UX Collective: Are we doing UX for AI the right way?

Katya Korovkina, a Design Manager and CX expert at Eleks with fourteen years of experience in digital product design, published this piece in January 2026 as a direct challenge to one of the most common assumptions shaping AI product work: that conversational interfaces are the natural default for AI features. She calls this assumption “chatbot-first thinking” and argues it is damaging the quality of AI UX.

The article is grounded in a straightforward observation: language models can respond to natural language input, so teams default to building chat-based interfaces. But the capability to do something does not mean it is the right interaction model. Korovkina’s central point is that conversational UI is appropriate for a narrow set of tasks and genuinely harmful for a much broader set.

She identifies several structural problems with over-relying on conversational interfaces. The first is what she calls the articulation barrier. Text-based prompting requires users to formulate requests in language, and a significant share of the population — including in high-income countries — struggles with that kind of complex written expression. Interfaces built around chat effectively exclude users who would have had no trouble with a well-designed visual UI. The second problem is that recognition outperforms recall in most discovery scenarios. When a user scans a list of options, they can identify what they need without being able to name it in advance. Conversational interfaces, by contrast, require users to know what they want before they ask. This makes them poorly suited to exploratory tasks.

Precision is a third issue. High-stakes interactions — signing a contract, adjusting a financial record, confirming a medical detail — benefit from explicit, field-level confirmation. Dialogue creates ambiguity, and in contexts where the cost of a misunderstanding is high, that ambiguity becomes a liability rather than a feature.

Korovkina’s recommended approach is to treat AI capability as one tool in a service design process rather than the interaction layer itself. She points to Microsoft 365 Copilot’s adoption rate across Fortune 500 companies as evidence that AI creates the most value when it enhances an existing workflow rather than replaces the familiar interface around it. Teams that have done this mapping — identifying the steps in a user journey where AI genuinely reduces friction versus the steps where a traditional UI does the job better — consistently report higher satisfaction than teams that rebuilt around chat from the start.

The article is most relevant to designers and PMs who are being asked to add AI to an existing product, and to teams building AI-native tools that risk conflating conversational capability with good UX. It does not cover specific tooling and assumes the reader is already building in an AI context.