Darren Martin, chief digital officer of AtkinsRéalis, has spent his career focusing on how big data can solve real human problems. His entry to the field came not from a computer science background, or engineering, but from his work helping those with severe, and life threatening, mental health issues. 

Darren Martin first started using big data to solve human problems in his work as an intern clinical psychologist at Toronto General Hospital. “I’d gone there,” he recalls,  “Because at the time, it was recognised as one of the best eating disorder clinics in the world. And I felt that that was an area where I could apply some of my expertise. We also had a very well known anxiety and depression clinic that many people came to.”

In the daytime, he was working with groups of patients using cognitive behavioural therapy and other techniques, to help them overcome mental health issues that can, for some patients, ultimately lead to their death.

“The idea of that was to connect patient types together to help each other and support each other, rather than it just being a traditional Freudian sort of therapist and patient, patriarchal type of relationship. It was empowering people to help themselves and overcome problems.”

The therapies being used at the hospital could be life-saving. But it was hard to see which therapies were most effective, for which patients.

“We were dealing with people that had extreme depression and anxiety. We were dealing with patients who had eating disorders. Some patients have really serious health challenges, and some of them don’t survive.

“People with depression can take their own lives. People with anorexia have a whole load of health issues, and the fatality rate with that group is quite high.”

And the hospital, along with other health services in Ontario, was facing spending cuts. It needed to focus its work on the therapies that offered the most promising outcomes. “It made me really start thinking at a meta-study level,” Martin says. “How do I look at the data for the outcomes that are achieved across a range of treatments and patients, and how many of those patients are getting better, how many of these patients are coming back for more and more treatment, and in some cases, how many of those patients are not recovering?”

The research sent Martin’s career in a new direction. “I found myself by day working with these dynamic groups and learning a lot, and also helping people. It was very important to me to have purpose and outcomes.

“My analytical side would come alive at night. So my empathetic side during the day working with patients, and then my analytical side would come through in the night. I ran these studies and was trying to understand where the focus should be. “

Meta-analyses like these have been an increasingly important part of medical research since the 1970s, inspired in part by Scottish physician Archie Cochrane’s 1972 book, Effectiveness and Efficiency: Random Reflections on Health Services, which made the argument for more evidence-based medicine. Cochrane lent his name to an international network dedicated to this type of research.

By the 1990s, the techniques that Martin was using to compare psychological treatments, were rapidly becoming key to understanding performance in the public sector, and in business.

“These days, it’s called big data. And there are analytical tools and so on that are used. But in the beginning, this would be normal research practice. You’re looking at huge datasets as a researcher.”

And the development of AI promises to speed up research across datasets.

“The capabilities for the technology to come to those inferences on its own are also available. They can see patterns that can’t otherwise be determined using traditional statistical analysis.”

Meta-analyses beyond medicine

Martin’s work in healthcare and data science took him into the private sector during the later part of the technology boom around the millennium. He then returned to the public sector as a business transformation and technology consultant, focused on the challenges of allocating constrained resources effectively for citizens.

“Taking that forward from dealing with complex people and people issues to complex organisations, has been a passion of mine. And at the point of some of the government transformation that was happening around the cash crisis, around 2008, and beyond, there were some significant things that had to change in government to continue to deliver citizen services and to protect our state during a wave of terrorist attacks that were happening. “

He worked for a leading global technology company, serving the National Health Service (NHS) in the UK, and then later, as an independent transformation consultant, on a turnaround project for the UK’s Child Support Agency.

“It’s really important to kind of look at the whole system, and understand what needs to be done. So part of my journey has been working on cradle to grave solutions for the NHS, and dealing with major digital transformation for the whole of the NHS, and all the complexity that that brings, through to being involved in a turnaround for the child support agency at the time, which owed unpaid maintenance from its community of parents. And it was in a really difficult situation.”

He had been hired as an independent consultant to improve the agency’s reputation, and as part of that was asked to implement a multi million pound new customer relationship management system. But it soon became clear that the problems it had in its relationship with customers were not merely technological.

“I was leading on enabling a different digital solution to help them unlock that problem: to ensure that children receive money, and to help them thrive in their communities.” 

