The rise of conversational user interfaces and the era of micro productivity
Business applications are changing due to generative AI. You’re about to see a change in how you work, and it’ll be fuelled by the conversational interface.
Many businesses have used artificial intelligence to implement conversational user interfaces (CUIs) over the years. The common chatbot is probably the most referenced example. When you think of conversational user interfaces, customer interactions likely come to mind, or perhaps internal use cases like HR and IT support. But what about the business systems you use at work every day? Your CRM, BI tools, distribution systems, stock management, finance, HR, CMS, email, word documents, PowerPoint, document management: all of these are all about to be overhauled with the conversational UI. And this will usher in a new wave of what I call ‘micro productivity’.
The concept of micro productivity
Micro productivity is the small but significant efficiency gains that accumulate throughout the working day. Consider a common task in a graphical UI. Searching and updating a contact in a CRM system, for example. Locating and updating, then publishing a web page. Checking on a customer’s claim status and understanding what the next steps are, so that you can advise the caller accordingly. These GUI-based activities might take, say, 10 steps to complete, and perhaps takes around 30-60 seconds.
Compare that with just asking for what you want. A conversational user interface (CUI), by contrast, could reduce these journeys to as little as a single step, and reduce the time taken by a non-trivial factor over time.
The value of aggregated marginal gains
On its own, this might seem like a minor improvement: a mere 30-second saving here and there. But when multiplied across an entire workday, these micro savings compound into some very real productivity gains, such as:
Time saving: Performing the same task dozens of times daily could mean hours saved each week.
Ease of use: Systems become significantly easier to interact with, reducing frustration, shrinking learning curves and improving experience.
Reduced cognitive load: Employees can focus on higher-value work, rather than navigating complex menus and searching for information.
This concept of micro productivity mirrors the idea of ‘the aggregation of marginal gains’, which was popularised by Sir Dave Brailsford, the performance director of British Cycling.
Brailsford and his team sought out every possible efficiency, no matter how small, from shaving a few grams off the weight of the bikes to putting alcohol on their tires for better traction. Each improvement may have seemed insignificant on its own, but together, they contributed to significant performance gains, ultimately leading to victories in multiple Olympics and the Tour de France.
In the same way, conversational interfaces introduce small but impactful efficiency gains across business processes, which when compounded, could lead to major increases in productivity and efficiency.
15 AI Trends in Customer Experience 2025
Learn how to find success with AI.
VUX World presents 15 key insights we’ve learned from working and speaking with enterprises deploying AI in CX during 2024. Download our report to discover key challenges, top use cases and 15 solid recommendations for how to make AI work for you in 2025.
The limitations of Graphical User Interfaces (GUIs)
As great as the graphical UI is, it has its limitations. All of these systems exist to do specific jobs, and every time you open a program, you have a specific outcome in mind. GUIs aren’t always ideal for achieving outcomes for the following reasons:
1. Usability
Users have to learn how the system works; what to click, what to search for, and which fields to complete.
2. Information architecture
Poor organisation of content and features with unclear labelling leads to a frustrating user experience.
3. Screen real estate limitations
The hierarchical design of most GUIs, constrained by screen real estate, often buries essential features within multiple layers.
4. Sequential progress
Completing a simple task often requires multiple clicks, searches, and navigation through various menus step-by-step. There are very few shortcuts.
Generally speaking, then, although we are historically comfortable with GUIs, and some GUIs work pretty well, they don’t work perfectly for every use case.
Jakob Nielsen, the well respected pioneer of User Experience Design, describes the graphical UI as a command-driven interface, wherein the user tells the system what to do, one step at a time.
The value of Conversational Interfaces
Conversational UIs address the shortcomings of GUIs with several key benefits:
1. Usability
Users interact with the system simply by typing or speaking; skills most people already possess. Instead of users learning to understand the system, the system learns to understand the user. If you can use natural language processing effectively, then when done right, conversational interfaces are generally more user friendly.
2. Information architecture
Conversational interfaces consolidate features and data into a single entry point, using artificial intelligence to do the heavy lifting of language understanding. This eliminates the need for extensive navigation.
3. Screen real estate
Because of natural language processing, you can simply ask the system for what you want using human language, so you’re not confined to what’s clickable on-screen.
4. Journey progress
A well designed conversational UI enables you to get to the root of your need without having to go step-by-step.
Jakob Nielsen describes this shift as “reversing the locus of control,” as conversational UIs are intent-based and outcome-driven. Users express their desired outcome, and the system handles the complexity.
Comparison between GUIs and CUIs
For simple comparison, here’s an overview of the differences between graphical and conversational interfaces.
Previous attempts at conversational interfaces
The idea of conversational UIs isn’t new. Companies have been experimenting with them for years. Previous examples include:
Capital One, which had an early advantage in conversational interfaces, with the creation and launch of its virtual assistant, Eno, in 2017. This was probably the first example of a multi- or omni-channel virtual assistant that you could access via Amazon Alexa, the capital one mobile app, and even via a shortcut in the Google Chrome browser.
Bank of America and its virtual assistant, Erica, launched in June 2018. The virtual assistant, accessed via the Bank of America mobile app, intended to collapse the entire mobile app into a single conversational UI. At a time, this was well regarded as best in class for mobile banking.
Spotify launched a ‘Hey Spotify’ conversational interface in its app in 2021 to make searching for music easier. In the same year, it debuted its own hardware, Car Thing, that was a device built for your car that provided a conversational interface as an entry point into your Spotify account.
