How Enterprise AI Agents Are Priced: Models, Real Prices, and TCO

See the six AI agent pricing models, with list prices from each vendor's own pricing page as of September 2026. Then learn the costs outside the price sheet and the questions to ask before you sign.

Blog17 min read

This guide helps you put a number in a budget and defend it later. It is for the line-of-business owner who sponsors the project, and for the finance, procurement, and platform teams who review it. It covers the six pricing models in use today, with real list prices from each vendor's own pricing page in September 2026. A worked example shows how design choices move a bill. The guide then covers the total cost of ownership beyond the license, and the questions that show the difference between a quote and a cost.

Start with a rule that most pricing pages omit. Microsoft's billing documentation for Copilot Studio states it. When a tenant on prepaid capacity reaches 125% of that capacity, Copilot Studio disables its custom agents. People who try to use one get a message that the agent reached its usage limit. The conversation in progress finishes. The next one does not start.

This is fair to Microsoft. It publishes the rule, the per-feature rates, and a usage estimator, and many vendors publish less. The rule shows that AI agent pricing has two questions: what the meter charges, and what happens to the work when the capacity ends.

What is AI agent pricing?

AI agent pricing is the way a vendor charges for software that plans and takes actions with delegated access. This software does more than answer questions. Vendors meter one of six things. They charge for the people who have access (seats), or for the interactions the agent handles (conversations, actions, executions, or credits). They charge for the compute and tokens the agent consumes, or for the agents or bots you deploy. They can also charge for the outcomes the agent delivers, or for a scoped engagement that bundles platform access with implementation. Most vendors combine two or more.

Pricing is unsettled because the old unit does not fit. Software seats assume that a person signs in. An agent that starts on a new order, a ticket, or a schedule has no sign-in to attach a seat to. One page shows the market as it solves this problem. Salesforce sells Agentforce per user, per conversation, and through prepaid Flex Credits, side by side. Buyers can pre-purchase, pre-commit, or pay as they go.

If you still need to decide what counts as an agent, start with what an AI agent is. If a quote is on your desk now, continue here.

The six AI agent pricing models at a glance

Every price below is a list price from the vendor's own pricing page, as of September 2026. Negotiated enterprise prices differ. Vendors change these pages often, so check the source before you use a figure.

ModelYou pay forPublished example (as of September 2026)Budget risk
Per seatEach named person with accessMicrosoft 365 Copilot, $30.00 user/month paid yearly (source). Agentforce add-ons, $125 user/month (source).Pays for access only. Autonomous and unlicensed usage often has a separate meter.
Per interactionEach conversation, action, execution, or creditAgentforce, $2 per conversation or $500 per 100,000 Flex Credits (source). Copilot Studio, $200 per 25,000-credit pack/month (source). n8n Pro, 50€/month for 10K executions (source).Unit definitions differ by vendor. Agent design multiplies consumption. Capacity limits can stop work.
ConsumptionCompute, memory, calls, and tokensAmazon Bedrock AgentCore Runtime, $0.0895 per vCPU-hour, model inference billed separately (source).The bill has many small meters and follows traffic and design. You own the engineering.
Per agent or botEach deployed bot or workerPower Automate Process, $150.00 bot/month paid yearly (source).The price is predictable. Throughput per bot and the build work sit outside the license.
Outcome-basedA defined resultIntercom Fin, $0.99 per outcome (source).The cost turns on how the vendor defines and counts an outcome.
Platform + implementationPlatform access plus scoped engineeringUiPath Standard and Enterprise, contact sales (source). n8n Enterprise, contact sales (source).No public benchmark exists. The scope decides the quote.

The next section takes each model in turn: how it works, where it fits, and what to watch.

The six pricing models, explained

1. Per seat: you pay for each person with access

Seat pricing charges a fixed monthly fee for each named user, usually on an annual term. As of September 2026, Microsoft lists Microsoft 365 Copilot at $30.00 per user per month, paid yearly. That price includes Copilot Chat and access to build in Copilot Studio. Salesforce lists Agentforce add-ons at $125 per user per month for unmetered Agentforce usage by employees. It lists Agentforce Industries add-ons at $150 per user per month. Agentforce 1 Editions start at $550 per user per month and include 2.5 million Flex Credits per org per year.

Where it fits: employee-facing use, where usage roughly tracks the number of people. What to watch: a seat pays for access only, and the inclusion has edges. Microsoft's billing rates page says that employee-facing agent usage by licensed Microsoft 365 Copilot users has no charge within fair usage limits. But agent flows that other triggers start consume Copilot Credits at the standard rate. Computer-Using Agents are not included. The autonomous work is where the second meter starts.

