AI Chatbot Pricing: SaaS Support Platform vs Custom Build

At 1,000 billable outcomes, Intercom’s $0.99 rate adds up to $990 in AI fees before seat costs. Freshdesk’s session pricing produces a different bill for 1,000 sessions, even though both figures describe support automation. AI chatbot pricing is comparable only when the billing unit and platform charges are shown together.

For a small SaaS support team, a subscription may cost $0 to several hundred dollars monthly, while usage fees depend on whether a vendor bills for sessions, replies, or resolutions. A custom build has no standard subscription price: its estimate depends on scoped engineering and ongoing operating costs. The useful comparison is the total cost of a defined support workflow, including platform access and human follow-up.

AI chatbot pricing at a glance

Free and starter plans can support a limited trial, and published SaaS support-agent plans can cost tens or hundreds of dollars monthly. Usage-based examples in Gleap’s pricing guide range from cents to dollars per billed unit. Custom development is a separate project budget, not another monthly chatbot tier. These are different cost structures, not guaranteed quotes for every team.

Zealoop lists a Free plan at $0, Starter at $29 per month, Growth at $99, and Scale at $299. Growth includes 1,200 conversations per month, 25 actions, 10 tables, and 10 seats; Scale includes 6,000 conversations and unlimited sources, actions, tables, and seats. Those figures describe Zealoop’s plan allowances, not a universal market rate. Zealoop’s pricing page lists flat plan prices without per-seat or per-resolution charges.

A low-looking usage rate is not a total monthly cost. The invoice can also include a help-desk subscription, agent seats, onboarding, integration work, and human follow-up. Compare the same support volume and list each component separately rather than treating an AI fee as the whole service price.

What kind of chatbot is being priced?

The term “chatbot” can refer to tools with different users and operating costs. A personal AI subscription is for an individual’s general use; it does not establish the cost of an embedded agent that answers customers using company documentation or account data. A custom-built agent is different again: the buyer funds engineering and ongoing operations rather than selecting a standard product tier.

For SaaS support, compare products against the work the team expects them to do:

• Documentation answers: The agent retrieves information from approved company content.
• Customer-data lookups: It accesses a relevant account or subscription record.
• Guarded actions: It performs a limited operation, such as updating an address, with permissions and checks.
• Human escalation: It transfers a case when the answer or action requires staff.

The more the workflow depends on a company’s systems and permissions, the less useful a comparison based only on a general chatbot’s subscription price becomes. A support agent’s price should be judged against the whole workflow, including what happens when automation stops.

Common AI chatbot pricing models

A vendor’s billing unit determines what is charged when the bot participates but does not solve the case. The prices below are examples listed in Gleap’s comparison; their units are not interchangeable.

A fixed subscription charges for a plan and its stated allowance. Per-seat pricing adds cost as more staff need access. Usage or credit pricing charges for activity, while per-conversation and per-resolution models apply different rules to customer interactions. Chatbase credits are not equivalent to conversations: 4,000 credits could cover different reply counts depending on the model and its credit use.

For each quote, identify whether the listed amount is the AI fee, the platform fee, or both. Also ask whether an escalation is billable, whether a resolution includes a handoff, and how the product handles usage beyond its allowance.

What drives the monthly cost of a SaaS support chatbot?

Monthly cost depends on how the agent fits into the support operation, not just on message volume. A practical estimate should account for:

1. Support volume and billable share. For a session fee, count sessions that include an AI reply. For outcome pricing, estimate cases meeting that vendor’s resolution definition.
2. Seats and help-desk access. Freshdesk’s cited per-agent plans run from $29 to $119 per agent monthly when billed annually. A multi-agent team must include the seats it actually needs.
3. Knowledge sources. A team with separate product guides or help centers should compare source allowances. Zealoop Growth includes 100 knowledge sources; its Free plan includes three.
4. Customer-data integrations. Account lookups require access to the right data and testing of permissions. Establish whether connector setup is included in the subscription or scoped separately.
5. Actions and safeguards. A refund or address change is not just another text reply. Confirm action limits and how permission checks and failures are handled.
6. Escalation and follow-up. If the AI cannot resolve a case, staff time remains part of the support cost. A vendor may also bill for the AI interaction that preceded the handoff.
7. Setup and limits. Onboarding and configuration can affect the first-year budget. Zealoop’s published policy is that the agent stops answering at its conversation limit until the next month or an upgrade, with no overage bill.

A plan limit has an operational cost as well as a billing effect: a pause at the cap may leave more questions for staff. Include the expected consequence of reaching that limit in the team’s decision.

Compare chatbot costs by business size and support volume

The examples below use stated assumptions to show how platform charges change the comparison. They are not equivalent product bundles: Zealoop’s published plan price is shown separately from Freshdesk’s AI session and seat fees. Freshdesk’s first 500 sessions are described as free once, so the table shows both the initial allowance and the recurring cost after it has been used.

Freshdesk’s $490 recurring AI fee is calculated as 10 blocks of 100 sessions at $49 each. The one-time 500 free sessions would reduce the first 1,000-session month’s AI fee to $245, making the known first-month total $274–$364 with one seat. Other fees are not included in these totals. A session is not necessarily a resolved issue, so this example should not be read as a direct quality comparison with a plan that counts conversations differently.

The table shows the value of adding the platform line instead of comparing a plan against a bare AI unit fee. Seat count, included services, and the definition of a billable interaction still need to match before the totals can support a purchasing decision.

