Could the public sector be a hotbed for AI innovation?
After almost a decade of digital transformation, there’s plenty to argue ‘for’. And with huge data sets ready to be put to good use and a history of being able to adapt to technological change, there’s a lot going in the sector’s favour.
And yet, we still aren’t quite seeing it embrace AI. Why?
Quite simply it is the inherent risks that come with such innovation. Ethical issues aside, investing (public) money in these technologies requires a certain leap of faith – a luxury the public sector feels it doesn’t have.
But this shouldn’t hold the sector back when it comes to investing in AI and driving innovation. It’s in a position to be considered a leader in the field if only it were to look at the successes that have come before and identify its strengths.
Lean into product thinking
Product thinking is one of these aforementioned strengths. The core concept behind product thinking is to make sure that change is purposeful and user-centric, which doesn’t just happen accidentally. Every team has had an idea that, at first glance, is revolutionary - a shiny new widget or a fancy layout to make a user’s journey easier. More often than not, these things don’t have the impact they need to, or worse, actively frustrate users.
These ideas don’t pan out because they’re siloed off from the rest of the user’s experience and don’t properly consider the surrounding elements that need to be cohesively connected. Thinking of the overall “product” rather than an individual feature helps us to properly understand a user’s problem and design a solution that helps, rather than tack on another feature to a fundamentally flawed system.
The best way to envision this is to think of a user’s journey like they’re climbing Mount Everest. The problem they face is that it’s a difficult journey from base to summit. If you came at the problem from a feature perspective, you might be tempted to install signposts every 100m saying “go up,” with the justification being that they guide the journey. Product thinking, however, would be installing a chair lift from base to summit, actually making that journey easier for the user.
AI innovation and the “feature trap”
Unfortunately, generative AI, the type of AI that’s currently in vogue, is particularly prone to falling into the “feature trap.” There’s so much excitement around this technology that it’s being shoe-horned into as many places as possible with little to no consideration as to whether or not it’s useful. Generative AI does genuinely have the potential to significantly improve a myriad of processes, products, and lives, but it needs to be a tool, not the whole arsenal.
This is to say that generative AI's capabilities and drawbacks must be better understood. As for what this technology does well, it’s incredible at utilising training data to produce text, images, and audio. It can also interpret extensive data sets incredibly quickly. That leaves it with two apparent flaws: it cannot create anything new as it’s entirely reliant on training data, and it’s very much at the mercy of the quality of said data. The old adage ‘rubbish in, rubbish out’ has never been more true.
Mitigating the issue of ‘rubbish in, rubbish out’ is all about having strong data fundamentals. Three principles are crucial to getting this right: data must be discoverable, accessible and trusted.
The reality is that not every team has the resource or capability to build and train a generative AI model, so off-the-shelf versions will be used en masse. While powerful, off-the-shelf models are more generic than specialist by their very nature. The more data that you have quality assured, packaged and shelved for onward use, the better results you will see from generative AI.
Public sector applications for generative AI
Research is already underway about the efficacy of generative AI in a triage capacity and has shown promising results. In a recent study evaluating the triage performance of AI chatbots for ophthalmic conditions, ChatGPT listed the appropriate diagnosis among the top three suggestions 93% of the time versus 95% for actual physicians.
This is a perfect example of where generative AI works best, taking in information from the user, interpreting it quickly, and passing on that analysis to a human. The public sector is often guilty of trying to reinvent the wheel, but utilising generative AI in specific, limited parts of the journey to uncork bottlenecks is where it’s best placed. Like chilli powder, sometimes a dash to spice things up is just enough.
A further example includes the monitoring of real estate; a huge manual undertaking involving repeated site visits, red tape and crucially, time. France, the United States and Australia are each using AI-powered tools to monitor undeclared changes in real estate. The French project recently allowed officials to uncover more than 20,000 undocumented pools at a time when the nation’s reservoirs and water supply suffer shortages. This project allowed the government to collect more than €10m (£8.5m) in revenue and is being expanded to monitor other types of illegal and undocumented modifications to real estate.
There is clear potential in the public sector for AI to make a discernible difference, which will only benefit the general public. Like any sector, if you standstill you are going to be left behind. For the sector that means not being able to deliver services. If the rest of the world moves on, developing products using AI, while the public sector lags behind, it is going to take a long time for the sector to catch up.
And we are already seeing examples of AI being put to good use, with product thinking enabling true creative thinking and application of the technology.
The public sector is well placed to be a real trailblazer when it comes to AI – let’s grab that opportunity by the digital horns.