What are the myths and misconceptions in using AI and ML in Public Services?
You've likely heard the phrase "data is the new gold/currency" thrown around in discussions about tech and business. This is not new - we heard it 7 or more years ago when “Big Data” was a favoured buzzword, but it has made a major comeback with the advent of mass adoption of AI. This catchy metaphor suggests that data, like gold, is inherently valuable. However, this comparison misses a crucial point: the value of data isn't in its mere existence, but in how it's used and understood.
Imagine finding a treasure map with no landmarks or scale. Without context, it’s just a piece of paper. Similarly, data without context has limited usefulness. For example, if a healthcare provider collects data on patient visits without information about treatment outcomes, the data won't help improve patient care.
The quality of data is just as important as its context. Let’s say you’re planning a garden. You gather loads of information about the plants you want, but if the data (like climate compatibility or soil needs) isn’t accurate, you’ll end up with a garden that doesn’t grow. In the same way, data must be accurate, timely, and relevant to be valuable.
Additionally, while one would most likely welcome “more” of something with inherent value, collecting more data isn't always beneficial. Take Amazon's Alexa as an example—it gathers vast amounts of user data, and cost the business a cool $10 billion in losses in 2022; or British Gas discontinuing much of its Hive IoT device portfolio. This demonstrates a crucial point: storing and maintaining data costs money. Not every piece of data collected leads to valuable insights, and unnecessary data can drain resources. We’ve seen this in our business, SkenarioLabs, in which we analyse buildings looking for energy efficiency retrofit opportunities - many people point us to 10-second smart meter data, but this isn’t really functionally more useful than weekly or monthly energy bills in deciding whether to replace boilers and windows - and massively increases costs and privacy concerns.
-
AI is Only for Tech Companies and Requires Large Teams of Data Scientists
A common belief among local government and public sector leaders is that implementing AI technologies requires extensive expertise and a dedicated team of data scientists. This perception can make AI seem inaccessible for non-tech organisations. However, this isn't necessarily the case.
Advancements in AI have led to more user-friendly tools that do not require users to have a background in data science. For instance, AI services and platforms, including from major players such as Palantir, are now available that provide low-code solutions, pre-built models and easy-to-use interfaces, enabling non-specialists to apply AI solutions effectively.
Organisations can also leverage partnerships with universities, existing suppliers, or even tech startups that have the necessary expertise. TechUK is a major supporter of these kinds of initiatives and has groups to create ecosystems to this end. These collaborations can provide access to the skills and resources needed to initiate AI projects without the overhead of building a large in-house team.
-
AI isn’t suitable for policymaking as the real world is too complicated
People have valid concerns about AI's ability to handle complex, nuanced decisions that typically require human judgment. However, the logical conclusion that AI is wholly unsuited for the policy environment is to miss how it can enhance current ways of working and help make decisions clearer.
AI excels at processing vast amounts of data quickly, offering insights that humans might overlook. For instance, it can analyse trends and predict outcomes based on historical data, which can inform more effective policy decisions. Rather than replacing human decision-makers, AI serves as a powerful tool that augments their capabilities.
In areas like environmental policy, AI has been used to model climate change impacts and develop more effective environmental regulations. At SkenarioLabs we utilise AI to help us to identify patterns in data which help us to do things like fill gaps in datasets - for example, where we are analysing all buildings in a region for a Local Authority for decarbonisation planning, data is not available for every single building. We utilise Machine Learning techniques to help us estimate the likely characteristics of buildings based on their neighbours and context using the information we do know about them.
While AI should not be blindly trusted to make autonomous decisions in sensitive areas, with the right framework for oversight and accountability, it can be a valuable ally in policymaking. This approach leverages AI's analytical strengths while safeguarding against its limitations.
-
AI enable us to fully automate services and drastically reduce costs
There's a widespread belief that AI can completely automate services and significantly cut costs. Articles such as this one from Deloitte suggest that 30% of savings could be made within 5-7 years, while containing helpful sections such as “What do government workers do all day?”. Considering that the work was published in 2017 it may be seen as an early example of over-egging the opportunity for making massive savings with AI.
To be able to achieve the very real opportunities which AI offers, we must be open an honest about what it can and cannot do. AI is typically better suited for partial automation, handling specific tasks within a larger workflow rather than replacing entire systems. Full automation is presently rare and usually limited to very controlled environments with predictable variables. Government services do not generally follow this pattern.
Fully automating services, especially in sensitive areas like healthcare or social care, raises ethical concerns and practical challenges. AI must be closely monitored to ensure it operates fairly and accurately, which can necessitate continued human oversight.
While AI can help streamline certain processes and reduce some costs, it is not a magic solution for full automation or financial challenges. Successful AI integration involves balanced expectations and a clear understanding of the technology's actual capabilities and limitations. For example, where positive outcomes are relatively straightforward and variables are known, such as public transport routes, there are good opportunities to have AI optimise routing.
Conclusion
As we explore the evolving landscape of artificial intelligence, it is essential to approach it with a balanced perspective. AI holds significant potential to transform industries, enhance public sector operations, and improve the quality of services provided to communities.
However, its integration into our systems and policies requires careful consideration, strategic planning, and ongoing oversight. By debunking common misconceptions and fostering an informed understanding, we can harness AI's capabilities effectively while mitigating its challenges.
Upcoming Local Public Services events
Insight NWT Simplifying the Complex - Accelerating the Public Sector roadmap for digital and data
techUK, 10 St Bride Street, London, EC4A4AD
Local Public Services updates
Sign-up to get the latest updates and opportunities from our Local Public Services programme.
Authors