2. Per interaction: conversations, actions, executions, or credits

Interaction pricing charges per unit of activity, and the unit is what varies. Salesforce offers Agentforce at $2 per conversation. It also sells Flex Credits at $500 per 100,000 credits. An Agentforce action costs 20 Flex Credits, and an Agentforce Voice action costs 30. At list price, that is $0.10 per standard action. Salesforce also sells an Agentforce User License at $5 per user per month that requires Flex Credits, so seat and credit meters run together.

Microsoft Copilot Studio sells prepaid packs of 25,000 Copilot Credits at $200.00 per pack per month. An up-front purchase saves up to 20%. A pay-as-you-go option bills monthly through an Azure subscription. Each feature consumes credits at a different rate, according to Microsoft's billing rates. A classic (authored) answer costs 1 credit. A generative answer costs 2, an agent action costs 5, and tenant graph grounding costs 10.

Workflow tools use the same logic with a different unit. n8n prices on monthly workflow executions, regardless of complexity. We checked its pricing page in September 2026. Starter was 20€ per month, billed annually, for 2.5K executions. Pro was 50€ per month for 10K, and Business was 667€ per month for 40K executions, self-hosted. Enterprise was contact sales.

Where it fits: variable, measurable work, such as customer conversations or event-triggered tasks. What to watch: the units of two vendors are rarely comparable. The traffic and the design of the agent together decide how many units a task burns. The worked example below shows how far design can move a bill. Ask what happens when prepaid capacity is exhausted. In Copilot Studio, custom agents stop at 125% of prepaid capacity unless you configure pay-as-you-go. New agent flow runs stop when the tenant uses all of its prepaid capacity.

3. Consumption: compute, memory, calls, and tokens

You meet consumption pricing when you build agents on cloud infrastructure. Amazon Bedrock AgentCore lists Runtime at $0.0895 per vCPU-hour and $0.00945 per GB-hour. It charges only for active consumption, so I/O wait with no background process incurs no CPU charge. Gateway API invocations cost $0.005 per 1,000, and short-term memory costs $0.25 per 1,000 new events. Identity costs $0.010 per 1,000 token or API key requests when used outside Runtime or Gateway. Bedrock bills the underlying model inference separately.

Tokens sit under most meters, even when they are not the headline unit. In Copilot Studio, an agent that uses a reasoning model pays the feature rate plus a premium rate. That premium is 10 Copilot Credits per 1,000 tokens, per Microsoft's billing documentation.

Where it fits: engineering teams that build and run their own agents and want to pay for exactly what runs. What to watch: the price sheet is a set of small meters, and your total is the product of all of them. The vendor meters the infrastructure. The engineers who design, secure, and maintain what runs on it are not on the page at all.

4. Per agent or bot: you pay for each deployed worker

Per-agent pricing charges a flat fee for each deployed bot or digital worker. This is the long-standing unit of robotic process automation. As of September 2026, Microsoft Power Automate lists Power Automate Process at $150.00 per bot per month. Power Automate Hosted Process is $215.00 per bot per month, and both are paid yearly. Power Automate Premium is $15.00 per user per month. UiPath lists a Basic Automation Cloud plan from $25 per month and lists its Standard and Enterprise plans as contact sales. UiPath says that users and robots are available for purchase in all three tiers.

Where it fits: steady, always-on workloads, and finance teams that want a number that does not move with traffic. What to watch: a per-bot price tells you nothing about how much one bot can do. The process design, the build, and the upkeep are separate work. For how that upkeep differs between scripted bots and reasoning systems, see RPA vs AI agents.

5. Outcome-based: you pay for a defined result

Outcome pricing charges only when the agent delivers a defined result. Intercom prices its Fin AI Agent at $0.99 per outcome as of September 2026. Intercom counts an outcome when a customer confirms that the issue is resolved. It also counts one when the customer asks for no more help after Fin responds. The third case is when Fin completes a workflow that Intercom calls a Procedure, including handoffs.

Intercom charges once per conversation, even if Fin answers several questions. Fin needs no seats when you use it with an existing helpdesk. On Intercom's own plans, seats cost $29, $85, or $132 per seat per month.