The initial goal had been to install a new CRM system. But that wasn’t going to solve all of the agency’s challenges.

“I walked around the floors, and I saw that after about four o’clock, there were very few people left to answer the calls in this public sector organisation. So for me, the real problem was that when I did the analysis, only one in 20 of the staff were available after five o’clock to answer phones, and the phones ran till eight o’clock. 

“Most of the parents that wanted to resolve their payments and correct incorrect payments or dispute certain things, typically, were in middle to low income jobs, and were tied to their jobs during nine to five. So when they were free to resolve these problems, only one staff member in 20 was available.”

Martin quickly worked out that there was no technological fix needed. Instead, he needed to identify a human problem and a human solution, and use technology to support that.

“It wasn’t about ripping out a CRM system. It was about using the millions that CRM system was going to cost to enable a significant portion—about 30% of the organisation’s 10,000 people, so about 3000 of them—to change their working hours so that they were available to answer more calls.

“If we got that problem sorted out at a foundational level, then any technology that can be applied was an even better win. And we could really be precise about the technology that was needed to solve some of the more fundamental problems. And so for me, this was kind of an interesting, complex situation where if you just run in with a hammer looking for a nail, you’re going to miss the point all together.”

The challenge isn’t to find problems your chosen technology can be used on. It is to identify the best solution for each problem.

Today, we’re at the peak of the hype curve for AI. The launch of ChatGPT, from OpenAI, and other forms of generative AI, have opened up a discussion on how these tools might be used. But, as Martin saw in the public sector, it’s vital to focus on solutions, not on technology in itself.

“If one is trying to just apply technology, for the sake of it, I think that’s really a futile pursuit. It’s really important to understand the outcomes that one needs to drive.”

An urgently needed tool

There are reasons to be afraid of the AI revolution. But we live in a world beset by challenges. There are plenty of problems that AI might be used to solve.

“I try to look at AI and the challenge that’s ahead of us. We need to enable a race to net zero, we have a huge dependency on carbon fuels and processes. We have a huge phenomenon of people moving to cities. We’ve got population growth and diminishing resources. And on our own humans cannot solve those problems.”

The tools Darren used in his meta-analyses made use of a machine’s ability to quickly spot patterns. Modern AIs, and generative AIs in particular, build on this pattern matching. But they do it in a very different way.

“What’s changed is the generative capability, the sense that the technology can not only mimic behaviours, observations and so on, that humans would otherwise do. We now almost have the concept of the AI generating its own designs, and not just mimicking but creating content.”

These are systems designed to handle complexity. They can work through thousands or millions of solutions, in the blink of an eye. A generative AI doesn’t just solve complex problems at speed. It can consider unorthodox solutions that no human would waste their time on.

“They can run through millions and millions of different scenarios, millions of what ifs—a crazy, crazy waste of time for a human being to do that. And it would take many lifetimes for someone to explore all the permutations that a generative AI could. That’s really where the opportunity lies.”

In the public sector, Martin had to find ways to use limited resources effectively. Today, as we move into the energy transition, it is not a shortage of resources that is the problem. The challenge is how to allocate a flood of investment, into one of the biggest crises humanity has ever faced.

“We’re in an unusual position at a global level in that, according to McKinsey, there’s about $130 trillion worth of investment available over the next decade to enable urbanisation, to enable the race to net zero, and to maintain a lot of the infrastructure that we already have.

“Part of the problem is accelerating the planning, the permissioning, the design of the new infrastructure that needs to be constructed, let alone building it. Our construction industry is still considered to be very latent in adopting the efficiencies that other adjacent markets and industries have. So this is really our time.”

Engineers have an unprecedented opportunity to help the planet, and humanity. But they face constraints, not in terms of finance, but in terms of labour.

“The labour and the skills that we need to move us on the race to net zero are in finite supply. There’s more money in the system. There are a lot of enabling technologies that could accelerate. And it’s just trying to solve the application problem, what problems do we really need to solve? Where does AI create an opportunity for us, rather than a problem? And I think it’s where we have design constraints, where human brains alone can’t solve those particular problems. I think that’s really powerful.”

