are applied only where explicitly stated.
2,000 conversations per month (small support team, ~5 agents)
| Selected model | Calculation | Illustrative monthly charge |
|---|---|---|
| Hypothetical per-seat add-on | 5 × $50 | $250 |
| Hypothetical per-ticket offer | 2,000 × $0.50 | $1,000 |
| Quickchat AI Enterprise, starting-rate assumption | 2,000 minimum × $0.50; 1,500 assumed resolutions | $1,000 |
| Fin with an external helpdesk | $49 + (1,500 − 50) × $0.99 | $1,484.50 |
| Salesforce Agentforce Conversations | 2,000 sessions × $2 | $4,000 |
The Quickchat AI illustration uses the minimum commitment, not just the 1,500 assumed resolutions. Its Enterprise terms require a quote. For smaller deployments, the separate self-service plans start at $9/month for commercial use with 150 reply credits; those credits are not resolutions. See how AI credits are counted before estimating a conversation budget.
10,000 conversations per month (mid-market, ~15 agents)
| Selected model | Calculation | Illustrative monthly charge |
|---|---|---|
| Hypothetical per-seat add-on | 15 × $50 | $750 |
| Hypothetical per-ticket offer | 10,000 × $0.50 | $5,000 |
| Quickchat AI Enterprise, starting-rate assumption | 7,500 × $0.50 | $3,750 |
| Fin with an external helpdesk | $49 + (7,500 − 50) × $0.99 | $7,424.50 |
| Salesforce Agentforce Conversations | 10,000 sessions × $2 | $20,000 |
Multiplying Salesforce’s conversation rate by 7,500 resolutions would undercount this scenario’s 10,000 assumed billable sessions. Conversely, a quote for Help Agent Resolutions would use its resolution rules. Select the intended SKU before doing the arithmetic.
50,000 conversations per month (enterprise, ~40 agents)
| Selected model | Calculation | Illustrative monthly charge |
|---|---|---|
| Hypothetical per-seat add-on | 40 × $50 | $2,000 |
| Hypothetical per-ticket offer | 50,000 × $0.50 | $25,000 |
| Quickchat AI Enterprise, starting-rate assumption | 37,500 × $0.50 | $18,750 |
| Fin with an external helpdesk | $49 + (37,500 − 50) × $0.99 | $37,124.50 |
| Salesforce Agentforce Conversations | 50,000 sessions × $2 | $100,000 |
At this volume, committed usage, negotiated rates and the actual mix of billable events can materially change the result. The table retains the stated assumptions to make the arithmetic reproducible; it does not estimate negotiated Enterprise pricing.
Use the same TCO boundary for every option:
Monthly TCO = AI charges + base platform and seats + human support cost + channel/integration charges + implementation amortization.
For example, suppose a deployment retains 15 staff at an illustrative fully loaded $3,000/month each, pays $600/month for its base platform and allocates a $12,000 implementation cost over 12 months. That adds $46,600/month to whichever AI charge is being evaluated, before any other fees. The hypothetical $750 seat add-on would total $47,350; the $3,750 starting-rate resolution illustration would total $50,350 under these same assumptions. This isolates pricing arithmetic, not a staffing forecast. Change headcount only when workload tests support that change, and avoid double-counting fees already bundled in a quote.
Use the existing chatbot ROI calculator with your own assumptions. The Quickchat AI Enterprise overview describes the offering; the signed proposal establishes its full commercial terms.
Behavioral incentives by pricing model
Seat revenue follows licensed access, conversation revenue follows eligible sessions, outcome revenue follows billable outcomes, and credit revenue follows metered operations. These relationships explain sensitivity to usage. They do not establish a vendor’s product quality, action coverage or willingness to automate work.
An outcome definition can reward a closed conversation without proving that the customer achieved the intended result. Track billing alongside task completion, repeat contacts, escalation and customer feedback. Use comparable test cases and keep the vendor’s billed metric separate from your own service-quality measure.
Three transparency questions to ask before signing
- What is the complete charging rule? Record the SKU, event definition, included usage, minimum commitment, allowance expiry, overage rate and any separately metered operations. Confirm whether commitments are monthly or annual.
- What sits outside the unit price? Itemize required seats, helpdesk subscriptions, implementation, integrations, channel fees and ongoing support. Identify which costs are bundled so they are not counted twice.
- What changes at renewal or during a dispute? Obtain the renewal mechanism and the rules for reopened, misclassified or disputed outcomes. A published starting price does not promise a fixed renewal rate; Quickchat AI’s terms also allow fee changes with notice under their stated conditions.
Related reading: outcome-based pricing, chatbot costs, customer-service AI agents and the AI agent platform evaluation guide.
Frequently asked questions
How are AI agents being priced in 2026? Customer-service AI can be billed by licensed seat, conversation, resolved outcome or usage credits. A subscription can combine several of these. Compare the exact product, billable event, included allowance, minimum commitment and required platform subscription before comparing dollar rates.
How much does a customer service AI agent cost? Quickchat AI Enterprise starts at $0.50 per resolution with a 2,000-resolution monthly minimum. Fin with an external helpdesk lists a $49 monthly base including 50 resolutions, then $0.99 per additional outcome. Salesforce lists both $2 Conversations and a separate $2 Help Agent Resolutions option. These prices use different billing definitions and exclude costs that depend on the selected contract and deployment.
What is the cheapest AI agent pricing model in 2026? There is no universal cheapest model or 3,000-conversation crossover. Costs depend on licensed seats, eligible usage, minimum commitments, included allowances and the capabilities being purchased. Compare the same workload and include human staffing, platform fees and implementation consistently in every option.
Why do some AI agent vendors not publish their prices? Some vendors provide a negotiated quote instead of a public unit rate. Ask for the billing unit, committed volume, included usage, overages, implementation charges and renewal terms in writing. A missing public rate does not establish a particular price, minimum commitment or requirement to sign an NDA.
What does per-resolution actually mean? A resolution is a vendor-defined billable outcome. It can include an inferred resolution when a visitor stops requesting help, rather than explicit confirmation. Some outcome plans also bill configured handoffs. Check the rules for inactivity, escalation, reopened conversations and refunds; a resolution metric and a billing total need not count the same events.
How does Quickchat AI’s $0.50 compare to Intercom Fin and Salesforce Agentforce? Quickchat AI publishes an Enterprise starting rate with a monthly minimum. Fin charges for outcomes under its selected plan. Salesforce Conversations charges by session, while Help Agent Resolutions is a different offering. A $0.50-to-$2 headline comparison alone cannot establish four times lower total cost or equivalent results. Apply each definition and contract to the same test workload.
Should small support teams pick per-seat or per-resolution pricing? Estimate the required seats and eligible billable usage, then apply each plan’s base charge, included allowance and minimum. A small workload can fall below an Enterprise minimum. Quickchat AI also has separate self-service plans using reply credits. Test required support workflows and human handoff before treating a seat-based assistant and an autonomous service agent as substitutes.
What enterprise pricing terms should I negotiate? Record the billing definition, monthly or annual commitment, allowance expiry, overage rate, setup costs, support scope and renewal mechanism. Ask how billed events can be reconciled with conversation records and how disputed or reopened cases are handled. Confirm any service levels and data-handling requirements in the agreement rather than assuming they follow from a headline price.