Upcoming WebinarsMore Webinars
Oge Marques explaining recent developments in AI for Radiology
Author of the forthcoming book, AI for Radiology
AI Business is part of the Informa Tech Division of Informa PLC
This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 3099067.
by Ciarán Daly
LONDON - The British retail landscape is littered with stories of doom, gloom, and decline. In the first 9 months of 2019, the UK saw 85 000 retail jobs disappear as retail spend moves online and high street rents rise.
However, the retail sector is also one of the most promising areas for the potential application of AI technologies. From fulfilment, supply chain and logistics to personalisation and the customer experience, the industry is ripe for innovation. The obstacles, however, are manifold in a sector yet to undergo a fundamental shift in culture and business models.
Founded in 1869, Sainsbury's is the second largest supermarket chain in the UK - and certainly its oldest. With 1415 stores of varying sizes nationwide, Sainsbury's possesses a huge workforce and supply chain. Bringing the supermarket into the 21st century hasn't been easy, but the company are already taking steps to keep ahead of the curve.
To find out how this giant of British retail is approaching AI, we sat down with Udai Chilamkurthi, Chief Architect for the Technology Strategy & Architecture practice within Sainsbury's retail and logistics division, where he and his team aim to modernise the firm's technology estate and enable innovation and agility using new technologies. As we learned, AI is just one of these, and is linked directly to growing data and analytics practices across the company.
U: There’s a few things that are important to us as a grocery business. One is the value and cost of the supply chain. This currently uses information which has been heavily optimised with mathematical models over the past 20-30 years. This is great, but the business is evolving, customers are changing, and it’s very hard to reoptimise these things regularly without using something more intuitive.
What we’re trying to do is gather more insights into how we can flex the supply chain—how we can save money, improve availability, and deliver new business models. At the same time, we want to think about what this means for the world of logistics or retail availability on the shelf. This is a constant process of trying, re-trying, and optimizing to gather insights without the need for 500 people crunching numbers.
U: We have invested a lot as an organisation over the last couple of years trying to establish an organisation that deals with data, analytics, and data science, build that capability, and try to capture data from across the organisation.
Every organisation – not just ours – has a lot of hypotheses
about what they think they should do. Whatever the hypothesis might be, we try
to re-evaluate and test it pretty quickly using all the data we have access to.
The flipside to that is generating new hypotheses and perspectives. Say we want to replace product A with product B, we can go back and look at what happened in the last 12 or 24 months and measure all sorts of variables that may be having an impact.
Sometimes it works out, and sometimes it turns out we don’t have enough information to prove it either way.
U: From a technology perspective, a lot of the insights are transferable because ultimately, the other parts of the business are either directly impacting, or impacted by, the supply chain.
But there’s also a completely separate group of AI experiments occurring at store level that are customer-facing, from video analytics looking at customer moments within the store to monitoring replenishment rates.
These are very analytics-driven, they’re AI and they use machine learning, but I personally see those technologies as distinctly different from an optimisation AI.
U: To be honest, I think a lot of these challenges are more people and organisational, rather than technology-centric. The retail business, logistics, and store operations spaces have been long-established in the UK. They have an ethos and principles around how they do things.
So while there’s a lot of organisational interest in understanding the value of these insights, there’s definitely resistance to accepting that some things might have to change based on data science.
Even if we have a good hypothesis that we have proved one way or another, there’s only so many people who can understand the nuts and bolts of it. It’s hard to rely on somebody that has a Ph.d in data science and take their word for it—even if we can prove that we have to change X to Y. So there’s a big gap between having an insight and that actually translating into value, and the journey can sometimes be quite painful as a result.
"While there’s a lot of organisational interest in understanding the value of these insights, there’s definitely resistance to accepting that some things might have to change based on data science. "
Being able to visualise these insights for this audience can be very effective. Making insights more visual and understandable without using any economic or data science terms can have a greater impact than saying, “this is what the algorithm says.”
One way is to reference a use case or the customer’s journey
using data, which almost takes the conversation to a place where it’s not
debatable. People can be opinionated about what to do, but you ultimately need to
present the information in an appropriate way that makes sense either from a
customer or profitability perspective underpinned by data.
U: To be very open and honest, I don’t think the business is there yet in terms of using it as a tool. What we tend to do is pick very narrow use cases which have value for the business, test the hypothesis and drive it. So it is very much ad-hoc and driven by specific use cases, rather than something that’s embedded into the organisation.
We’re not there yet—I don’t think anyone is there yet, either from the perspective of our industry peers, or other industries.
There’s a need to look at the key drivers, rather than just
using AI or machine learning for their own sakes. Having a 0.1 percent
efficiency gain in how we fulfil our products from a supply chain perspective
will, in itself, mean millions in terms of the bottom line. The pot of money at
the end is quite big, and that’s the carrot at the end of the stick that we are
going to try and chase. We’re not there yet—I don’t think anyone is there yet,
either from the perspective of our industry peers, or other industries.
U: Logistics, supply chain, and replenishment are areas where people have started investing in AI and ML, but I don’t think it’s seeing large-scale adoption yet. However, within supply chain and logistics, there’s already a convergence between large, established companies either directly implementing AI or adding AI companies to their portfolio to improve that. So that’s going to happen irrespective [of adoption].
When it comes to the in-store and customer experience, that is a space which is already exploding. I think the challenge there is capitalising on the value of AI, as a lot of things can be done, but if I’m a business owner trying to look at my bottom line, I’m not sure if it’s going to give me value because I haven’t seen a sensible business case that makes sense yet. The technology itself isn’t mature and comes with its own problems, but I think the experimentation is going to explore and in 12 months, people will be asking, ‘what can we do with it? What’s the value?’
Udai Chilamkurthi is Chief Architect ofTechnology Strategy & Architecture, Sainsbury's Retail & Logistics. Catch him speaking at The AI Summit London, June 12-13
Based in London, Ciarán Daly is the Editor-in-Chief of AIBusiness.com, covering the critical issues, debates, and real-world use cases surrounding artificial intelligence - for executives, technologists, and enthusiasts alike. Reach him via email here.
Author of the forthcoming book, AI for Radiology