We’re at a moment in time when technology has opened up a world of possibilities. Young people are growing up in a world of remote collaboration, learned in games like Fortnite ans Minecraft. A new generation will enter their professional or educational careers with these skills, and this will help us transform how we all work.

“That capability is going to be available in the workforce, and it will become a commodity. So when everyone can design and know how to use tools, what do we really want to be training people at university to do?”

Trust and governance

Generative AI is an advanced form of information processing. The idea that we can process information in a series of steps was first identified by the Persian scholar Muhammad ibn Musa al-Khwarizmi, a thousand years ago, as he worked in Baghdad’s House of Wisdom. He gave his name to these techniques, known as algorithms.

Al-Khwarizmi, like later computer scientists, could define each step taken in an algorithm, just as his successors could read through a piece of code. Generative AIs don’t work like this. They make up their own rules, within a black box.

“AI doesn’t know when it’s got it wrong, and it can hallucinate and it can make mistakes,” says Martin. “For those people that are really close to AI technology, I think they have no inherent trust of the technology whatsoever. Because they know it’s a probabilistic, deterministic technology that’s figuring out what is the likely answer.”

This is one time when we cannot just trust the process. We need to find new ways to verify and validate, without replicating the work the AI has done for us.

“I think it’s really important to understand what level of confidence one wants to apply around AI. Generating a substantive answer saves time, that then needs to be collaborated and enhanced and worked through to the next level.”

This is where skilled professionals are needed. And now, they can spend their time on tasks that cannot be replicated by machines.

“If I look at the engineering community, it’s a highly intelligent, highly qualified community, but actually, some of the work that they do at the basic level is quite repetitive and quite time consuming. Tapping into automation capabilities early on in that process with robust validation and corrections and assessments, allows those qualified people to enhance the design, once some of those basic things have been done.”

We shouldn’t blindly accept the solutions that AI suggests. We must remember that these are fast machines, not experts. 

History shows us that even experts cannot always be trusted. In the 1990s, and into the 2000s, the UK Post Office relied on evidence from computer systems to convict more than 900 people of fraud and related offences. This evidence was itself later discredited or undermined. But it continued to be used. This was a human failure, not a technological one.

“What’s come across clearly [from the Post Office Horizon IT Inquiry] is that there were a number of the examples where the problems with the system were identified, and the right people were informed and there was an awareness of it, but there was a human decision to ignore it, and disregard it.”

When we trust an AI, or any system, blindly, the fault for errors lies with us, not with the machine.

“What AI does is it allows us to spot anomalies faster than the human eye. And it’sup to us to look at the information and make ethically sound decisions based on information to decide for the right reasons when to ignore it and when not to.”

It’s vital that we maintain this human element, in our society, and in our workplaces.

“That will be key for firms like AtkinsRéalis, that you will be able to authenticate and validate and so on without replicating the work.”

The right tool for the job

Throughout his career, Martin has focused on how information processing tools can help us achieve human outcomes. In Toronto, he helped identify effective therapies, using meta-analysis. For the UK government, he introduced new technologies, but also identified where humans were needed. In his new role, as CDO of AtkinsRéalis, the meeting of the human and the technological will remain key. Architecture, engineering and construction, professionals must identify what work can now be done by machines, and what must remain in human hands.

“I think that AI will impact organisations like AtkinsRéalis in many ways,” he says. “I think we need to look at those commodity workloads, traditional time and material workloads that will be automated in a race to the bottom. It’s important that we don’t just chase the traditional business and traditional models in the hope that they will continue. “

Engineers must not get caught up in excitement about technical capabilities, but rather must think about what these capabilities can contribute. Then, they can share in the outcomes of clients and other stakeholders.

“We’ve got to turn AI disruption into shareholder value,” says Martin. “Our commercial strategy has to mature rapidly. I think in some cases, we’ll lower the price to gain market share. We need to share outcomes and gains with our clients when we innovate and solve problems faster. 

“I think there are certain markets that we want to move into, emerging markets, where there just aren’t enough people on the planet to solve some of the challenges we face. AI will help us scale global cities and help us with decarbonisation of the planet. We should really be focusing on using AI to accelerate that journey.”