AWS has three core automation primitives that most business teams either do not know about or only partially understand: Lambda, Step Functions, and EventBridge. Together they can replace Zapier, Make, and a significant amount of custom scheduling code — at a fraction of the per-task cost and with substantially better reliability at scale.
This post explains what each service does, when to use each, and what it actually costs to build and run common business automation patterns on AWS.
The Three AWS Automation Primitives
Before comparing them, it helps to understand what problem each solves distinctly.
Lambda is compute without servers. You write a function, define what triggers it (an HTTP request, a file upload, a schedule, a queue message), and AWS runs it. You pay per invocation and per GB-second of execution time. There are no servers to provision or maintain. Lambda scales from zero to thousands of concurrent invocations automatically.
Step Functions is a workflow orchestrator. It manages state between steps, handles errors and retries, supports parallel branches, and provides a visual representation of your workflow execution. Think of it as the glue that connects multiple Lambda functions (and other AWS services) into a reliable multi-step process.
EventBridge is an event bus. It receives events from AWS services (S3, RDS, CloudTrail, etc.) and from external systems, matches them against routing rules, and delivers them to targets. EventBridge decouples the thing that produces an event from the thing that responds to it.
These three services work together rather than competing. A typical pattern: an external event arrives via EventBridge, triggers a Step Functions workflow, and each step in the workflow calls a Lambda function to do the actual work.
Lambda: Pricing and Real-World Cost
Lambda has two cost dimensions: request count and compute duration.
| Dimension | Free tier | Beyond free tier |
|---|---|---|
| Requests | 1M/month free | $0.20 per 1M requests |
| Compute (512 MB) | 400,000 GB-seconds/month free | $0.0000166667 per GB-second |
What does that mean in practice? A function allocated 512 MB running for 200 milliseconds on average, called 1 million times per month, costs approximately $1.67 in compute. The first million requests per month are free. A modestly busy business automation function running 5 million times a month costs roughly $8–10 per month total.
The important limitation: Lambda has a maximum execution time of 15 minutes. It is not suitable for long-running processes like generating a large PDF report, running a database migration, or processing a large file. For those use cases, Lambda triggers an ECS Fargate task or a Batch job instead.
Step Functions: Pricing and Real-World Cost
Step Functions has two workflow types with different pricing models suited to different volumes.
| Type | Pricing | Best for |
|---|---|---|
| Standard | $0.025 per 1,000 state transitions | Long-running workflows, audit trail required, human approvals |
| Express | $0.00001 per state transition | High-volume, short-duration workflows (<5 min) |
A practical example: a 10-step Standard workflow running 10,000 times per month generates 100,000 state transitions, costing $2.50/month. The same workflow on Express costs $0.01. Express workflows are designed for high-throughput real-time processing. Standard workflows provide full execution history and are auditable, which matters for compliance use cases.
The real cost of Step Functions is not execution: it is development time. Designing a robust state machine, handling error states cleanly, and building in compensation logic for failed steps takes an experienced engineer two to five days for a moderately complex workflow.
EventBridge: Pricing and Real-World Cost
EventBridge pricing is straightforward: $1 per million custom events published. The first 100,000 events per month from most AWS services are free. Pipes, which provide point-to-point integrations between sources and targets, cost $0.40 per million events.
For a typical business automation setup — CRM webhook events, S3 upload notifications, scheduled rule triggers — EventBridge costs are negligible. A team processing 500,000 events per month pays $0.50 in EventBridge costs. The value is not in cost savings relative to alternatives; it is in the architectural cleanliness of decoupled event routing.
Common Business Automation Patterns: What They Cost
Here is what typical business automation use cases cost to run on AWS each month, once built.
| Pattern | Services | Monthly cost estimate |
|---|---|---|
| CSV file processing (upload → parse → write to DB) | S3 + Lambda + RDS | $5–20 |
| Nightly report generation and email delivery | EventBridge + Lambda + SES | $1–5 |
| Multi-step approval workflow with email notifications | Step Functions + SES + Lambda | $5–15 |
| Real-time event processing (webhook → enrich → write) | EventBridge + Lambda + DynamoDB | $10–50 |
| API integration pipeline (poll source, transform, sync target) | EventBridge + Lambda + SQS | $5–25 |
These running costs are dramatically lower than equivalent Zapier or Make subscriptions at volume. The difference is that AWS requires a developer to build the initial solution, which has its own cost. The economics favor AWS at sustained automation volume; they favor no-code tools for simple, infrequent workflows or when non-technical operators need to own maintenance.