Where it fits: work with a countable, attributable result, which is why customer support leads here. What to watch: the definition is the contract. Ask how the vendor tells a customer who got help from one who stopped replying. Ask whether you can audit the count. Outcomes are harder to define for cross-system back-office work, where several systems and people touch one result.

6. Platform plus implementation: a scoped engagement

The last model has no number on the website. The vendor scopes an engagement that bundles platform access with configuration, engineering, and support. It quotes the engagement after discovery. UiPath lists Standard and Enterprise as contact sales, and n8n lists Enterprise as contact sales. Salesforce also tells buyers on its pricing page to contact sales for details.

Where it fits: regulated, cross-system, or on-premises work, where scope drives cost more than traffic does. What to watch: you cannot benchmark the price from the outside, so the burden moves to the scope document. Insist that the scope names what is included, what success looks like, and what the price becomes after the pilot. Mission Control prices this way. The structure is in the section below.

Hybrid pricing is the norm

Few vendors run one meter. Salesforce's $5 user license requires Flex Credits. Intercom combines seats with per-outcome charges. Copilot Studio includes some employee usage with a Microsoft 365 Copilot license and meters the rest in credits. When you compare quotes, list every meter that each vendor uses, and include the meters behind the headline one.

A worked example: how design moves the bill

Unit prices are easy to compare and nearly useless on their own. The number of units that a task consumes decides the bill, and the design of the agent decides that number. Here are two examples, worked from published rates. The arithmetic is ours, at list price, before discounts.

Same traffic, different design: Copilot Studio

Microsoft's billing documentation includes a customer support example. An agent serves 900 customers a day, and each run uses four classic answers and two generative answers. The agent consumes 7,200 Copilot Credits a day. Over 30 days that is 216,000 credits, or nine prepaid packs at $200 each: $1,800 a month.

Now keep the traffic and change the design. Suppose the same agent answered all six questions generatively and grounded each one in the tenant graph. Each run would then cost 72 credits: six answers at 2 credits plus six groundings at 10. That is 64,800 credits a day, 1,944,000 over 30 days, or 78 packs: $15,600 a month. The customers and the questions stay the same, and the agent uses nine times the credits.

Same vendor, different meter: Agentforce

At Salesforce list prices, a conversation costs $2 on the per-conversation meter. A standard action costs $0.10 on Flex Credits (20 credits at $500 per 100,000). The two meters cost the same when a conversation averages 20 actions. Below 20 actions, Flex Credits cost less per conversation. Above 20, the per-conversation meter costs less. You cannot pick the cheaper meter until you know how many actions your real conversations take, and only a pilot can tell you that.

The lesson applies to every vendor. Ask for a consumption estimate on your own workload, designed the way you will actually run it. Then repeat the estimate at three and ten times today's volume.

AI agent pricing examples, as of September 2026

This table puts the same figures in one place. All are list prices from each vendor's own pricing page, accessed September 2026, in the currency and terms shown there.

Vendor / productModelPublished price (as of September 2026)Source
Microsoft 365 CopilotPer seat$30.00 user/month, paid yearly.microsoft.com
Microsoft Copilot StudioCredits$200.00 per pack/month for 25,000 Copilot Credits. Pay-as-you-go is available.microsoft.com
Salesforce AgentforcePer conversation$2 per conversation.salesforce.com
Salesforce AgentforceCredits$500 per 100,000 Flex Credits, with 20 credits per action and 30 per Voice action.salesforce.com
Salesforce Agentforce add-onsPer seat$125 user/month. Agentforce 1 Editions from $550 user/month.salesforce.com
Microsoft Power AutomatePer bot / per seatProcess $150.00 bot/month, Hosted Process $215.00 bot/month, Premium $15.00 user/month, all paid yearly.microsoft.com
UiPath Automation CloudPlatform (tiers)Basic from $25/month. Standard and Enterprise: contact sales.uipath.com
n8nPer executionStarter 20€/month (2.5K executions), Pro 50€ (10K), Business 667€ (40K, self-hosted), billed annually. Enterprise: contact sales.n8n.io
Amazon Bedrock AgentCoreConsumptionRuntime $0.0895 per vCPU-hour and $0.00945 per GB-hour, with model inference billed separately.aws.amazon.com
Intercom Fin AI AgentOutcome-based$0.99 per outcome, charged once per conversation.intercom.com
Mission Control (Swarm)Platform + implementationScoped per pilot. Contact sales.usemissioncontrol.com

Total cost of ownership: the lines beyond the license

A price sheet shows the meter. It does not show the cost to get a system into production and keep it there. Benchmarkit and Mavvrik surveyed 372 organizations, and CFO Dive reported the results in September 2025. In that survey, 85% of companies missed their AI cost forecasts by 10% or more, and 24% missed by more than 50%. The top drivers of unexpected cost were data (acquisition, storage, cleaning, and security) and hidden costs from historical business practices and legacy systems. Neither appears on a pricing page.

