Just because the hallmark of generative AI is natural language, that doesn't mean free form inputs are the best way for enterprises to integrate genAI.
With SaaS application integrating Generative AI into their platforms, there's a tendency to try to expose infinite possibilities. While this sounds good in theory, it turns out that humans don't do well with too many choices.
In his book, The Paradox of Choice, psychologist Barry Schwartz argues that eliminating consumer choices can greatly reduce anxiety for shoppers. This is why In-N-Out with their simple "burgers and fries" menu does so well. If you want Chinese food or tacos, you head somewhere else, but when you are at an In-N-Out, chances are you already know what you are going to order.
When one looks for poorly implemented AI in SaaS interfaces, you can see it front end center in the new version of Google Analytics (GA4) which was released in late 2020. They were an early adopter of AI, but some of the key problems are with the user interface, even though it was designed to be more "user-centric" and customizable. Be retreating from having any opinions about how a user should behave, the user is left to figure it out for themself. As a result, they will often get frustrated and/or do nothing at all.
Voice powered platforms face a similar conundrum to the paradox of choice, but the heart of the issue is that without a visual queue, people forget what they are able to do. You can see this in the Amazon ecosystem where once installed Alexa "Skills" are often never used more than once or twice. Things have gotten so bad in the Alex ecosystem, that starting in June 2024 developers are going to have to pay out-of-pocket to host their applications.
The possibilities for improving user experience and unlocking business value with generative AI within SaaS applications is tremendous. However, it's important that product leaders integrating GenAI follow these framework:
The value of a SaaS application, vs. custom software, is that the creators have opinions about the industry. If you have a SaaS platform servicing the timber industry, you should have personas (mill workers, mill managers, etc) that tell you what people are doing on a day-to-day basis. Your SaaS application should serve these customers, and should have opinions about the right way to do things.
Never expect a user to know what's possible within your SaaS app. Always provide visual queues. If Gen AI unlocks a piece of functionality, have it available as a button or a dropdown, not a free form text that they need to guess the write words for.
Rather than waiting around to see how open ended prompts are used by hapless users, use generative AI to enhance information you already have. Have a bunch of customer call data? Ask a Gen AI system to grade the calls from best to worst or check to see if appointments are being made.
Any time you are asking users for a bunch of data, like writing a bio on a dating site, you can use Generative AI to create customized default text. This should then be easy to overwrite or modify, but if you can get people over the hump when it comes to free form text inputs, your app will be one step ahead.
--
We are just scratching the surface. What have you seen that works? Want to talk through idea on your latest project? Message me.
Our VP of Engineering outlines the questions a properly designed proof of concept (PoC) needs to answer, particularly for complex technologies like...
Steps to use Amazon Bedrock to get process Documents.
Economies of scale around compute and data favors tech giants in the Generative Ai space. But firms of all sizes benefit from the giants'...