Thunk.AI is an enterprise-ready platform for human workflow automation using no-code AI agents. It automates workflows that are traditionally human-intensive by enabling the creation and execution of AI-native applications, called thunks.
In this article, you will learn how to build and execute custom thunks.
Step 1: Sign In or Create an Account & Get Onboarded
Sign up for Thunk.AI using your Google Workspace or Microsoft 365 account.
During sign-on, grant Thunk.AI access to your drive. This access allows your AI to read and index files to synthesize relevant information for your projects and generate new files or images as needed. For more details, refer to our Privacy Policy.
Complete onboarding and tell us a bit about yourself. Your AI agent works like a work assistant—it needs to know who you are, your role, and your company. This information helps it tailor its actions to your needs. It’s not sold or used for ads, just to personalize your experience on the platform.
Step 2: Understand Key Concepts
The Thunk.AI mental model centers around "thunks," AI-native applications or workflows that represent a business process. Thunks are first designed and then executed.
Each thunk describes a workflow with steps. Thunk owners provide workflow logic in natural language at design-time using a no-code interface to outline business processes, goals, and outcomes.
Every user who designs or participates in a thunk is paired with an AI agent that assists them in the design and execution of work in a thunk. AI agents leverage LLMs (e.g., GPT-4) to automate tasks and perform meaningful application work within a thunk.
Thunks can define and utilize Content Collections, including documents and images, to provide the AI relevant information for tasks.
AI tools, organized into modules, extend an agent's capabilities, enabling actions like web searching, updating databases or file management.
Thunks also feature an Inbox for routing emails, form submissions, or events from external applications.
In execution, AI agents collaborate with human participants to efficiently process work and complete tasks. The user's AI agent shares the same capabilities, information, and instructions as the user, enabling seamless collaboration to complete work within a thunk.
Hosted on a scalable platform, Thunk.AI supports users building and running any number of thunks to bring AI automation to diverse business processes. Read this article for details on the concepts introduced above.
Step 3: Browse Sample Thunks
Upon login, you'll find a collection of live, intelligent AI applications across various industries—these aren’t just templates; they’re fully functional sample Thunks.
For example, in the Expense Receipt Processing sample, your AI extracts receipt details, checks them against company policy, recommends a decision, and drafts an email—automating the entire process. Watch this video to explore the Thunk and learn how to build one yourself!
You can copy any sample to make it your own, add or remove steps, alter step instructions, add your own documents and try it out on your data. Your AI will help process your data through the sample thunk’s plan.
Step 4: Build your first, custom Thunk
We will use this sample thunk for "Processing incoming purchase orders" as a running example in this tutorial.
Thunks are first designed and then executed. In this section we will cover how to:
4.1 Design your thunk with AI-Driven Planning
To start designing your thunk, prepare information that aligns with your workflow. Using the example of a thunk for "Processing Purchase Orders," you’ll gather the following:
Thunk Goal: Define a concise identifier for your workflow, such as "Process incoming purchase orders."
Workflow Process Description (Optional): Outline your process in detail, as you would explain it to a co-worker. For example: "We receive purchase orders from clients. Each order should include product details, quantities, and payment information. I need to validate the order details, ensure payment terms are met, and forward confirmed orders to the fulfillment team."
Initial Data (Optional): Provide any starting data that can guide your AI in setting up the thunk's structure, such as example purchase orders. The AI can import this data to begin working on it immediately.
To begin planning, Create a New thunk from the thunk homepage.
Enter the Goal: This required information defines what the thunk automates and serves as its identifier.
Specify the human workflow process you want to automate in this thunk. This step is optional—if you skip it, your AI will use your Goal to suggest an initial workflow you can edit.
Next, specify any data you already have (optional). This helps your AI plan what data to track in the thunk and becomes part of the work your AI will start processing.
AI-driven Planning: Your AI will now navigate you through (i) Plan generation (ii) Data Structure identification and (iii) Plan details to build a thunk in a matter of minutes.
(i) Plan generation: Your AI uses information you provided to create a workflow with sequential steps. For now, ensure all necessary steps are included; details will be refined in a subsequent planning step.
You can edit the plan via the UI or by chatting with your AI.
(ii) Data Structure Identification: Your AI determines the key data to track during execution—e.g., invoices, payment amounts, approvals for invoice processing. Column properties, types, and descriptions help ensure accurate outputs. You can refine this structure at any stage.
(iii) Plan Details
Finally, your AI refines the Plan Details for workflow execution. This crucial planning step allows the thunk designer to specify how Steps should be completed. You can do this by clicking Detail Instructions for each step.
Instructions: Edit suggested, natural language directions on how your AI should assist. These can be updated at execution time as well as your AI does work.