The automotive industry jumped on the first proper wave of conversational interfaces, too, with companies like SoundHound providing voice assistants to companies like Hyundai, KIA, Honda and Mercedes. Then you had BMW that built its own in-house voice assistant. And others, like Volvo, Audi, Ford and many more having the same capabilities provided by Amazon Alexa and Google Assistant.
Voice assistants. Then, of course, you have the industry leaders and, at the time, best in class conversational UI, the voice assistants; Amazon Alexa, Google Assistant and Apple Siri. You could even throw Samsung Bixby into the mix, however this wasn’t as widely adopted as initially hoped.
The conversational UI goes a lot further back than this, but these were the first wave of widely used and popularised examples in recent times. So popular and so exciting, that many of us in the field in 2018 truly believed that the conversational interface will be the key enabler of a whole new paradigm for human computer interaction. I wrote in the Harvard Business Review in 2019, about how voice assistants will have an impact on our lives and change some fundamental behaviours, such as how we shop.
Despite these early efforts, widespread adoption and sustained usage faced significant challenges due to the limitations of available technology. We couldn’t quite make conversational user interfaces work consistently enough for everyone.
Limitations of previous conversational user interfaces
Most of the applications referenced above use traditional machine learning natural language understanding (NLU) to classify a customer utterance and match it to an intent (with the exception of SoundHound). Once this intent has been classified, it uses deterministic rule-based logic to decide what to do next.
While this was best-in-class, and the standard until 2021/22, it has limitations.
1. Limited Understanding
NLU-based systems rely on natural language understanding (NLU) models. Typical accuracy rates of these machine learning systems is around 80-90%. This means that 10-20% of interactions with conversational interfaces resulted in the dreaded “Sorry, I didn’t understand that” response. It also meant that longer utterances or multiple intents were impossible to deal with for most (again, with the exception of SoundHound).
2. Predefined Logic
Natural language understanding systems operated on pre-programmed “if-this-then-that” logic, making them inflexible and unable to handle unexpected requests, which is all too common when you’re dealing with the complexities of human language.
3. Pre-written Dialogue
All responses sent back to users were scripted during development, leading to rigid, unnatural interactions and an inability to handle novel requests. These limitations meant that the overall experience of conversational interfaces felt clunky.
To make conversational interfaces work as well as they need to for them to be useful, a technology upgrade was needed, and that was brought about by generative AI.
How generative AI is changing the game for conversational interfaces
With generative AI hitting mainstream in 2022, with the launch of ChatGPT, much experimentation has gone into how to leverage large language models to create more natural and more useful conversational interfaces. Generative AI gives us an opportunity to address the limitations of previous technologies and offers some additional benefits, such as:
Better understanding
With advanced retrieval-augmented generation (RAG) techniques and prompt engineering, generative artificial intelligence can comprehend user inputs more effectively.
Dynamic task execution
Although in its relatively early days, AI agents have the potential to autonomously plan and complete tasks without requiring pre-defined workflows, enabling more adaptive and personalised experiences.
More natural interactions
Even with pre-defined workflow logic, generative AI can create context-aware, personalised responses, rather than static, pre-written dialogue, which makes interactions feel more natural.
Tool usage
Having large language models use function calling, and giving them access to tools, means that CUIs can now affect the real world by making changes and updates to applications and systems easier than they could before.
Dynamic user interfaces
With access to tools, generative AI models can now dynamically change the nature of the interface, based on the user need. I call this the Dynamic User Interface.
As such, we’re seeing many SaaS companies now implementing conversational interfaces. A trend that I see will only grow over the course of 2025 and beyond.
Real-World examples of contemporary conversational UIs
Over the last few years, there have already been some examples of more contemporary conversational interfaces, built on generative AI, such as:
Google Analytics. These days, you can query data directly using natural language instead of navigating complex dashboards.
Enterprise AI Platforms. Platforms like Cognigy and Kore.ai allow businesses to describe their automation use cases in natural language, and the system generates a first draft of an AI application.
Zapier. Need to build a specific workflow in Zapier? It’s conversational interface, which it calls an AI agent, will let you describe what you need and it’ll build the first draft of the workflow for you.
Hear.ai. The inspiration for this whole post came from Hear.ai. I recently interviewed co-founder and CEO, Noam Fine on the VUX World podcast. Hear.ai centralises customer interactions across channels and allows contact centre managers and agents to extract insights simply by typing.
ChatGPT. Probably the most popular example of an application that makes conversation the primary modality is ChatGPT itself.
These examples highlight how CUIs can reduce friction, improve adoption, and drive efficiency across applications and industries. And not just by providing a chat bot interface, but by moulding with the program, affecting what it does and shaping the creation of outcomes.
It also shows how CUIs are making their way into your business, and they’re just getting started.
The proliferation of conversational UIs
You probably already use conversational UIs every day without realising it, whether in Slack, Teams, or even email. These platforms, along with the popular voice assistants mentioned earlier, have normalised natural language interactions. The problem has been that they’ve all lacked the ability to create productivity gains for the enterprise.
As you can see, conversational user interfaces, powered by generative AI, are now starting to appear in the business application layer, from analytics to workflow creation. These conversational interfaces will begin to proliferate across the enterprise in the coming years. Everything from your CRM to your HR system will have a conversational interface, and I predict that this will see an explosion of micro productivity across business functions. Micro productivity that amounts to big changes in the effort distribution of resources and even the removal (or significant reduction) in the amount and type of tasks performed in a given role.
The question is; what will you do with your new found time? What opportunities will this open up? And how will you use that time to scale, improve and transform what you do?