Why public sector procurement needs a serious rethink to deliver on the promise of AI and tech

If two or more suppliers of digital services into government get together over a coffee or a beer, it’s only a matter of time before they start to share war stories about the sorry state of government procurement.

To be fair, that sentence could have been written in any year since government procurement began, and I suspect the topic will remain a perennial source of grumbles far into the future; such is the way of the world. Yet for a new government considering digital reforms to help reduce casework backlogs and rescue front-line workers drowning in paperwork, turning their attention to the unglamorous issue of how the state chooses commercial partners to work with could bring lasting results.

Afterall, government has done this previously, before important aspects of those reforms were rowed back on. In the early 2010s the Government Digital Service (GDS) introduced the Digital Marketplace, GCloud and cross-government spend controls that opened a route for modern digital agencies to bid and win contracts – an often overlooked factor in the early success of the GDS reforms. 

Having personally spent 20 years in government and now two years on the other side of the fence with Faculty, I would recommend four priority changes for how procurement can be a route to improved AI and technology implementation. First, we should end the multi-year, big prime mega contracts for digital data and technology, especially for new innovative technologies. Second, we should do more to prevent data infrastructure providers creating walled gardens around their technology. Third, we could prevent incentives for low pricing at the expense of quality in cutting edge tech markets. Lastly, we argue for more support for homegrown tech companies in UK government procurement processes. Let’s explore each of these in more detail.

1. End the impenetrable markets created by multi-year mega IT contracts – or at least require diverse supply chains

The last few years have seen a gradual movement back towards multi-year, very large digital contracts with big IT suppliers and management consultancies. These suppliers are intended as one-stop-shops for meeting each government department’s diverse technology requirements. Whether that’s support for a legacy software renewal, some additional capacity for new customer-facing service builds, or the use of Generative AI. 

The attraction of big contracts for government officials is understandable. Going through the procurement process has become a seemingly never-ending palaver. Their admin teams are shrinking, and those staff are needed for other priorities. Having a few giant contracts may seem simpler for accountability, but in reality big contracts often hide risks and make accountability harder when things don’t go to plan.

For those of us working in tech companies that aren’t multinational corporations or global management consultancies, this administrative convenience for officials is becoming a headache. Nearly two years after the launch of ChatGPT and a large dose of excitement in government about the potential for AI that resulted from this, we have seen relatively few AI projects in domestic Whitehall Departments or their arms-length bodies come openly to market. Instead, this work – where it needs assistance from external suppliers – defaults to their existing strategic IT contract. In many cases this is expected to continue for the next (e.g. five) years of that contract, regardless of whether the provider is truly the most skilled in these niche, rapidly evolving techniques, or whether this approach best serves the public.

Despite this, for specialist companies like Faculty (and there are plenty like us active in the UK public services market), the overall market remains fairly healthy, and the positive noises from the UK government are encouraging. There are sufficient areas where public service teams find it worthwhile to navigate additional procurement processes to access our expertise in AI implementation and safety. 

There are also opportunities for smaller more agile specialist tech companies like Faculty in those big contracts when they come up for renewal, especially when large primes have to demonstrate diverse supply chains that include cutting edge AI expertise. Many would prefer it if such large, multi-year contracts would be split up or reformed. But if they are to remain a feature of public sector life, then government should at least insist they include a coalition of partners representing smaller and more specialist companies – including those of us that have happily outgrown the strict definition of a Small or Medium Sized Enterprise (SME). Where these partnerships exist, the reality is that from the supplier side, working with these larger companies is often a pleasure, with genuine respect and engagement regardless of contributing company’s size or background. 

While the government could limit or add conditions on these types of large contracts, they could also target the incentives that drive officials towards this approach. The procurement and finance processes have become so byzantine and lengthy – 9 months for a procurement process is fairly normal – and any sensible CDIO would try and do as few as they can. As a result, only companies with a particular size and scale can afford to bid and promise the required breadth of services and people. The result is a set of closed procurement frameworks for digital work that limit pre-qualified companies to access projects. As a result these frameworks are the sole route to acquire large volumes of work, and are refreshed only every 2 to 5 years – a long period amid rapid technology changes.  

The digital colleagues in the public sector I used to work with want to work alongside the best in the business. They want to work with companies that share their entrepreneurial zeal. If government procurement and finance could shed some of their burdensome processes, public service buyers would have more freedom to choose either smaller companies where they met their needs or rely on big IT firms where they offer genuine advantages.

2. Prevent infrastructure monopolies from building walled gardens

It would be easier to say ‘prevent monopolies’. Acknowledging the economic harm of rent seeking behaviour from suppliers within monopolistic or oligopolistic markets is hardly controversial. But we all also live in the real world, and in some areas of digital services a constrained market is already a fact of life, and there are at least some network effects that bring practical advantages. For example, despite trying (and largely failing) a decade ago, government is unlikely to move away from the ubiquity of Microsoft Office for civil servants any time soon.

