What AI Workflow Automation Actually Costs

Before comparing AI workflow automation options, it helps to separate the two cost components clearly, because they behave differently and scale differently.

Platform cost

The platform cost is what you pay for the orchestration layer: the workflow editor, the agent configuration, the human review system, the task inbox, the audit log, the role-based access controls, and the API. This cost is typically a subscription, either per seat or per usage tier. It does not change based on how many tokens your agents consume.

Model cost

The model cost is what you pay the AI providers for the actual inference: the tokens sent and received by your agents on every workflow run. This cost scales directly with usage volume. More workflow instances, larger context windows, more agent calls per instance: all of these increase token consumption and therefore model cost.

Why the separation matters

On platforms that proxy your model calls and charge a markup, the two costs are merged into one bill. The platform's revenue grows as your AI usage grows. This creates a structural misalignment: the more value you extract from automation, the more you pay the platform, not just the model provider. On DAVE, the two costs are separate. Your platform bill is determined by your seat count. Your model bill is determined by your usage, paid directly to the provider. Scaling your automation volume does not increase your platform bill.

DAVE is priced by seat count, with a 30-day free trial for up to five seats. No credit card is required to start the trial. After the trial, pricing is:

• Starter: $20 per month. Includes 1 seat.
• Team: $50 per month. Includes 5 seats.
• Business: $250 per month. Includes 25 seats.
• Beyond 25 seats: A flat $10 per seat per month, added to the Business plan.

A seat is a user account in your DAVE workspace. DAVE ships with six default tenant roles: Admin, Create, Curate, Use, Reporting, and Financial. All roles are editable by your administrator. A user can hold multiple roles simultaneously. The role determines what the user can do: Admins configure the workspace, Create users build agents, Curate users design workflows, Use users run instances and complete tasks, Reporting users access analytics, and Financial users have access appropriate to financial oversight.

The platform cost covers everything in the orchestration layer: the workflow editor, agent configuration, human review and task inbox, role-based access controls across 20 permission types, full audit logging, per-tenant encryption using AES-256-GCM authenticated encryption, SOC 2 and GDPR compliance modes, MCP integration for AI assistants, and the full 305-endpoint REST API. There are no per-instance fees and no storage quotas.

Your billing plan and seat count are managed at Admin, then Billing in the sidebar. Only the Admin role can access the Billing section.

Your model cost depends on three variables: which providers you use, which models you select for each agent, and how many tokens your workflows consume per run.

Provider and model selection

DAVE connects to seven provider types: OpenAI, Anthropic, Azure OpenAI, Google AI, OpenAI-Compatible, Local (Ollama), and Custom. Each provider has its own pricing, published on their website. Model prices vary significantly within a single provider: a fast, lightweight model may cost a fraction of a cent per million tokens, while a frontier reasoning model may cost many times more. The model you assign to each agent in DAVE determines which price applies to that agent's calls.

Because DAVE does not proxy or mark up AI usage, the price you pay is exactly the price the provider publishes. You can verify this by comparing your provider's billing dashboard with DAVE's workflow run logs.

Token consumption per workflow run

Each time a workflow instance reaches an Agent Interaction node, DAVE calls the configured provider and model. The number of tokens consumed depends on the system prompt length, the context passed to the agent (documents, prior outputs, retrieved records), and the length of the agent's response. A simple extraction task on a short document might consume a few thousand tokens. A complex analysis task with a large context window might consume tens of thousands.

To estimate your monthly model cost:

1. Identify the agents in your workflow and the model assigned to each.
2. Estimate the average token consumption per agent call (input tokens plus output tokens).
3. Multiply by the number of workflow instances you expect to run per month.
4. Apply the provider's published per-token price for each model.
5. Sum across all agents and all workflows.

This is an estimate, not a guarantee. Real token consumption varies with input length and agent behavior. Start with a conservative estimate, run a representative sample of instances, and check your provider's billing dashboard for actual consumption figures before committing to a volume projection.

Local models and cost

If you connect a local Ollama instance as a provider, the per-call API cost for that provider is zero: you bear the infrastructure cost of running the model, not a per-token charge. For high-volume, well-defined tasks (classification, extraction, formatting), a local open-weight model can significantly reduce the model cost component of your total bill. The platform cost is unchanged.

Total cost of AI workflow automation on DAVE equals your platform subscription plus your model spend. Here is a practical framework for estimating both before you commit.

Step 1: Count your seats

A seat is a user who needs a DAVE account. Count the people who will build workflows (Curate role), build agents (Create role), run instances and complete tasks (Use role), view reports (Reporting role), and administer the workspace (Admin role). One person can hold multiple roles and counts as one seat. Match your seat count to the appropriate plan tier.

Step 2: Map your workflows to agent calls

For each workflow you plan to build, list the Agent Interaction nodes and the model you intend to assign to each. This gives you a map of which models are called, how many times per workflow run, and which provider each call goes to.

Step 3: Estimate monthly instance volume

How many times per month will each workflow run? For some workflows (a weekly report, a monthly invoice batch), this is straightforward. For others (a customer-triggered onboarding flow, a content pipeline driven by submissions), you need to estimate based on your current volume or your growth target.

Step 4: Estimate token consumption per instance

For each agent call in your workflow map, estimate the average token count: system prompt tokens plus input context tokens plus output tokens. If you have existing prompts, count them. If you are starting fresh, use a conservative estimate and refine after your first production runs.

Step 5: Apply provider pricing

Look up the current per-token price for each model on your provider's pricing page. Multiply token consumption by price, sum across all agent calls in a workflow, then multiply by monthly instance volume. Sum across all workflows for your total monthly model spend estimate.

Step 6: Add platform cost

Add your DAVE subscription cost based on your seat count. The sum of platform cost and estimated model spend is your projected total monthly cost for AI workflow automation on DAVE.

Run this calculation before your trial ends. The 30-day free trial gives you enough time to build a representative workflow, run a meaningful sample of instances, and check your provider's billing dashboard for actual token consumption. Use real consumption data to refine your estimate before you choose a plan.

Three things that commonly appear in AI platform pricing are absent from DAVE's cost structure. Understanding what is not in the cost is as important as understanding what is.

No per-instance fees

DAVE does not charge per workflow instance run. You can run as many instances as your workflows require without a per-execution fee. The platform cost is flat for your seat tier. Your model cost scales with usage, but that cost goes to the provider, not to DAVE.

No storage quotas

DAVE does not impose storage quotas on workflow definitions, instance records, audit logs, or task payloads. You are not charged for storing workflow history or audit data.

No AI usage markup

DAVE does not sit between you and your model providers and charge a percentage on top of their prices. You bring your own API keys, connect them directly to the provider types you choose, and pay the provider at their published rate. DAVE's revenue comes from the platform subscription, not from a cut of your model spend. This means your model cost is transparent, auditable, and fully under your control.

These three absences are not incidental. They are the result of a deliberate architecture: a flat platform subscription for the orchestration layer, and a direct billing relationship between you and your model providers for the inference layer. The two costs are independent, visible, and predictable.