‘Accountants have a key role in making AI work’

Ahead of her appearance at the CA Summit this month, we catch up with Authentic Innovation founder Zoë Webster to talk regulation, bias and the new skills to cultivate

Words: Lysanne Currie | Photography: David Titlow

I’ve heard it said that AI is like teenage sex,” Zoë Webster says mid-photoshoot, in a vault below London’s Waterloo station. “Everyone is talking about it, but not everyone who claims to be doing it really is.”

It’s a surprising thing for Webster to say: she has just been named by AI Magazine as one of the top 10 AI leaders in Europe and has herself been thinking about and working with the technology for more than 20 years. Thirteen of those years were spent at Innovate UK, culminating as Director of AI and Data Economy, followed by three as BT’s Director of Data and AI Solutions.

Webster left BT earlier this year to set up her own consulting company, Authentic Innovation, to help businesses think pragmatically and strategically about the use of AI. “We are in a hype cycle,” she says. “And there’s a bit of Fomo [fear of missing out] around AI. Everyone thinks everyone else is doing it and doing it at scale and getting it right – and that’s not really the case. Everyone is in the same period of uncertainty.”

There is, Webster says, something of a frenzied “where can we put AI?” theme going on in boardrooms. “But AI is one of several tools,” she says. “It can provide insight from data, it can find patterns, generate content, enable conversations between chatbot and customer – but we must use it for the right reasons.”

“Human ingenuity, empathy and working as a team will really come to the fore as AI grows – we will always need the human connection”

Since launching Authentic Innovation, Webster has been helping organisations in sectors ranging from procurement to energy to charity. She looks for the “space of opportunity” based on an organisation’s goals, then assesses options according to what it already has in place, the potential value of candidate ideas, her own practical knowledge of AI and the feasibility of execution. “Something like ChatGPT is only one possible solution,” she says. “AI is far broader.”

So how should businesses approach an AI strategy? “I first ask clients to think about their goals: what is it they want to change in the organisation? Are they looking to attract more customers or maintain the current ones? What insights would be most valuable or what questions do they have? What would you need to know about customers to speak to them in a way that resonates?”

Next, Webster looks at the available data. “AI is nothing without the data,” she says. “But there’s a lot to think through: where is the data coming from, is it representative, is it sustainable, do you have the rights over that data, could it be biased?” 

Webster next looks at what form of AI makes most sense for the particular company, being clear all the time that it is not some kind of magic bullet: “AI is not the answer to all your problems – if you have rubbish data then AI is not going to help. If, however, you want to make sense of the data you have got, and you want to get insight to make things better for your customers, for colleagues, there are a range of opportunities.”

Group think

The one thing Webster stresses to all companies is that AI adoption is not a solitary affair: “It’s a collaborative effort and it’s really important to have different skills around the table: law, marketing, HR, change management and, of course, accountancy.”

Finance professionals have understandably been anxious about the advent of AI, worrying predominantly that it will take their jobs. Webster believes strongly that the situation is, in fact, the opposite – that accountants, in particular, are crucial for the technology’s growth.

Return on investment (ROI) is one area which needs the skills and training of CAs. With organisations finding measuring ROI challenging, how best to assess AI’s value and how to account for it?

“When I did my degree I was one of two women doing computer science, but I haven’t seen radical progress. AI is still pretty male dominated”

“Sometimes it’s straightforward, say, if a company ran an AI model and attracted 5% more new business. But often it’s much more difficult to assess – for example, how it also affects productivity within wider systems and potentially impacts society more widely,” she says.

Webster adds that costs are often uncertain and accounting for innovation is inherently difficult. She also points out that AI lacks the personal touch that CAs use in business: “Human ingenuity, empathy and people working as a team will really come to the fore. We will always need the human connection. And as AI grows, these soft skills will be increasingly critical.”

Webster urges everyone to remember that AI’s knowledge can be limited. “Machines can only learn from things that are in digital form,” she says, “so unless something is in noughts and ones, they can’t learn from it. Think about all the experiences you have learnt from that are not digitised – things that have gone well or not so well that are not accessible to machines. There is so much that machines don’t know. They have harvested a lot of data but there’s a lot of knowledge and life experience that’s within us all – but not on the web.”

Education
Studied computer science at University of York and master’s in intelligent systems at University of Sussex

2000
Joins Defence Evaluation and Research Agency as Research Scientist, moving into its spin-off QinetiQ as Lead Researcher; undertakes PhD in AI

2007
Appointed Lead Technologist at Innovate UK, rising to Director, AI and Data Economy in 2018

2020
Moves to BT as Director, Data and AI Solutions

2024
Leaves BT to launch Authentic Innovation consultancy

Webster believes if we are to have widespread adoption of AI, we also need regulation – part of which is about ensuring that accountability lies in the right place. “It may be a lot of fancy maths, but it’s easy to forget there’s a whole supply chain involved with AI,” she says. “For example, where’s the data from, who’s supplying the algorithms, who maintains it, who’s keeping track, and what happens if something degrades over time or customer behaviour changes? Making sure the right people are accountable in the right way is really important.”

The EU AI Act came into force in August, aiming “to foster artificial intelligence and deployment in the EU”. It’s a framework for use across all member countries and states that it “provides developers and deployers with clear requirements and obligations regarding specific uses of AI while reducing administrative and financial burdens for businesses”. Webster welcomes the Act and believes we’ll see regulatory change in the UK in the not too distant future. “There are emerging AI applications being developed by people with bad intent, including ‘deep fakes’,” she says.

However, what if it goes right? What if the adoption of AI becomes really useful, we start to rely on it more and more and it feeds into our daily lives? How to we make sure we have safeguards in place and we’re making the most of the opportunity?

Balancing act

One issue that needs to be solved from within the industry is AI’s bias problem. UN Women reported earlier this year on a study by the Berkeley Haas Center for Equity, Gender and Leadership. It analysed 133 AI systems across different industries and found 44% of them showed gender bias and 25% exhibited both gender and racial bias. (Indeed, when CA’s art director put “entrepreneur” into ChatGPT, a young, white man emerged.)

Webster believes there are two big obstacles that need to be overcome. Not only is the underlying data itself a problem – “It comes from a western view of the world which can be full of poor data and horrific stuff,” she says – but also there is not enough diversity among the people working in that space. “We need people in the room to be able to say, ‘I’m not comfortable with that.’”

More diversity could go a long way to balancing the bias but, says Webster, EDI progress is stalling across business in general – and in some cases is going backwards.

“I’ve been in this space for a long time,” she says. “When I did my degree I was one of two women doing computer science, so I am used to it. But I haven’t seen radical progress. AI is still pretty male dominated.

“There is lots of amazing work being done out there by people such as Maggie Philbin [the former presenter of BBC’s Tomorrow’s World] and Anne-Marie Imafidon [Stemettes founder]. But if you think about who is going to play around with algorithms at the weekend to gain experience, it’s less likely to be a woman with kids or anyone with caring responsibilities. When do women have time to experiment with it?” 

Webster believes time carved out in the working day through training is realistically the best way to get more women playing with AI.

If these problems can be addressed, AI is a “fundamentally useful technology”, she says. “It harnesses lots of different varieties of maths to solve problems, but as it grows we’ll increasingly need human context and human understanding. The organisations who are going to win will be those who bring those things together in the most effective, inclusive, empathetic way.”

linkedin.com/in/zoewebster

Zoë Webster will be talking about AI at the CA Summit. Register here