If we look carefully we can see in some areas like data infrastructure that a new wave of providers are offering solutions that can join up data using one platform across organisational boundaries. There are pros and cons at work here – any decision to do this risks embedding a new monopoly that will be extremely difficult and painful to move away from over time. But it can also result in rapid action – avoiding the quagmire of non-interoperable technologies, viciously guarded organisational silos and overly cautious assurance, finance and procurement processes that dog more incremental approaches to reform. 

Yet where monopolies exist in (data) infrastructure, government commercial teams, and those they report to, should be ruthless in policing the boundaries to ensure that any ubiquitous new networks have genuine open API access, with oversight from technical teams. This hard-edged separation between data and applications, including AI applications, is required to enable competition in digital, data and AI services, and to avoid the type of full-stack monopolies central government has only recently managed to move away from.

However, there remains a risk that the government is sleepwalking into contracts that allow infrastructure providers to create walled gardens around their core offers, inhibiting others commercially and technically from adding the value that is needed. Government’s commercial and leadership teams, no longer connected to government technologists as they were after the original GDS reforms, risk failing to insist on the right contractual clauses, asking the wrong questions and failing to police contracts with sufficient technical expertise. 

And there are arguments to suggest that infrastructure providers should be subject to more than internal governance to ensure these new data markets can thrive. The MHCLG digital planning team that I led previously did something similar in the 2023 Levelling Up and Regeneration Act in local government, with clauses around requirements to meet data standards to participate in the market.   

3. Prevent buying of market incumbency with (too) low pricing

The UK government does not operate a simple ‘lowest cost’ process for procurements, but instead has a fairly sensible balance between quality, cost and social value. A typical contract scoring balance is 60/30/10. Pricing is always an important area when bidding for government procurements, with contracts often won or lost in correctly anticipating and undercutting a rival’s price. This should be healthy – the system is designed to deliver value for money to the public and often it does a good job.

But in fast-moving markets, like technology (and especially data), companies with deeper pockets can gain incumbency with a public sector organisation by offering low prices before rates inevitably rise once the provider becomes deeply embedded in the service’s infrastructure and processes. While ‘abnormally low pricing’ is in theory investigated, it remains challenging for smaller companies to win contracts against such practices. And this behaviour is seemingly most prevalent in the cutting-edge markets where smaller, more nimble and niche companies like Faculty offer genuine advantages, but where ‘like’ is not always measured against ‘like’. Also, in markets like defence, the small proportion of spend on AI can be dangerously compromised when quality is the difference between a mission’s success and failure – yet 30% or so of the scores are judged on price. This is exacerbated by procurement scoring, where quality scores tend to cluster closely – a weaker bid might score 50 out of 70 points, while a stronger one scores 60. This makes the 30 points for cost disproportionately important, creating strong incentives to prioritise low pricing over quality.

The procurement process could be improved with better connection and alignment between the government technical teams that score quality and the commercial teams that score price (and social value) – ensuring overall bid scores reflect genuine differences in quality and value. Other reforms could include more dogged policing by both commercial and operational teams against ‘abnormally low’ pricing, especially in innovative markets to enable a competitive market to flourish. 

4. Support home-grown tech companies in procurement scores

Working in AI can sometimes feel like working in an industry without borders. Scientific papers from around the world get devoured by technical teams in companies like Faculty, who then rapidly apply those new findings to solve practical problems here in the UK. Regulation clearly has a localised aspect, but there is an ongoing global conversation about frontier models and public risk. At Faculty, we’re involved in red-teaming those models on an international basis. It is an exciting time to be working in this field.

However, new technologies, trade and geopolitics are deeply entwined and the UK would do well to understand the implications ensuring it has homegrown AI talent, especially in the face of unpredictable global relationships. Current access to global (including US) AI models and markets may not continue in a straightforward manner over time.

Faculty is proud to be part of a cluster of globally-impactful AI companies in the UK. But the government doesn’t make it easy to invest in its home grown talent. For example, UK government procured £379bn worth of goods and services in 2021/2, around 15% of GDP; so procurement decisions it makes can impact the whole economy. But Faculty – a company that has grown in the UK, pays taxes in the UK and helps develop the ecosystem of AI jobs in the UK – regularly competes for government contracts with international companies. These companies are headquartered elsewhere to avoid paying taxes that ultimately pay for public services, including the procurement processes they bid through. This can affect pricing, margins and the speed at which we are able to grow as a company.

To retain its current global position in this emerging and likely highly strategic market, there is an argument that the UK government should reflect home-grown advantage in its domestic procurement rules. For instance using the Social Value Act to ensure half of the required 10% of scoring allocated to social value is determined on whether the company is homegrown and headquartered here for tax purposes.

Conclusion

These suggested reforms shape only part of a long possible list of options. They reflect the experience of those of us in the company who have spent time in both government and in the private sector, and of course reflect the priorities of a London-based tech company of around 400 highly capable specialists that has been operating at the cutting-edge of technology in government and other sectors for a decade. 

However, Faculty is not alone in openly advocating procurement reform, and we see this blogpost as a contribution to a wider public debate about changes in this often overlooked area of government. Although there will be divergent views among newer, more nimble and smaller tech companies, we suspect at heart views on procurement reform are likely to be fairly similar.