Standards-based governance of commercial and pre-commercial procurement can enable sensible adoption of Artificial Intelligence (AI) to improve public service delivery.
Blog post by David Kershaw and Warren Smith, from the Cyd team.
On Thursday 9 November 2023 we delivered a webinar on ‘Procuring Artificial Intelligence (AI) in the Public Sector’, in collaboration with the Centre for Digital Public Services (CDPS). This blog post provides some further reflections on the subject.
Scene setting
Digital, and procurement, are two key public sector enablers to achieve a more equal, more sustainable and more prosperous Wales.
Wales is widely considered a leading nation on sustainable development, having enshrined in law the ambition, permission and legal obligation to improve social, cultural, environmental and economic well-being.
This provides the overarching framework for a broad range of public bodies’ activities, including digital and procurement in Wales.
Digital in Wales
The digital strategy for Wales sets out how digital, data and technology will be used to improve people’s lives.
Missions 4 and 6 of this strategy respectively cover:
- Economic prosperity and resilience by embracing and exploiting digital innovation, including AI and machine learning (ML), as well as procurement practices and policies; and
- Service improvement by working together, with data and knowledge being used and shared – including using data-driven innovation, to support the adoption of automation and AI solutions, done ethically and with integrity.
A key element of this strategy, and critical to the delivery of better services, is the role of CDPS and the Digital Service Standards for Wales. This is structured around three high-level elements – meeting user needs, creating digital teams, and using the right technology – setting out what’s expected from new or redesigned digital services funded by Welsh public sector organisations.
Procurement in Wales
Various reforms are taking place that affect the way public procurement is undertaken by Welsh public bodies. Alongside the Wales Procurement Policy Statement (WPPS) and the Well-being of Future Generations (Wales) Act 2015, two new pieces of legislation have received Royal Assent this year:
- The Social Partnerships and Public Procurement (Wales) Act 2023; and
- The Procurement Act 2023 (which next year will replace the current Public Contracts Regulations 2015).
Once Royal Assent is received for the Health Service Procurement (Wales) Bill, this will join the legislative framework for procurement in Wales.
Overall, this framework ensures Wales is in the best position to use procurement as an effective lever to achieve its policy ambitions.
As previously trailed, this links to Mission 4 of the digital strategy for Wales, which also sets out how procurement practices and policies will:
- Support innovation and economic prosperity, allowing businesses in Wales to thrive; and
- Support the public sector to work with a responsive market of companies.
“Life moves pretty fast…
…If you don’t stop and look around once in a while, you could miss it.”
In the context of AI, this meme from the hit mid-80s film ‘Ferris Bueller’s Day Off’’ couldn’t be more apt.
Today marks the first anniversary of OpenAI’s launch of ChatGPT (other generative AI (GAI) tools that use Large Language Model (LLM) algorithms are available), which now has over 100 million weekly users.
Three months after its launch, ChatGPT’s Application Programming Interface (API) and OpenAI’s Whisper API (a multilingual speech-to-text GAI capability) became available. There’s now over 2 million developers building these LLMs into their apps and products.
Gartner identifies ‘Emergent AI’ as one of the top themes in its Hype Cycle for Emerging Technologies 2023, which includes:
- GAI;
- AI simulation;
- Causal AI;
- Federated ML;
- Graph data science;
- Neuro-symbolic AI; and
- Reinforcement ML.
According to Gartner’s research and analysis, these technologies offer immense potential to enhance digital customer experiences and make better business decisions, among other opportunities.
With any emerging technologies it’s important to be able to discern what’s commercially viable now and in the near future.
Way before a procurement starts, technology decisions made can have a huge impact on how organisations create, iterate and run services. This can also significantly affect wellbeing outcomes and impact being achieved throughout the life of the contract, particularly if locked into using specific technologies.
It’s therefore imperative that practitioners in digital, data and technology teams, and commercial and procurement teams, work together to optimise their decision making.
Good commercial practice, whatever the procured technology
Throughout the webinar we referenced the guidelines for AI procurement developed by the UK Government’s Office for AI.
These guidelines contain specific considerations to address throughout the commercial (and pre-commercial) lifecycle when procuring AI capabilities:
- Algorithmic transparency;
- Process-based governance frameworks and auditability; and
- Model testing.
If you’re thinking of procuring AI capabilities, we strongly encourage you to read these specific considerations from the Office for AI’s guidelines.
In our view, aside from these points, the rest of this guidance represents digital commercial fundamentals that should be applied irrespective of what digital, data or technology capabilities are being procured.
In an increasingly fast paced and ever-changing landscape of new, emerging and evolving digital, data and technology capabilities, situational awareness is vital.
Understand the landscape
Both the Digital Service Standards for Wales and the guidelines for AI procurement link to guidance on choosing technology to help understand this ever-changing technology landscape, by:
- Making value-chain maps;
- Using prototypes to test what best meets users’ needs; and
- Allowing for evolution.