Build the total cost of ownership from these lines:

Cost lineWhat drives itQuestion that exposes it
Licensing and usageSeats, credits, conversations, executions, bots, outcomes.What is the unit, what is included, and what is the overage rate?
Inference and infrastructureModel tokens billed outside the platform meter, and hosting if you self-host.Is model inference inside the price, or a separate bill?
ImplementationScoping, configuration, process capture, and the time of your subject matter experts.Who does the work, for how many weeks, and whose people?
IntegrationIdentity, connectors, legacy systems without APIs, data preparation.Which of our systems are covered, and which are our problem?
Security and governanceSecurity review, audit logging, approvals, access reviews.Which controls ship in the runtime, and which do we build?
Maintenance and changeTool, model, and interface changes, and the retesting after each.Who fixes it when a system it uses changes, and at whose cost?
Overage and continuityUsage above prepaid capacity, and what stops when the capacity ends.Does work stop, slow, or continue to bill at a higher rate?
ExitMoving configurations, logs, and learned procedures elsewhere.What can we take with us, in what format?

Costs that sit outside the meter

Two published examples show why the license and the bill are different numbers. Amazon states on its AgentCore pricing page that Bedrock bills model inference separately from AgentCore. Microsoft's Copilot Studio billing rates exclude bring-your-own-model configurations, including Azure Foundry models, which Microsoft bills separately. Neither cost is hidden. Both are easy to miss when a spreadsheet has a column only for the platform.

Implementation, integration, and maintenance

The build is the line that surprises buyers most often. A consumption price assumes that someone already designed the agent, connected it to your systems, and secured it. A per-bot price assumes that someone already mapped and scripted the process. Price that work honestly, and include the hours your own experts spend to teach and review. If you are weighing a build on open-source agent frameworks, count the governance that you will have to build and maintain yourself. The AI agent security checklist lists the controls to price.

Maintenance is the line that grows. Tools change, vendors upgrade models, interfaces move, and every change needs retesting. If an outside provider does the work you are pricing today, compare against that provider's full run-rate too. Synthetic workers vs managed service providers shows that comparison.

Deployment model changes who pays

Where the system runs moves costs between you and the vendor. A vendor-hosted service includes infrastructure in the meter. A self-hosted deployment, such as n8n Business, moves hosting, patching, and monitoring to your team. On AgentCore, the instance option bills the EC2 instance cost plus a management fee of 12% of the On-Demand rate. Your data may fall under a regime that keeps it inside your perimeter. If so, settle that before you compare prices, because it excludes some pricing models entirely.

Questions to ask before you sign

Take these questions into every vendor call. A vendor that answers them plainly, in writing, is easier to budget for than a cheaper one that does not. For the capability side of the evaluation, pair this list with how to evaluate AI agents for the enterprise.

About the meter

  • What exactly is the billing unit, and where is its definition written?
  • How many units does one of our typical tasks consume, designed the way we will actually run it?
  • Which features consume at higher rates (grounding, reasoning models, voice, computer use)?
  • What does the estimate look like at three and ten times our current volume?

About limits and overage

  • What happens when prepaid capacity ends: does work stop, queue, or bill at an overage rate?
  • Can we cap spend per agent or per environment, and who gets a notice before spend reaches a cap?
  • For outcome pricing: how do you count an outcome, and can we audit the count?

About everything outside the meter

  • Is model inference included, or billed separately? Can we bring our own model, and what does that cost?
  • What implementation work is included, who does it, for how long, and what do our people have to do?
  • Which of our systems do the included integrations cover, and which need custom work?
  • Which security controls ship in the product, and which would we have to build?
  • Who pays for fixes when a connected system or a model changes?

About the contract

  • What is the price after the pilot or first term, and on what unit?
  • Which commitments are annual, and what happens to unused capacity?
  • If we leave, what do we keep: configurations, logs, and the procedures the system learned?

How Mission Control prices synthetic workers

Here is how the vendor that wrote this guide prices, so you can ask us the same questions. Mission Control builds synthetic workers, which run on the Swarm platform. Mission Control does not publish list prices. We scope each engagement as a pilot, so pricing is a conversation with sales. In the six-model framework above, Mission Control is model 6, platform plus implementation.

