Generative AI Agents and the use of Large Language Models are all the hype.  On the one hand, stakeholders are hearing the appeal of switching their standard user interfaces for conversational ones.  “Everything will be a conversation”. 

On the other hand, standard web interfaces, products and user experiences have stood the test of time, proven their value and effectiveness, making organizations reluctant to switch things up.  Should one switch the ROI one knows for the promise of AI efficiency they believe in?

This type of polarised technological view suddenly makes organizations pick between becoming obsolete or taking the risk of throwing proven conversion out of the door, to the benefit of the newest hype that ‘hopefully’ will yield equal or better results.

In reality, Conversational AI Agent interfaces – at OpenDialog we refer to them as ChatUI, and graphical user interfaces don’t have to be mutually exclusive.  Both of them have their advantages and pitfalls.

The standard webchat user interface typically contains a few interesting out of the box UI elements with clear benefits::

• A free text input field allowing users to ask their questions in natural language, on their terms
• Personalised options through carousels to present different options to the users, using a set of rich cards and associated buttons 
• Quick reply buttons to allow users to make quick decisions
• 2-way interactions: allowing users to interact directly with content and information, rather than just being on the receiving end

Conversational AI Agents, even in the well-known right-hand bottom corner webchat widget, provides users and organizations with a unique way of truly communicating : through conversation, in natural language.

However, they are also limited in many ways : 

• Difficult to discover as it sits in the corner of a page when minimised
• When starting of maximised, it hinders the scannability of the original page
• Loudness when popping up incessantly
• Limited UI real estate for more complex interactions

When users think of Conversational AI Agents, more often than not they associate them with the bubble in the right-hand corner of a website that pops up and that you can ask questions. Whether or not those questions always yielded the right answers was unclear – leading these interfaces to suffer from a tarnished reputation.

Graphical user interfaces are familiar to users.  For all their flaws, they have made up the daily bread of how users interact with information on the web and how they interact with digital products.  These user interfaces use well-known patterns and interaction paradigms that users have learnt to use in earnest.

User interfaces as we know them, to date, offer the following advantages:

• Higher discoverability through clear information architecture : navigation menus, scrollbars, search bars, and structured layout make it easy for users to locate information at a glance
• Better information density: Unlike webchat interfaces, which present content sequentially, graphical UIs can display large amounts of information simultaneously, improving scannability.
• More Control for Users – Users can jump between sections, filter content, and browse at their own pace without waiting for a Conversational AI Agent’s response.
• More flexible interaction patterns – Graphical UIs support complex workflows, such as filling out multi-step forms, comparing multiple products side by side, or using interactive dashboards.

However, these graphical user interfaces are very one-sided, allowing users very little leeway as it comes to the content and the way they interact with it.  One could almost compare them to a one-way street : you only get to interact on the graphical interface’s designer’s conditions.  This puts a lot of pressure on the graphical user interface: 

• to convey the information you want, 
• to contain the information users need,
• to make sure the user understands and acts upon that information as you would expect,
• to be optimised for conversion

At OpenDialog, we have always believed that Conversational AI applications need to be flexible, modular and personalised: the perfect combination of conversation design, generative and explicit reasoning.  In the OpenDialog framework for Conversational AI Agents this has meant giving our customers the power that comes with using Large Language Models for free-flow conversations, as well as the controls needed to harness process-led conversations.

Not content with the constraints represented by the interface side of things, we have therefore set out to break those boundaries as well to provide customers with more flexible ways to interact with their end-users on their terms : ChatUI SDK.

Custom Interface Layouts – Move beyond traditional chat widgets with customizable, UI-native conversational experiences that enhance your platform’s engagement.

Conversational Search – Turn AI-powered search into a dynamic, real-time guide that delivers instant, relevant responses while adapting the interface dynamically.

Inline Interactions – Users can highlight text or interact with UI elements to trigger inline conversations, turning static experiences into dynamic engagements.

Rich, Dynamic Content – Markdown and CSS support enable branded, immersive, and highly engaging messaging experiences.

On-Page Event Triggers – Prevent churn, enhance retention, and improve UX by triggering just-in-time conversations based on user behavior—from button clicks to navigation actions.

Full-Page Immersive Experiences – Think beyond chatbots! OpenDialog’s ChatUI enables fully integrated AI assistants that redefine what Conversational AI can do within your product.

Thanks to the combination of graphical user interface elements, user behaviour, and Conversational AI meeting users where they are, the OpenDialog ChatUI SDK will allow our customers to maintain the benefits they have reaped from their legacy interfaces, all while allowing them to harness Conversational AI on their user’s terms.