There are four broad process patterns where an AI-powered thunk adds value in your work environment. In all these cases, the value of AI comes from automatically and efficiently doing what human beings would otherwise have to do slowly, laboriously, and with significant expense.
Person-to-person processes (often conversations conducted through messaging)
Person-to-system processes (often conducted through applications and forms)
System-to-person processes (often notifications conducted through messaging)
System-to-system processes (integrations conducted through backend software)
The Thunk.AI platform enables all four of these process patterns. This article describes these patterns in the abstract. Subsequent articles list specific instances of these patterns for different organizational functions: Sales, Marketing, Customer Success, etc.
External systems can send messages to a thunk (via email or HTTPS webhooks) to kick off the automation.
The AI agents in the thunk can fetch data from or send messages to external systems (via custom AI tools) as part of their AI processing. Both the incoming and outgoing integrations are easy to create and configure.
Here is a basic thunk sample that takes input from a form submission (via a webhook) and that uses a custom AI tool to send a chat message as the result of the workflow.
1: Person-to-person processes
Many processes involve a person A sending messages (information, a request, a notification, etc) to another person B, especially within the same team or organization.
An intelligent AI-powered thunk can add value to the recipient by understanding and processing the message. The result might be a summary or a proposed response or an action taken on account of the message.
Sometimes a process is required to route a message from one person to an appropriate other person. For example, a business may receive a message from a customer, and it may need to be routed to the appropriate customer success representative. An AI-powered thunk can act as an intelligent router, understanding the content of the message and other contextual factors in making an intelligent routing decision.
2: Person-to-system processes
Much of the work in an organization starts with a person and ends up in an application or system.
In the simplest case, a user sends a message to kick off the process. For example, a customer may send a message to the email address of a business. An AI-powered thunk can read the message and convert it into an appropriate structured payload to save into an internal system. In such a pattern, a thunk acts as an intelligent message processor.
A very common case is to use an application (like a Forms package) that presents a user interface and accepts inputs from a user. These inputs get sent to an API. Every submission form is an example of such a system. An AI-powered thunk can act as an intelligent workflow engine to process the submission, validate it (were the appropriate fields correctly filled in), annotate it (assign a priority, for example), enhance or summarize it, and then save it into an appropriate backend system.
Many existing applications already have a workflow engine at the mid-tier and a database to store data. In these cases, an AI-powered thunk can act as an intelligent step added to the existing workflow. A thunk can be used to provide recommendations for submitted information (eg: does the candidate meet the job criteria) or to validate the correctness of submitted data (eg: does the submitted photograph have sufficiently good lighting).
3: System-to-person processes
Many systems and applications initiate processes that communicate with users. A delivery service informs customers that a parcel is on its way. Many email marketing applications send messages to users with information or content. An AI-powered thunk can be used to construct and transform the content per user. For example, something as simple as translating the content depending on the specific user or choosing appropriate content on a per-user basis.
4: System-to-system processes
Finally, it is very common in business environments to have system-to-system processes. The most common reason is to do "integration" across systems. Companies may have a separate payroll system and a separate HR system, so when a new employee gets added to one system, the change needs to cause a corresponding entry to be made in the other system. However, a number of details may differ -- things as simple as property names and formats, for example. Sometimes, integrations involve rich conditional logic or judgment calls (some fraction of expense reimbursement requests may get pulled into a special review system for further scrutiny). An AI-powered thunk can act as an intelligent system integration pipeline.