Value-chain maps help you understand systems (technical architectures, organisation structures, etc), particularly parts of those systems that will change frequently (e.g. GAI) and those that are mature enough to treat as stable (e.g. cloud compute and storage).
Value-chain map courtesy of Simon Wardley, made available under the Creative Commons ‘Attribution-ShareAlike 4.0 International‘ license
Test what it’ll take to meet users’ needs using prototyping to get an early understanding of the different components in your map. These should be made by diverse multidisciplinary and cross-functional teams, and should be regularly revisited to re-evaluate decisions, as things within the map will change. Guaranteed.
It’s important that commercial, procurement and contracting approaches are designed to reflect this, can accommodate evolution and be able to benefit from innovation.
Applying these practices across Wales
Aligning digital, data and technology procurement with the Digital Service Standards for Wales and the WPPS, will help:
- Achieve the 7 well-being goals;
- Support the 5 ways of working; and
- Advance research and development (R&D).
Achieving the 7 well-being goals
To do this:
- Start with a problem statement and outcomes to achieve, which meet the needs of and benefit the people of Wales;
- Make user-led technology decisions that enhance the team’s ability to meet people’s needs, today and for future generations;
- Consider ethics, privacy and security throughout; and
- Assess public value through the 4 lenses of social, cultural, environmental and economic well-being.
Supporting the 5 ways of working
To do this:
- Ensure every team has a diverse mix of people, experience, expertise and disciplines;
- Build and maintain a culture of responsible innovation, including when working with supplier partners;
- Make decisions in diverse and inclusive, multidisciplinary and cross-functional teams, at every stage of the commercial / pre-commercial lifecycle;
- Engage honestly, openly and constructively with the market;
- Participate in and actively contribute to community-based collaborative networks to share better working practices, knowledge and lessons learnt, such as:
- Cyd;
- The Crown Commercial Service’s Digital Data and Technology Buyers’ Community of Practice;
- The Office for National Statistics Cross-Government and Public Sector Data Science Community; and
- The Small Business Research Initiative (SBRI) Centre of Excellence collaboration network.
Advancing R&D
We’ve referenced ‘pre-commercial’ several times. Pre-commercial procurement allows the public sector to incrementally and competitively contract for R&D capabilities from industry innovators, where no readily available commercial solution exists in the market.
In Wales, the SBRI Centre of Excellence funds R&D into new, innovative solutions to address unmet needs, in this case within healthcare. This is hosted by the Betsi Cadwaladr University Health Board (UHB) in Wrexham and the Life Science Hub Wales in Cardiff.
This particular SBRI Centre of Excellence case study on AI technology to improve prostate cancer diagnosis caught our attention. As a result of this funded challenge, Betsi Cadwaladr, Swansea Bay and Aneurin Bevan UHBs began their own testing and development, and jointly became the first to use the application to assist in diagnosing patients. Further ‘Scale and Spread’ funding then enabled the pilot to expand across NHS Wales, including Cardiff and Vale, Cwm Taf and Hywel Dda UHBs.
This SBRI example chimes with key findings from recently commissioned research and analysis by the Centre for Data Ethics and Innovation, on public perceptions towards the use of foundation models in the public sector. Research participants felt most positively about use cases that directly benefited the public (e.g. advancing healthcare R&D).
Although the SBRI sits outside of the current Public Contracts Regulations 2015, more innovation-friendly procurement processes are anticipated as part of the new Procurement Act 2023. Through what’s termed the ‘Competitive Flexible Procedure’, public bodies will be able to:
- Design and run a process that reaches the best solution, through developing their understanding of what’s available;
- Integrate phases within the process to improve collaboration, including negotiation or dialogue;
- Run a multi-stage procedure, including R&D stages, with interim assessments and funding; and
- Refine award criteria as the process develops and understanding improves, before a final assessment to establish the ‘Most Advantageous Tender’.
Interesting times ahead
In conclusion, for any technology but acutely so for AI and other emerging capabilities, it’s imperative that standards-based governance is used to optimise decision making and collaboration, throughout the whole commercial and pre-commercial lifecycle.
This will help to maximise transparency, objective and evidence-based approaches, and competition. In turn, this will also help to promote a culture of integrity, responsible innovation and responsible business conduct.
However, to anticipate, detect and mitigate unfair biases in AI systems, it’s critical to ensure that teams developing and deploying such capabilities have a diverse mix of people, experience, expertise and disciplines, and decisions are made as a team.
This QuantumBlack report from December 2022 found that “increasing diversity on AI teams is a work in progress”. So in anticipation of a soon-to-be published update, it will be interesting to see what progress has been made in this vital area since last year.
2024 will be an exciting year for digital, innovation and procurement reform, which will drive behavioural and cultural change across whole organisations in Wales. We look forward to sharing further reflections on these areas in the near future.
What’s next?
For more information about the CDPS project on secure and ethical automation and AI, click here. The next CDPS webinar in their AI series will focus on managing bias, and will take place at 9.30am on Thursday 7 December 2023. To book your place, click here.
Join the Cyd user testing group by clicking here.