You probably use AI chatbots like ChatGPT and know that the underlying AI models are powerful and intelligent. At the same time, these chatbots are difficult to rely on to get work done. You have to sit in front of the AI chatbot to instruct it and provide it context, and then check, proof-read and correct its responses. Besides, an AI chatbot does nothing when you don't sit in front of it. It does not automate your work.
What can I do with Thunk.AI?
The simple answer is "reliable AI automation". In other words, AI logic runs automatically on your work items or your data without a human being having to sit in front of a chat window and tell it what to do.
AI logic that runs automatically is called an "AI agent". You can set AI agents up to be fully or partially automated. The main benefit comes from eliminating repetitive work and by ensuring that work is handled by AI through an established process or workflow.
If your work involves repetitive manual tasks that involve reading or creating or analyzing documents, messages, images, or web pages, it is likely that Thunk.AI can automate some or all of it
The particular processes or workflows that you choose to automate are usually custom to your team or business. They do not need to be defined perfectly. Instead, you can give your AI agent high-level natural language instructions and expect it to follow the instructions sensibly and reliably. Here are some articles with examples and more details:
Why should I use Thunk.AI?
The reason to adopt AI agents is of course the productivity gain that comes from automating work, taking advantage of the dramatic breakthroughs that have recently occurred in AI technology. Here is a high-level description about the promise of automated AI agents at work.
Once you decide to use AI agents to automate workflows, why use Thunk.AI as the specific platform of choice? There are three reasons:
Steering and control: For a business workflow, there is often already a process, there are certainly rules to follow and policies to comply with. The workflow definition needs to be a combination of intentional instructions that must be followed, and flexible instructions where AI can make subjective judgments. In technical terms, you need to be able to steer and control the Autonomy (decisions it makes) and the Agency (what actions it can take in each situation) of the AI agents in a very intentional and granular way. The Thunk.AI platform is designed to achieve this. There is an explicit planning phase separated from a runtime execution phase. AI instructions and constraints are provided at a very granular level with a number of checks and balances.
Reliability and enterprise-grade governance: Most AI agent platforms create compelling demos, but they fail to deliver reliable, predictable, and consistent behavior. As a result, most actual AI agent projects fail to graduate past the proof-of-concept phase into actual production deployments. This is because the underlying AI models are fundamentally probabilistic. The platform on top of the AI models needs to build in mechanisms to account for the errors and unpredictability, and ensure that the resulting automated work is reliable. More broadly, the key governance requirements of an AI agent platform in an enterprise environment are Compliance, Human-in-the-loop collaboration, fine-grained Automation controls, Reliability, and Modularity. These are called the "CHARM" requirements. Thunk.AI provides rich capabilities to address these requirements. Of course, in addition to standard enterprise security concerns (Thunk.AI maintains SOC2 compliance), many enterprises also want AI automation software to installed in their own private cloud tenants. Thunk.AI fully supports that deployment model when required.
Time to deployment: For a traditional business application, one common approach is to hire software engineers who then write code to implement the application. However, AI agent applications are different. It requires specialized software engineering talent to build automated AI agent applications directly against AI/LLM models. Even despite the high costs, long development cycles, and slow iteration time, the final results still lack the reliability and enterprise-grade governance needed in a business environment. With Thunk.AI on the other hand, the really complex part of AI agent automation is available out-of-the-box, and most AI automations can be built entirely "no-code" using natural language. This leads to rapid development and iteration, which is critical to tuning and improving reliability. In addition, there are built-in mechanisms like auditing and scoring that help you improve your workflows and role-based access controls to ensure that team members can participate appropriately, and integrations with the other business applications that carry your existing data and context.
How do I use Thunk.AI?
As for the "How?" question, that's answered by the information in the rest of the articles. We recommend initially reading the article on Platform concepts and then moving on to the next level of detail.
You don't need to know or internalize all of this information before you start. Remember that Thunk.AI is a self-service platform and it is easy to sign in, look at a couple of samples, and then experiment by creating your own thunks.