When to Use Each Service
Lambda alone: single-step event processing
Lambda is the right tool when something happens and you want to respond with a single action. A file is uploaded to S3: resize and compress it. A webhook arrives from your payment processor: write the event to a database. A scheduled trigger fires every night at 10pm: pull today's sales data and send a summary email.
Lambda alone is reliable for these patterns because the failure mode is simple: either the function runs successfully or it fails and you can configure automatic retries. There is no complex state to manage between steps.
Lambda alone breaks down when you need to know whether step three succeeded before running step four, or when you have parallel branches that need to complete before proceeding. That is when Step Functions earns its cost.
Step Functions: multi-step orchestration
Step Functions is the right tool when your workflow has dependencies between steps, needs guaranteed ordering, or requires visibility into exactly where an execution failed. A purchase order approval workflow is a classic example: receive purchase request, check budget availability, route to appropriate approver, wait for approval, update ERP, send confirmation. Each step depends on the previous; failure at any point needs to route to a compensation path.
Step Functions provides a visual execution graph that shows exactly which step is running, which have completed, and what the output of each step was. For debugging and auditing, this is invaluable. Building equivalent observability into a chain of Lambda functions with SQS queues is possible but takes significantly more engineering time.
Step Functions also handles human-in-the-loop patterns via callbacks. A workflow can pause at an approval step, send an email with an approve/reject link, and resume when the human acts. Building this pattern with raw Lambda requires persistent state management that Step Functions provides out of the box.
EventBridge: event routing and decoupling
EventBridge is the right tool when you want to decouple event producers from event consumers. Your CRM fires a contact-updated event: EventBridge routes it to a Lambda that syncs the contact to your billing system, a separate Lambda that updates your support ticketing system, and a Step Functions workflow that triggers your onboarding sequence. The CRM does not need to know any of that exists.
EventBridge is also the right tool for scheduled automations via EventBridge Scheduler, which replaced CloudWatch Events and is more reliable and easier to manage at scale. Cron expressions, rate-based schedules, and one-time schedules are all supported.
EventBridge Pipes provide point-to-point integrations without writing code: read from a source (Kinesis stream, SQS queue, DynamoDB stream), optionally filter or enrich, deliver to a target. For simple data pipeline patterns, Pipes eliminate the need for a Lambda function entirely.
Development Cost to Build AWS Automation
Running costs are almost irrelevant at typical business automation volumes. The meaningful cost is build time. Here is an honest range by project type:
Simple Lambda function with one trigger: $1,000–3,000. A developer sets up the Lambda, configures the trigger (S3, EventBridge, API Gateway), writes the handler, adds error handling and logging, and deploys via Infrastructure as Code (Terraform or CDK). A developer who has done this before can complete a straightforward function in one to three days.
Step Functions workflow with five to ten steps: $5,000–15,000. Each step requires a Lambda function with its own error handling, the state machine definition, IAM permissions for each transition, and testing of the full happy path plus major failure scenarios. A robust 10-step workflow typically takes a week to two weeks to build and test properly.
Full event-driven architecture: $15,000–40,000. Multiple EventBridge rules, several Step Functions workflows, a dozen or more Lambda functions, SQS queues for reliability, DynamoDB for state storage, and CloudWatch alarms for monitoring. This is a system, not a script, and it requires architecture decisions that affect long-term maintainability.
AWS vs Zapier and Make: Honest Comparison
Zapier and Make solve the same class of problems. The honest comparison is not about features but about where each excels.
Where AWS wins over Zapier/Make:
- Volume cost. Zapier at 100,000 tasks/month costs $299–599/month. AWS costs $10–30/month for the same volume. The economics flip decisively at scale.
- Data residency. When automation logic touches sensitive data, running it in your own AWS account with VPC isolation is materially more secure than routing data through a third-party SaaS platform.
- Complexity ceiling. Zapier's branching logic, multi-step error handling, and custom code support are improving but remain constrained. Lambda has no ceiling: you can write any code you want.