The price is scoped, but the structure is the same for every engagement. The synthetic workers page publishes it, in this order:

  • Scope and configure: Select the use case, ingest your documents and standards, and configure the worker.
  • Single-user learning loop: Your designated expert works with the synthetic worker. The worker learns, converges, and improves.
  • Cross-user validation: A second user starts from a higher baseline. The knowledge transfers without degradation.
  • Results: Analysis, documentation, and go-forward recommendations.

The pilot includes platform access, worker configuration, all engineering support, and a forward-deployed team that works inside your organization for the whole pilot. This design answers several of the TCO questions above. Implementation and engineering are inside the engagement, with no separate services line. The plan books the expert's time in the learning loop from the start, so that cost does not appear later as a surprise.

Here are the limits, stated the way you should expect any vendor to state them. Mission Control has no public list price, so you cannot benchmark it from its website. You get a number after scoping. A pilot with an embedded team is a heavier commitment than a credit pack that you buy today and use tomorrow. Ask us about post-pilot pricing, units, and inference in your scoping conversation, and get the answers in writing. Ask every other vendor the same questions in the same way.

Before you sign: a one-page pricing sheet

AI agent pricing looks like a choice between meters. In practice it is a forecast of how your work will run. The forecast is only as good as the workload you test it on. Before you sign anything, complete one page for each vendor:

  • Unit and definition: the billing unit, quoted from the vendor's own documentation.
  • Your consumption: units per task for your real workload, designed the way you will run it.
  • Stress test: the monthly bill at today's volume, three times, and ten times, including overage rates.
  • When capacity ends: whether work stops, queues, or continues to bill.
  • Outside the meter: inference, infrastructure, integration, security, and maintenance, each with an owner and an estimate.
  • After the first term: the price and unit when the pilot or discount ends.
  • Exit: what you keep if you leave.

If you can complete the page for a vendor, you can budget for that vendor. To complete it for Mission Control against your own work, talk to our team.

References

Common questions.

How much does an AI agent cost?

The cost depends on what the vendor meters. As of September 2026, published list prices include Microsoft 365 Copilot at $30 per user per month, paid yearly. Microsoft Copilot Studio costs $200 per month for a pack of 25,000 Copilot Credits. Salesforce Agentforce costs $2 per conversation, and Intercom Fin costs $0.99 per outcome. Many enterprise platforms list their larger plans as contact sales. The license is only part of the cost. Implementation, integration, inference billed outside the meter, governance, and maintenance add to it.

What are the main AI agent pricing models?

Six models are in use. Vendors charge per seat (each named user) or per interaction (a conversation, action, execution, or credit). They charge for consumption (compute, memory, calls, and tokens) or per agent or bot. Outcome-based pricing charges for a defined result, such as a resolved conversation. The sixth model is a scoped platform-plus-implementation engagement. Most vendors combine two or more, for example a user license that also requires credits.

Is outcome-based pricing better for buyers?

It can be, when outcomes are countable and clearly attributable, as in customer support. The risk sits in the definition. Ask exactly what counts as an outcome. Ask how the vendor tells a customer who got help from one who stopped replying, and whether you can audit the count. For cross-system back-office work, where several systems and people touch one result, outcome definitions are harder to agree on.

What hidden costs should I budget for with AI agents?

Budget for the lines outside the price sheet. These include model inference that some platforms bill separately, infrastructure if you self-host, and implementation and process capture. Add integration with identity and legacy systems, data preparation, and security and governance work. Add ongoing maintenance as tools and models change, and overage when usage passes your prepaid capacity. Benchmarkit and Mavvrik surveyed 372 organizations, and CFO Dive reported the results in September 2025. In that survey, 85% missed their AI cost forecasts by 10% or more.

How does Mission Control price synthetic workers?

Mission Control does not publish list prices. We scope each engagement as a pilot, so contact sales for pricing. Every engagement follows the same structure. First comes scoping and configuration, then a single-user learning loop, then cross-user validation. The pilot ends with analysis, documentation, and go-forward recommendations. A forward-deployed team works inside your organization for the whole pilot.

Read next.

Start in one shop.

Two Bob units

Head mount, chest mount, or one of each.

The decryption computer

It decrypts each recording into your own cloud.

Swarm

Installed in your own environment.

Start with Record. Add Live and Co-Engineer when your team asks for them.

Get a demo