One day, we’ll say goodbye to the GUI

AI is  changing the relationship between humans and digital technology. Join us as we establish our vision for the future of human computer interaction (HCI)

“The function of good software is to make the complex appear to be simple.” — Grady Booch

Good software shields the end-user from underlying complexity, a process called abstraction. One button click can send a message to contacts across the globe, with the dizzying complexity of the global internet behind that single click … data centres, servers, databases, APIs, application software, fibre cables, operating software, virtualisation technology, content delivery systems, domain name servers, protocols, routers, proxies and everything in between, are abstracted away from the end-user.

The human-computer interface, where we ‘converse’ with a computer, has evolved since the days of punch cards. Today’s graphical user interface (GUI) gives interface designers control of individual pixels on a screen, a huge leap forward from text interfaces and anything else before that.

And it’s fair to say GUIs have been hugely successful. We use them on our phones and desktops computers, for gaming, in our cars, on home appliances. They have become critical to daily life.

User experience, how users connect with software functionally and emotionally, is critical to get right for organisations serving a large customer base. It’s been shown that a user experience is highly correlated to customer satisfaction, retention, referrals and costs. Get it wrong and customers will emigrate to the competition.

Let’s consider the technology underpinning Uber, abstracted away to a simple app which lets you book a short-term chauffeur-driven vehicle has been a global success … Uber is 15bn rides in.

With an easy to use GUI, Uber cracked the historical issue with cab hires, making sure driver and passenger know exactly where each other are, leading up to the journey. All fear and doubt was removed. And now, services without an Uber experience feel low rent by comparison.

Cognitive Load

What about the enterprise software we use at work? Organisations invest in enterprise software for strategic benefits … improvements to market share, customer retention, operational efficiency, competitiveness and costs are frequent justifications.

Systems borne from such motivation only have the internal user’s interests at heart in so far as they will become more efficient and effective in their work. Rarely are end-user’s personal and emotional interests taken into account. Learning one comprehensive business system is hard enough, but it’s not uncommon to see end-users in the back office navigating their way around multiple software applications to cover all their duties.

Scale this up beyond just one application … in 2023 the average knowledge worker needed 11 software applications to do their job (Gartner) … and we get a sense of the cognitive load expected from end-users working with GUIs.

Imagine being trained on that number of business systems. No wonder churn is high in these environments.

AI to the rescue

AI has demonstrated this need no longer be the case. Swapping a convoluted point & click experience with natural language is the big change coming to software. No longer will we stress while locating find the feature and data we want. Instead, we’ll make requests in our own natural language … “assess claim ref 12345”, “send apology email to customer Sue Smith with a 30% off voucher”, “I want to report a car accident”, are valid instructions we can now type or speak into computers.

What’s Caseblocks response to AI?

Caseblocks is for organisations looking to optimise business processes, today, rendered through a typical browser GUI. Every Caseblocks solution includes automation that happens with and without direct user intervention, such as when a case changes state, or when the end-user selects a case task to run.

As a model-based platform Caseblocks has a significant advantage over application-based solutions, which spread business logic in different locations, in different software languages, knitted together in a single stack. Caseblocks’ process models are contained in a text-based repository of case templates, schemas, state models, scripts, service levels, workflows and more. This content can easily be included in prompts, improving the LLMs chances of providing an accurate response and in the context of the end-user, not just the context of the trained LLM.

For example, if the prompt includes a definition for the loss ratio on an insurance policy, such as:

loss ratio  = (claims received / by premiums collected)  x 100

then the user can include the words “loss ratio” in any prompt and expect the LLM to use it appropriately, for example:

user > what's the average loss ratio on all policies written in 2024?

user > are there policies which are outliers on loss ratio?

user > are there policies which risk their loss ratio going over 80%?

These prompts and their responses feel like a destination in the evolution of HCI. The abstraction is human-level, hiding the complexity of the GUI interfaces we work with today.

Caseblocks’ current focus is to replace every point & click interaction in the GUI with a natural language alternative. And within that milestone, we have started on reporting, an area which will benefit greatly from the power of LLMs.