Thunk.AI is a self-service platform to automate work with AI agents. You need to know four things about Thunk.AI before you dive in:
It is a self-service platform. You can sign in and start using it. You don't need a software engineer (there is no code to write) and you don't need to talk to a sales person.
It uses AI agents based on the newest GPT AI technology. Each user in the platform is paired with an AI agent that helps them get work done.
It is a project platform. Each project is called a "thunk" and it has a high-level goal. Your AI agent works with you within each thunk to automatically process information and act upon it, informed by the goal.
It is a collaboration platform. While a single person can use it standalone, it is common to have a team of people and their AI agents collaborating together in each thunk.
So with that background, here's how you get started. You will have built your first thunk in 5 minutes.
1: Create an account
Start here and go through the sign in process using your Google account. After you sign in, you go through a few introductory screens and then get to a "My Thunks" page.
There are a few questions that come up at this stage:
Is it Google signin only? Yes, at the moment, though we're going to add other authentication providers. We generally suggest you sign in with the email address you use at work (if that is also a Google account). If not, most people have a personal Gmail account and that is fine.
Why does it need permission to access Google Drive? During sign-in, we request you provide permission to be able to (a) write files to Google Drive, (b) use a Google file picker to give you the ability to add files to your thunks. This limited permission is essential for the platform to work. In fact, once comfortable with the platform, many users decide to provide fuller read/write access to the Drive platform.
Why does it ask me to for information "about yourself"?: This is not meant to gather information about you for nefarious purposes (no, this data is not going to be sold and you won't be shown ads!). These few questions are really important in guiding your AI agent towards doing things that are useful. You should think about your AI agent like a work assistant. Your work assistant would almost certainly need to know who you are, where you are located, what your team does, what your role in the team is, and what your organization or company does. So we encourage you to provide meaningful answers here.
2: Look at a few samples
Once signed in, you will be on the "My Thunks" page. You will see a collection of sample thunks covering different use cases. It is possible one of these samples is very closely related to what you want to do. More often, new users don't quite know yet which problem they really need to solve. They need to first experiment and understand what the platform can do. Browsing the samples is a great way to get a sense of the diversity of use cases.
You can read more about the samples here.
3: Read a few articles
Usually, at this stage, new users have some idea of what the platform is doing. But there is so much hype and confusion about AI in our daily lives that it is also likely there is come confusion. It is a good time to read a few articles in the Help Center if you are so inclined. We suggest starting at the Core Concepts.
4: Copy a sample
The samples are read-only (i.e. you can see them but you cannot change them or interact with them like a user). However, you can copy a sample and when you do that, you'll get your own copy in your account. The copied sample will have the same goal and plan and data structure as the original, but the data isn't copied over. This allows you to start using the sample from the beginning, populating it with data, getting familiar with the UI, and interacting with the AI agents.
We are also starting to pre-create some copied samples in each new user's account, so you may be able to directly engage with these copied samples.
5: Create a new thunk related to your work
Now that you have some familiarity with things, create a new thunk.
Type in a goal (this is the intent for the new work project). How complex or detailed should this be? You'll see a few examples in the UI to help guide you. Usually a short goal (about 5 words or less) is a sensible choice. You will have the opportunity later to provide more details.
Then you get to choose between a few different project patterns to create your thunk.
The Task List pattern is suitable for projects where individual tasks or workitems are not repeated. There is almost no upfront planning that happens here, so this is the quickest and easiest kind of thunk to start with. But conversely, this isn't suitable for processes that have a lot of repeated work that needs reliable automated AI logic.
The Workflow pattern is suitable for projects or processes where the same sequence of tasks (the "Plan") needs to be applied automatically and repeatedly to different work items. If you choose this pattern, you can go through an AI-assisted planning process to choose and refine the plan to be followed.
The File Folder and Spreadsheet patterns are special-cases of the Workflow pattern that choose a simple plan automatically without your engagement. These are particularly useful when you are getting started and want to rapidly explore the platform.
Whichever of these paths you take, you'll be done with the creation phase in a minute or two and then your thunk moves into the execution phase. Your AI agent starts the first task. Usually, this is a task that involves bringing in data entries. There are different ways to add entries:
In many of the samples, the easiest way to add entries is to use the UI (add an entry or upload a collection of entries)
Or you can tell your AI agent to add entries (ask it to add something specific, or ask it to search the web and find entries, or ask it to fake some entries -- it's your AI agent and does what you ask it to do!).
Or you can forward email messages to your thunk and they get automatically converted into new entries.
A couple of questions come up at this stage:
I've added entries, why isn't it doing anything? Most likely because the entries are in "Draft" mode. By default, new entries get added in "Draft" mode just to make sure you don't inadvertently add a lot of data and do a lot of AI work by mistake. Click on the "Draft" button to change its status and then your AI agent will start executing the planned tasks on it. There is an AI agent setting that you can change to avoid using the Draft mode. Usually, users enable this setting after they have checked that the thunk does exactly what they want.
How do I get it to automatically move from one task to another? You'll find that your AI agent finishes the work of one task but then it doesn't move on until you click on "Finish this task". This is also by design for the initial stages of using a thunk. It is also controlled by an "auto-complete" setting for the thunk. If you enable "auto-complete", your AI agent will automatically decide when a task is complete and move onto the next task in the plan.
Congratulations! You are now ready to AI-enable many aspects of your business with thunks.