When should you use AI in your product?

It’s easy to want to build AI-powered features in your products, but the complexity and upskilling costs might not be worth it. AI solutions, in my opinion, are appropriate when you need to support an explosion of inputs, workflows, or outputs.

Explosion of inputs

If your UI is or will be an overwhelming set of features/dials/knobs, where customers have to figure out which of those to use (if you have the ones they need) and how to use them, that user experience cost could be eaten by AI. In that, you could now allow them to express their needs succinctly in natural language, without learning your UI language. It’s also an explosion of input, in that, you now have so many variations to support in terms of language and expressed needs.

Explosion of workflows

If your product wants to offer a large number of workflows, consisting of many different combinations of possible operations, you might be better off using AI. Manually supporting the number of combinations could be expensive to build and that cost continues as you want to add new operations. With AI, specifically agents, you let the agent determine the set of series of actions to take based on the user’s needs. You provide the tools/operations, and the agent figures out the workflow. Adding a new operation becomes adding a new tool (it’s plug and play) without worrying about how it fits into a web of combinations.

Explosion of outputs

If you want your messaging to be personalized to the customer’s situation, then AI can help. Otherwise, you’d need to group your customers and have generic messaging for each group. It can be difficult to find and maintain the right customer groupings and messaging. Instead, an AI can take in data about the customer’s situation, some guidance on style/constraints for its response, and respond uniquely to that customer.

Choose wisely

Previously, I wrote about the differences between chatbots, workflows, and agents. Chatbots and workflows are good at handling the explosion of inputs and outputs. Agents, can handle all three: inputs, workflows, and outputs.

Your product may not need AI. Choose wisely.