It's the End of Forms as we know it, and I feel fine
Forms act as tiny silos of data that inherently create averages of averages when aggregated.
Forms, in the traditional sense, are a way for Humans to collect and structure data. They are generally not an ideal way for humans to interpret said data. That's what Analytics is for.
Forms also have no real way of integrating other contextual information that's relevant to that analysis. They act as tiny silos of data that inherently create averages of averages when aggregated.
A Large Language Model, however, doesn't need a form in order to capture and structure information. It just needs whatever unstructured data you have and the example output you want. That output can be in formats that are easy for humans or machines to interpret and take action on. And now, they can output multiple formats at the same time based on each unique use case.
A real risk with Generative AI is generating data for the sake of generating data. So, if companies are using AI to populate forms, such as Quality Assurance in the Contact Center, then you are merely automating and scaling an already suboptimal process. Queue the status quo.
Now, I can't throw too much shade on my AI brethren for going this route. I invested a lot of time and resources into the idea of letting people create a form and then build prompts for GPT to auto-complete them.
Seems cool in concept, until you ask the question “now what do we do with thousands of auto-completed QA forms?” Ouch.
As example, almost every QA form ever created has a [Y/N] box for if the agent gave the proper greeting. But in reality, I don't care much if an agent didn’t give the proper greeting in 1 out of 100 calls. But I do care if they aren't adhering to our HIPAA guidelines, or if they used profanity, or if there is an extended amount of silence while they search for an answer to a caller’s question. And I want to know that in as near real-time as possible.
Similar, it's not all that helpful to look at a report and see how many calls an agent transferred to another department. What is insightful though is identifying under what conditions that happens and if it represents a coaching opportunity for that agent or if it’s a wider trend that needs a closer look.
That type of precision analysis and coaching can be done continuously using AI without needing to populate or surf through hundreds or thousands of forms.
So we’re building that instead.