AI has become the focal point of digital products. There are few tools on the market that don’t boast new AI capabilities, or have them on their roadmaps. But the common implementation of this game-changing tech is often confined to standalone or siloed experiences. It is being delivered as product-adjacent, rather than elevating a product’s primary purpose or user experience. We quickly need to evolve our thinking beyond the chatbot.

The chatbot interface is not a new idea, but it has been supercharged. It’s now the go-to tool for any business desperate to shout, “we have AI, too!”

But the chatbot pattern comes with baggage. Users are left to essentially build their own micro-product with each and every engagement. They need to train both it - and themselves - on how to work for them, a meandering and time-consuming process that’s difficult to repeat. We believe that new interaction models are needed to put control back in the hands of the user.

New paradigm, new problems

Jakob Nielsen recently referred to AI as the “first new UI paradigm in 60 years”. But should this mean we abandon everything we’ve learned about creating intuitive user experiences to date? Although in its infancy, we’re seeing patterns emerge in AI-based interfaces that we’ve collectively spent decades improving and reimagining in traditional digital products.

Problem 1
Hyper-linear workflows

For those who remember terrestrial television, refining output in a conversational format feels similar to shouting to your Dad, who’s up on the roof, fiddling with the arial, furiously attempting to interpret your instructions until you shout, “THAT’S IT!“. Chatbots hold the knowledge, and the user holds a vision of what they want to achieve. Getting to a desired outcome becomes a tedious back-and-forth of long-form instructions.

Without full visibility of the levers that shape responses, it becomes much harder for people to get what they want. We’ve reverted back to command-prompt interaction - but with prose instead of code.

Problem 2
Siloed, one-on-one interactions

With the rise of apps such as Notion, Slack and Figma, the way we work has evolved into much more streamlined process—where users collaborate in real-time and content is shared seamlessly.

But users address chatbots in isolated, one-on-one engagements, without the full context of what they, their team or the business are doing or trying to achieve. The wider context of their project or goal, and the steps they took to get to the outcome, are forgotten.

Problem 3
One-dimensional output

Chatbots will answer any question—but typically only that question, before distilling responses down to a short piece of text. Patterns and insights that could have been derived organically risk being left undiscovered.

This is especially relevant in the world of data and analytics, where understanding patterns and anomalies is a key activity for many users. We should remember the concept of organic discovery, and reduce the need for users to imagine every possible scenario or questioning line.

Problem 4
Unfocused and vague user journeys

The very nature of a conversation with a chatbot is open-ended, so when we simply apply them as a layer on top of existing interfaces, they can lack a sense of purpose, structure and progression.

When imagining new products and features that integrate AI, we should think about how we can empower users with the ability to create more focussed and task-specific tools, best suited for the task at hand.