Dr Jon Rimmer, CXO at Mercator Digital, takes a look into the UK Government’s knowledge silos, their impact on service design, and what AI could do to help
During his tenure as HP’s CEO in the 1990s, Lew Platt famously said: “If HP knew what HP knows, it would be three times more profitable.”
We’re now decades on from that, yet the issue of drowning in valuable, underutilised research remains, with one of the most prominent examples being right here in the UK Government.
Could AI be the tool that finally helps us make sense of what we already know - and stop valuable insight gathering dust? This very question is one that businesses, institutions and particularly governments can no longer afford to ignore.
Knowledge silos and their drawbacks
It’s no secret that the UK government isn’t short on data. HMRC alone has amassed over 10 years’ worth of user research, equating to a powerful treasure trove of data that could help build new and transformative digital services for the general public. The challenge now lies in unlocking that value in a way that drives meaningful transformation.
The reality, however, is that many of the Government’s current processes - across not just HMRC, but all departments - relies heavily on ‘asking around’ or laboriously combing through old PDFs and free-text documents. This makes any attempts to pull out the useful bits from what are essentially fragmented archives very tricky.
Just imagine working on something for two years, for someone to then come along and say, ‘Oh, we did that last year. You just needed to add this.’ These kinds of examples - which happen far more often than you’d realise - not only lead to delays and duplication of effort, but also drive up costs. And ultimately, it’s the public who pays the price.
In order to change this, transformation at scale is needed - a complete overhaul of the way in which the government delivers its services across the UK. But this is a massive undertaking that demands a level of digital proficiency that government bodies have consistently lacked.
Just look at the Government Digital Service (GDS), which launched over a decade ago with bold ambitions of revolutionising digital delivery. Instead of rethinking services from the ground up, many departments opted for digitising old processes. The main reason for this is that the wildfire of delivering government services at pace has completely outstripped the knowledge, skills and experiences we have available. Put simply, there were not, and still are not, enough digital professionals with the experience and skills to make it happen. And the few who do exist are often promoted or move on, creating a vacuum that continues to widen.
Rebuilding services afresh is also expensive and inefficient. Because discovery phases are often under-resourced, rushed, and take on a one-size-fits-all approach, you’re left with services that lack a clear view of users’ actual needs, with more services that fail to deliver.
How AI can help
I’ll start by saying AI is not a silver bullet, but it does offer a solution to a very human problem: the difficulty of knowing what we already know. So instead of spending a week searching through folders, repeating research or endlessly re-scoping existing services, AI can tell you who worked on this, what themes came up, what’s changed in the last five years.
By affording us the opportunity to ring fence these knowledge pools and better use the knowledge stored within them, the potential gains here are massive. We can not only save time, and compensate for headcount and skill shortages - particularly during discovery phases - but when research is reused effectively, the benefits ripple outwards.
Smarter access to old data means more informed design, better service delivery, and fewer unintended consequences. It’s a shift that could stop the wasteful cycle of reinventing the wheel. It could also help researchers and policymakers stay aligned, creating feedback loops that adjust services as real-world needs evolve.
Aside from smarter service design, AI - if used appropriately - can also address another major issue at play, and that’s bias. Given the fact the internet itself is largely built on the experiences of middle-class, techie men - and is known for not being hugely representative - any open internet-trained models are at risk of reflecting sizable demographic and systemic biases. That’s why any AI deployed in this context must operate in a ‘walled garden’, trained only on internal content. This won’t just ensure that recommendations and outputs are based on real service data and not biased public internet content, but it will also address data protection and trust.
As many will know, public confidence in AI is very much up and down in general, especially as staff experiment with open AI tools in the absence of clear guidelines. But this worry is particularly strong when it comes to the government. So in order to trust AI, people need a bias-mitigated environment that contains only government data, ensuring secure and efficient knowledge sharing that stays within government walls.
Better services start with better use of what we already know
The bottom line here is that, while we know that the public doesn’t see the internal mechanisms of government, we do know that they feel the difference when services work better, faster, and more humanely.
AI, if used properly, can help achieve that - not by replacing human judgment, but by surfacing the human insight we already have in a fair, efficient and cost-effective way.
Dr Jon Rimmer