Software subscription versus building a custom chatbot

A subscription replaces much of the initial product engineering with a recurring plan. It may suit a team whose workflow fits the product’s documentation, data, action, and escalation capabilities. Zealoop’s published plans, for example, range from $29 Starter to $299 Scale per month; Scale includes 6,000 conversations and unlimited sources, actions, tables, and seats.

A custom agent shifts more responsibility to the SaaS team or its development partner. A scoped estimate should separate work into these cost lines:

• Discovery of support intents, exception paths, and success criteria.
• Documentation retrieval and evaluation against representative questions.
• Data connectors, authentication, permissions, and guarded action logic.
• Testing for incorrect answers, failed actions, and escalation behavior.
• Deployment, monitoring, maintenance, and model or hosting usage.

The cited pricing comparison provides no engineering-hour estimates, hourly rates, or hosting and model-usage budget for custom projects. Giving a numeric build range from it would create a false benchmark. A buyer can still compare a quote responsibly by requesting hours and costs for each workstream, identifying one-time implementation separately from recurring hosting and maintenance, and asking which changes trigger additional work.

Custom development is most defensible when a required integration, permission model, or workflow cannot be delivered safely by an available subscription. It is not automatically cheaper at high volume: the team must account for the ongoing work of keeping a bespoke agent functional as its product and support policies change.

Calculate the true cost per resolved support issue

Use successfully resolved customer issues—not AI replies—as the denominator:

Cost per resolved issue = (platform fee + AI fees + setup allocated to the month + human follow-up cost) ÷ successfully resolved issues

For a like-unit example, assume 1,000 conversations each qualify as an Intercom outcome and one agent seat is needed. The cited prices produce $990 in outcome fees plus $29–$132 for that seat, or $1,019–$1,122 before any other costs. If all 1,000 outcomes meet the team’s definition of a successfully resolved issue, that is $1.019–$1.122 per issue. If some are handoffs or do not solve the customer’s problem, the cost per successful resolution rises.

Help Scout’s listed $0.75 fee would make 1,000 billable resolutions $750 in AI fees, but the platform fee is not included in that figure. Its total cannot be calculated from the listed unit price alone. For any vendor, add the support platform and staff time before calling the result a cost per resolved issue.

Compare quotes and avoid surprise charges

Ask for a sample invoice based on the team’s actual support data, including cases the AI does not resolve. A pricing calculator is useful only if its inputs reflect the vendor’s billing rules.

• What precisely counts as a message, action, session, conversation, outcome, or resolution?
• Are seats, channels, knowledge sources, and integrations included in the quoted plan?
• How many credits does a typical answer or multi-step lookup consume?
• What happens at the limit: does the agent pause, prompt an upgrade, or add overage fees?
• Does a human handoff count as a billable interaction, and is staff-side AI priced separately?
• Which onboarding, connector, or maintenance tasks are one-time, and which recur?
• Can the team cap usage, export its data, and cancel without a longer commitment?

The main comparison risk is unclear unit definitions. A quoted “resolution” can differ from the support team’s own measure of a solved case, and a session can include an unresolved interaction. Ask the vendor to apply its rules to a few real, anonymized examples before forecasting a monthly bill.

Which should you choose?

• Choose a free or starter subscription to test a small support workflow. Zealoop Free includes 50 conversations per month and three knowledge sources; Starter includes 300 conversations and 25 sources.
• Choose a flat plan when monthly predictability matters and the included allowance fits expected volume. Check what happens at the cap, not just the listed price.
• Choose usage pricing when activity varies and the team can forecast billable units. Compare the AI fee plus seats and platform costs; Freshdesk’s 1,000-session example shows how a recurring AI charge can sit alongside a seat fee.
• Choose a custom build when the required workflow or controls are unavailable in packaged products and the business can fund the continuing engineering and operational work.

The better-priced option is the one that resolves the team’s intended cases with predictable limits and manageable follow-up—not necessarily the option with the lowest advertised fee per interaction.

FAQ

Do AI chatbots cost money?

Some support products have a free plan or a limited free allowance, while paid products may charge subscriptions, seats, usage, or a combination. Zealoop’s Free plan includes 50 conversations per month and three knowledge sources. A free tier is a starting allowance, not evidence that a full support workflow has no operating costs.

How much does an AI chatbot cost per month for 1,000 support conversations?

The answer depends on the billing unit and platform fees. Zealoop Growth is $99 per month for up to 1,200 conversations. Freshdesk’s listed session fee for 1,000 sessions is $490 after its one-time free allowance is used, plus a $29–$119 monthly fee for one agent seat.

Is an AI chatbot charged per message, conversation, or resolution?

It can be charged by any of those units, or through seats, sessions, actions, or credits. For instance, Help Scout lists $0.75 per resolution, while Tidio lists $0.70–$0.78 per conversation with at least one Lyro reply. The invoice depends on each provider’s definition of the unit.

What should a custom AI agent quote include?

It should distinguish discovery, integrations, permissions, evaluation, deployment, and maintenance instead of giving one undifferentiated project total. The estimate should also identify recurring hosting and model usage separately from implementation. Without that breakdown, a buyer cannot compare a custom build’s ongoing ownership cost with a subscription plan.

Sources

• www.gleap.ai/blog/ai-chatbot-pricing
• www.zealoop.com/pricing
• www.zealoop.com/