Optimisation by design: AI as a critical enabler in UK regulated infrastructure

Driving business performance subject to price and service regulation is a complex task, especially when managing critical infrastructure assets. AI can transform this challenge into a competitive advantage. Here’s how.

2024-07-25Infrastructure & Environment
Andrew Glennie
Customer Director, Energy Transition and Environment

At the heart of regulation is the ambition to deliver better consumer outcomes. However, achieving this in practice requires regulators to navigate a delicate balance between ensuring appropriate safeguards, while being cautious not to stifle investment and innovation.

On the face of it, innovation and regulation might not appear to go hand-in-hand. But there’s an exception when it comes to artificial intelligence (AI). Let me explain why.

The Context

In this article, I’ll focus predominantly on UK Regulated Infrastructure: Aviation, Electricity and Gas, Water and Wastewater, Rail and Telecoms networks, where Faculty have worked extensively in recent years. However, much of this discussion is also relevant to other sectors where safety and security of supply are paramount, such as Healthcare and Financial Services. 

These are all critical industries that are subject to ‘Economic Regulation’, whereby regulators act in advance to specify the combination of price and service that entities must offer, with accompanying incentives and penalties. See, for example, Ofwat’s recent Draft Determinations for the water sector in England and Wales.

Why? In short, these sectors are natural monopolies characterised by high fixed costs, economies of scale, and where duplication of infrastructure isn’t typically viable. Regulators therefore act to promote (or enforce) competition, drive efficiencies and ultimately to protect consumer interests.  

In such safety-critical environments, with billions’ worth of infrastructure under management, it’s understandable why decision-makers might lean towards risk aversion. Especially given the need to meet testing service obligations with a constrained spending envelope. 

I would like to make the case for the opposite. Namely that the presence of such constraints makes the role of innovation even more critical. And that regulation provides the conditions for AI solutions to become your key competitive advantage. Below I’ll explain why, with the aid of some examples.

Optimisation by design

Within these frameworks, regulated infrastructure companies are responsible for determining the best allocation of resources to perform against their service requirements, subject to the constraints placed upon them (e.g. expenditure allowances). It’s an optimisation problem by design. 

These constraints manifest themselves in complex operational trade-offs, for example:

  • Electricity and Gas: How and when should I schedule asset (e.g. substation or overhead line) maintenance, given capacity constraints and allowed expenditure?

  • Water and Wastewater: What actions can I take to minimise use of Combined Sewer Overflows, given current network capacity and engineer field force?

  • Rail: How can I identify and prioritise pre-emptive interventions (e.g. vegetation cutbacks) given my existing budget and resource envelope?

  • Telecoms: How can I optimise traffic flow across my data network to best serve customers?

  • Aviation: How should I optimise my equipment inventory and workforce to minimise aircraft downtime and disruption?

While the regulatory framework differs, there’s of course a read across to other industries such as healthcare. For example, how can I optimise the allocation of fixed resources: hospital beds, medical staff, and equipment during peak hours to handle patient admissions effectively?

Delivering superior performance in this environment requires solutions capable of navigating these complex trade-offs to inform decision-making. The stakes are high. Not only are these decisions safety and service critical, but modern regulatory frameworks are increasingly outputs-based, with revenue shares subject to delivery of industry-wide KPIs. Reactive approaches will not suffice.

If we look at the characteristics of the trade-offs above, they also all have several features in common: 

  • They are all complex, frequently repeated decisions

  • They are all influenced by a range of confounding (exogenous and endogenous) factors. 

  • They all relate to core business functions. 

  • The environment in which they are being made is constantly changing.

It is under these conditions that AI solutions excel.

Converting constraints into competitive advantage

Some organisations simplify decision-making in these environments through business rules. However, these approaches don’t work in settings where a) fundamentals are changing (see: the evolution of electricity networks) and b) the number of relevant features or choices far exceeds our capacity as human decision-makers. 

This is the core value proposition for AI solutions. They are able to consider the range of (often confounding) factors influencing your business outcomes. They learn from past decisions, and critically for regulated infrastructure, are able to directly account for your constraints. Designed well, they empower your decision-makers to take better actions. 

At Faculty, we specialise in Decision Intelligence, that is, custom-built AI solutions that we design to integrate seamlessly into decision-making processes. Our “Decision Loop” methodology for designing solutions effectively for decision-makers was recently explored by my colleagues John Gibson and Rosalind Berka. Critically in regulated infrastructure settings, we build solutions transparently, exhibited through our ongoing work with Northern Powergrid, SGN and Utonomy under the Ofgem Strategic Innovation Fund, which promotes methodological transparency and open data. 

So how do we go about implementing these solutions? Recently I spoke at a UKRI-hosted event “Applying Machine Learning and AI in the Future Energy System”, where I shared some lessons from our experience supporting regulated infrastructure companies to adopt AI, and how you can mitigate some of the typical pitfalls. In short:

  1. Don’t start with the tech: Identify a problem statement that resonates first, and design the tech to address it. It won’t work the other way around. 

  2. Seek to augment first, not replace: Keeping human decision-makers in the loop in the first instance maximises their expertise and showcases value early.

  3. Ensure you have the right governance: To maximise chances of success you need to mix operational users with business ownership and technical expertise.

  4. Be realistic: Build towards successful deployment in stages, and consider both technical and operational challenges in advance. 

  5. But don’t wait for data perfection: As an infrastructure company you’ll have a lot of data, which will never be in a perfect state. Don’t let this stop you from getting started.

In summary

Regulated infrastructure provides the conditions under which AI can deliver comparative advantage. On operational timescales, you broadly have to make the most of the network and resources you have, which is where better decision-making becomes your differentiator.

To sum up:

  • In regulated infrastructure, the onus is on companies to determine the right resource allocation to drive business performance, subject to constraints. It’s optimisation by design.

  • AI solutions are your competitive advantage in this environment, enabling you to navigate challenging operational and strategic trade-offs.

  • In adopting these solutions, start by developing a robust problem statement(s) that resonates with day-to-day users. Your data will never be perfect; start and learn as you develop. 

  • There’s significant ROI to target given the scale and criticality of your core infrastructure.