- Integration with your own systems. Lambda can call your internal APIs directly over a private network. Zapier requires your systems to be publicly accessible.
Where Zapier/Make wins over AWS:
- Time to first automation. A Zapier automation can be running in 20 minutes without writing code. An equivalent Lambda function requires setup, deployment, and testing.
- 6,000+ pre-built connectors. Zapier's library of connectors handles authentication, pagination, and API quirks for thousands of tools. Replicating even one of those connectors in Lambda is several hours of work.
- Non-technical maintainability. When a non-technical operator needs to change the filter condition in a Zapier automation, they do it themselves in the UI. Changing the same logic in Lambda requires a developer and a deployment.
- Lower upfront investment. If you need five simple automations and your volume is low, the Zapier subscription is cheaper than hiring a developer to build the AWS equivalent.
The practical guidance: use Zapier or Make for automations that non-technical operators will modify, involve standard SaaS tools with existing connectors, and run at low enough volume that per-task pricing is acceptable. Use AWS for automations that process sensitive data, run at high volume, require complex logic, or need to integrate with your internal systems over private networks.
A Concrete Example: Nightly Sales Report
A professional services firm needed a nightly report: pull closed deals from their CRM via API, calculate rep performance against quota, format a summary, and email it to leadership at 7am every weekday.
On Zapier, this runs $99–299/month depending on the task volume and multi-step complexity. The automation breaks occasionally when the CRM API returns an unexpected error, and there is no visibility into what happened without digging through Zapier's task history.
On AWS, the equivalent uses EventBridge Scheduler to trigger a Lambda function at 6:45am on weekdays. Lambda calls the CRM API, processes the response, generates the HTML report, and uses SES to send the email. Running cost: under $2/month. When the CRM API returns an error, Lambda retries automatically, and a CloudWatch alarm sends an alert if the function fails three times in a row.
Development time to build the AWS version: two days. At a developer rate of $120/hour, that is roughly $1,900 in build cost. Break-even against Zapier at $149/month is 13 months. For a report the firm plans to run indefinitely, AWS is the better long-term choice.
Project summary:
Use case: Nightly CRM report emailed to leadership
Build: 2 days · Cost: ~$1,900 one-time
Running cost: <$2/month vs $149/month on Zapier · Break-even: 13 months
Frequently Asked Questions
When should I use Lambda vs Step Functions?
Use Lambda alone when your automation is a single step triggered by an event: a file is uploaded, a webhook arrives, a schedule fires. Lambda handles the action and returns. Use Step Functions when your workflow has multiple sequential or parallel steps where each step depends on the previous result, you need retry logic between steps, or you have human approval steps mid-workflow. Step Functions is a state machine that orchestrates Lambda functions and other AWS services. The cost is higher but the observability and reliability for multi-step workflows is significantly better than chaining Lambda functions manually through SQS queues.
How much does AWS automation cost compared to Zapier?
At low volume, Zapier is often cheaper because AWS has learning curve and setup costs that exceed the software cost difference. At high volume, AWS wins decisively. Zapier's Business plan at $299/month includes 50,000 tasks. The same 50,000 tasks on Lambda plus EventBridge costs under $10/month in execution. The break-even depends on developer time: if you spend 20 hours building the AWS equivalent at $100/hour, that is $2,000 in setup cost. The math favors AWS at sustained volumes above roughly 20,000 tasks per month and when workflow complexity exceeds what Zapier can express cleanly.
Can AWS Lambda replace Zapier for business automation?
Lambda can replace Zapier for most automation use cases, but not blindly. Zapier's value includes 6,000+ pre-built connectors, a no-code interface that non-technical operators can modify, and instant setup. Lambda requires a developer to write and deploy code for every integration. If the operations team needs to change automation logic without developer involvement, Lambda is the wrong choice. If technical operators run the automation, the logic is complex, the volume is high, or the data must stay in your VPC, Lambda is the better long-term choice.
Get an Estimate for Your AWS Automation Project
Most business automation use cases are simpler to implement on AWS than they appear, and cheaper to run than Zapier at any meaningful volume. If you have a specific workflow in mind, the fastest path to a real answer is a short conversation about your current process and what you want to automate.
