Two such firms - Lloyds Banking Group and M&G Prudential - took to the stage this week at a Fractal Analytics event in London this week to share how they are thinking about an approach to AI that makes sense not just for data science teams, but for the business.
Throughout their discussion both Lloyds and M&G spoke about, and shared learnings on, how they’re creating an AI/ML community internally within their organisations, as well as discussed the importance of getting the C-Suite on board.
For example, Priyank Patwa, Head of AI & Machine Learning, M&G Prudential, said:
“My role is about creating an AI/Machine Learning community. The idea is moving from a central function that does everything, to a bottom up, where our businesses are empowered to adopt AI. Our vision is that our AI will be something like the spreadsheet was in the 1990s.”
Whilst Abhijit Akerkar, Head of Applied Sciences, Business Integration, Lloyds Banking Group, said:
“I’m in the matrimonial business. I marry machine intelligence with human intelligence, to fundamentally change the way we do business at Lloyds Banking Group. To create completely new opportunities for our 23 million customers and 75,000 colleagues.”
Where do you begin?
M&G’s Patwa said that part of the solution is adapting your language and approach, depending on who you are talking to. For example, a fund manager will want to know the art of what’s possible to make more money using AI, whilst a data analyst will be interested in discussing models. However, he said that the financial services industry is still on a learning curve, when compared to the likes of Google or Netflix. Patwa said that the challenge is scale.
“We are doing fantastic stuff in pockets, but the biggest challenge for the financial services, is getting out of the that trap to making it a reality. This is not new for finance - we are a data driven firm. We couldn’t make any decisions if we didn't have data. But what’s different is the new techniques, which is what is very relevant for us. The ‘how’ is the most interesting bit for finance.”
Akerkar said that at Lloyds, his approach is to help business leaders reimagine their business with the help of these new technologies, which translates into a long list of potential use cases. However, prioritising those use cases is important and dependent on a number of factors. These include:
“Number 1, what is the size of the prize for this use case? The second is, what’s the feasibility? That has different dimensions - do we actually have the right data? Is this use case proven or are we trying to solve for the first time? Do we actually have the API pipelines running through? Also, how excited is the business sponsor about the opportunity? Will this be approved through a risk and compliance committee?
“The final criteria is, will this create impact this year? A mix of this helps us decide priorities and create a portfolio. The mindset we follow is more of a venture capitalist. Let’s try it out with small ninja teams. See what the result is and if it is showing positive results, then double down.”
Patwa said that his journey at M&G has been an interesting one, whereby he was part of a team with just 2 people in head office back in 2016, where he managed to get buy-in from the C-Suite after they managed to predict which branch the firm should open in Uganda, using data models, whilst sitting in London. He said that this got him some traction in the firm. Patwa said:
“You have to start with a proof-point, rather than starting with ‘I want to do AI’. We failed a lot of times because we started with asking for money, or getting a consulting company, or creating a Powerpoint. The other part of the paradigm is that you need then to start quickly moving fast, you can’t just keep doing prototypes.
“That’s where I think you need to give them the imagination. It’s not about those couple of dashboards, it’s about where it’s likely to transform the whole business. That’s frankly a constant uphill battle. It comes down to human behaviour, mindsets that you’re trying to shift.”
Akerkar agreed and said that any AI team has to ‘talk the language’ of the business and focus on things like changing the balance sheet, or how AI can improve the customer journey - not just talking about technology.
The challenge of change
It’s fair to say that any new technology, but particularly AI/ML, has the potential to introduce a fair amount of disruptive change on the culture and processes at an organisation. At Lloyds Banking Group, Akerkar is particularly focused on how AI can better guide the decision making process internally, whilst not making the experience of employees and that knowledge learned redundant. He said:
“We look at machine learning and AI as the evolution of decision making. The ability to make better decisions. We call it the ‘last mind’ - translating your better prediction through the model into value. We need to change the decision making journey. How do we ensure that the right insights are available to the right person at the right time? How do we create the right incentives and right culture that those decisions are used? How do we ensure the right human-machine interface happens?
“For examples, we don’t want our experts to blindly trust the model output. Because by doing that it will be big data, small brain. And that is not what we want. They have accumulated years of experience. So, how do we marry those two particular things so that they’re able to create better forecasts? There’s a huge amount of effort that goes into it.
“We do this where the AI model runs in parallel to the traditional model and we get people involved in using those particular things. So by the time we switch off the traditional and move onto the AI model, we have learnt through the feedback look, what do we need to change?”
Akerkar added that he would rather focus on the internal, constructive disruption that AI brings, rather than worrying about the external disruptions. He believes that by doing this, and by introducing better decision making capability using AI, Lloyds will be able to weather any external storms better. The challenge he has is two fold - 1) scalability, and 2) data.
On the scalability point, Akerkar said:
“How do we move from grandma kitchen to an industrial kitchen? Catering to do 500 customers in the restaurant? How do we effectively scale what we are doing, so that we are able to create a number of models at a faster pace? And in a consistent fashion. How do we get create the pipeline of people on a regular basis, data scientists, machine learning engineers. It’s not just recruitment, it’s creating a name for Lloyds Banking Group on the campuses, where the brand gets associated with data science.
“It’s also establishing scalable processes, for example, data science lifecycle. When a problem comes up, do I have scalable data science lifecycle through which it moves?”
However, all of this is redundant if an organisation is unable to get access to the data it needs, or if that data isn’t useable for the use case. Akerkar added:
“If you look at data to value as a value chain, the first mile is data. We already have that in place to a certain extent. Although we might have a lot of data, getting that data ready through the data pipelines is moving a mountain for the first time.
“Once the data pipeline is created, the next problem you solve with that particular data, is going to be faster. It’s taking the right stakeholders along and getting the infrastructure in place to move ahead.”