Examples: Share examples to clarify the desired outputs, including what to aim for or avoid.
Inputs and Results: Select the properties your AI will use for this step. These choices determine the inputs available and ensure the outputs align with expectations of the Step.
AI Modules: Choose from various platform tools, organized into modules, to enhance your AI's capabilities. Unselected tools will not be available for this step. To add custom tools, refer to the AI Tools section.
Pro Tip: Detailed Step instructions ensure consistent execution and reliable results. Learn more about how Thunk.AI maintains dependable agent behavior.
Proceed through the final steps of planning with your AI and that's it! Your Thunk is live and ready to execute!
4.2 Execute your thunk
Your thunk is now ready to accept work. As work aka data arrives, your AI will process them step by step, following your instructions. After each step, you can review outputs, adjust data, update instructions, and rerun the work if needed. Let’s dive into the details!
How your thunk accepts & processes incoming data
Every Thunk starts by collecting workflow requests. In this Purchase Order Processing Thunk, requests come as purchase orders—added manually (e.g., uploading purchase orders) or triggered automatically (e.g., form submissions, events). To begin, upload work requests to Data.
Integrations & Triggers
Every Thunk has an Email Inbox and Webhook Inbox to receive work requests—emails, form submissions, or external events. Configure them to accept requests and define how your AI processes them.
In all cases, when a request comes in, a new workflow entry is created and your AI begins working on it by taking every Data through every Step of your plan.
Pro Tip: New thunks are optimized for quick onboarding. By default, when data is added, your AI processes it immediately (instead of starting it as a draft) and moves through each step without waiting for approval. You can adjust both settings under AI Governance > Human-in-the-Loop to enable more control.
How your AI Executes a Step
Step Assignment
In Thunk.AI, work is assigned to a team member (human, not AI). If assignment logic is set, the owner’s AI distributes tasks accordingly. Each team member’s AI assists them in completing their steps. Learn how to add users and define assignment logic.
Step Execution
If you own the Step, your AI follows the standing instructions laid out for the step, relies on the inputs provided, utilizes tools available to it, does the work and records results in the appropriate result data properties.
Monitor & Review AI Work
Track your AI in real-time or review completed tasks in the AI tab. For each step, see how it plans, executes, and adapts based on the work state. Once done, results are stored in output properties, with data and work state managed for you!
Steer Your AI
Refine your Thunk by iterating with your AI. Adjust instructions or experiment with tools to shape its work. Re-run a step with the circle icon to apply changes instantly. Edits at the Instruction level apply across all data in the Thunk.
Chat with Your AI
Engage with your AI anytime through the chat interface. Ad-hoc chats on specific data stay isolated—keeping the overall Thunk plan intact.
Finish the Step
Once you're satisfied with the outcome, click "Finish the Step" to advance the workflow—your AI will take it from there!
Congratulations! Your first Thunk is live with your data! Now, explore these resources to further customize and refine it.
4.3 Customize your thunk
User Roles & Work Assignment
Owners can add users using the Team pane of a thunk. There are three user roles within a thunk:
Thunk Owner/Designer – Authors the logic of a Thunk using natural language.
Work Participant – Executes assigned workflow steps with AI assistance.
End User – Interacts with the Thunk by making requests.
Owners can also specify work assignment logic in any step’s instructions and the thunk owner’s AI will utilize it during step assignment. When participants are assigned a step, their AI helps them in its execution.
AI Governance
Every Thunk has customizable controls to fit your team’s automation needs—balancing compliance, human oversight, automation, reliability, and modularity. For example, you can allow AI to complete tasks without approval. Manage these settings in the AI Governance pane.
AI Tools & Modules
AI tools expand an agent’s capabilities—fetching data or performing actions within a business environment. Tools are grouped into AI modules, which Thunk designers can enable or disable in Step Instructions to refine AI behavior. Designers can also create custom tools, integrate with enterprise systems, and share reusable modules across Thunks. Tutorial on building custom tools coming soon!
Content Collections
Each Thunk can define Content Collections to ground AI responses with relevant information—like indexing product docs for support questions. Collections can include documents, images, videos, and web pages, from small datasets to extensive libraries.
To create & use Content Collections, upload documents via the Content tab to let your AI index and use them in Step Instructions. Your AI can also pre-extract structured properties, like document authors, for more precise automation.
Compliance Policies
Each Thunk has a dedicated collection for organizational compliance requirements. For example, in a purchase order workflow, policies on payment terms can be stored here. With Compliance settings enabled in AI Governance > Compliance, your AI automatically enforces these policies, ensuring all actions remain compliant.
Congratulations! 🎉
You've successfully customized your Thunk and your AI to fit your specific needs.
Contact Us
Happy thunking! Please reach out to us at [email protected] if you need assistance in building your thunks.