There is a great deal of buzz around Artificial Intelligence – or AI – at the moment, with many leading figures in wholesale saying it is the next big thing to hit our industry. But is it?
In the tech space, AI has been discussed for many years and – like many terms before it – inevitably means different things to different people. “Actionable Insight” used to be my favourite example of this – everyone referred to it but very few achieved it or could even verbalise what it was. Ironically, it also shares the acronym AI! “Digital” is another, a word that is highly subjective in its meaning from one business, or person, to the next.
But back to Artificial Intelligence, what is it? It is a broad term for several technologies and methodologies, encompassing machine learning, natural language processing and data analytics. What it isn’t, yet, is computers walking around, emulating human behaviour, and expressing emotion, abstract thought or sharing philosophies.
So, where is AI already impacting Grocery? A subject close to my heart – business intelligence, or BI, is already totally embedded into wholesale and retail. Sales reporting platforms (Dunnhumby, IRI, TWC) are basically data analytics environments that are processing and reporting “big data” so that wholesalers, retailers and suppliers can understand sales performance and base business decisions on sales facts.
Product recommendations – anyone who has used Amazon has experienced “customers who bought product x also bought product y”, which is a form of machine learning. What Amazon’s analytics is doing is watching for patterns in purchasing behaviour and, the more that two disparate products are purchased together, the more likely it is that this occurrence will become a cross-sell product recommendation on the Amazon platform. Product search functionality on eCommerce sites is another great example. Algorithms in search engines learn word correlations over time so that a search tool will start to connect Coke 330 with Coca Cola 330ml can or that Dairy Milk is a chocolate bar, not a carton of milk for use in tea and coffee.
All these functions within Artificial Intelligence are based on maths and probability. An IT developer has created a base programme and written logic into that programme, a little bit like starting to create a dictionary. Over time, the logic library increases as the programme is used in real life and more correlations are discovered and added to the dictionary. Another way of looking at it is that it is a little like building a set of rules – computers are still binary remember – so x = y or Coke = Coca Cola, systems have simply got cleverer and are now able to cope with millions or trillions of rules in hundredths of seconds.
My caution with people getting excited about Artificial Intelligence is two-fold. Firstly, our industry is awash with data, data makes up the very building blocks of AI and yet we are still working out how to harness it and use it effectively. Before any Artificial or Machine Learning can be applied, all of your data must be clean, tidy and in one place – a step that in itself can prove challenging (and coincidentally is something TWC can do for you, we talk about the need for accurate data in a previous blog).
Secondly, it feels to me as if people think that it is an instant solution to growing sales or a “magic box”, if you will, that you can install into a Head Office and it will run itself and leap to amazing conclusions that will transform customer purchasing behaviour. Plot spoiler, even Amazon does not have a magic box, instead it has a team of very capable developers who are working flat out on spotting correlations and adapting their technology around emerging customer behaviour.
But how can AI help wholesalers and retailers? This brings me back to the heading of this article; the first step must be to effectively harness the data that we are already sat on. Applying some rules to that data and starting to build predictive analytics is an ideal next step for our industry, in my opinion. We should really sweat the business intelligence environments that wholesalers and retailers have invested in and start to really push the boundaries on reporting capability.
What would this look like and what would it deliver? About four years ago, TWC started to investigate predictive analytics and devised some exciting applications. What if you are a national account manager looking after a large, national wholesale group. You know what your year-on-year growth target is for each of your skus and you can probably have a rough guess at how incrementally month-on-month performance needs to be delivered in order for you to hit your 12 month target. These facts can be added to the database so that the platform can tell you each month whether you have hit or missed your target. You would no longer need to go in and look at dashboards, instead you could receive an email with top line data that informs you how the wholesaler is performing against your targets and you could dive in for more detail if required.
Equally, a wholesaler knows what its minimum stock holding needs to be on each sku and a simple mathematical calculation can assess current run rates against stock to produce red flags when a product is in danger of falling below minimum stock levels. If that wholesaler also has a retail estate, the same calculation can be executed at store level. Again, it is a relatively simple piece of functionality for a reporting platform to monitor retail stock, adjust wholesale fulfilment accordingly and inform the supplier when a new delivery is required and let the retailer know when it is being replenished.
It sounds quite basic when I describe it like that, right? That is because AI currently is fundamentally about taking processes we already do, automating them and allowing the analytics platforms to learn faster and work harder – and with fewer errors than human beings.
TWC is already using the components of AI in our platforms. We have fuzzy logic; we use predictive analytics; and we are also building correlations on consumer purchasing behaviour. However, our biggest challenge is not creating the technology, it is often helping our clients to change their internal processes so that they are comfortable embracing the insights our platforms generate and embedding a truly data-led approach to decision making.
So back to the original question – AI isn’t as complicated as it might appear, but what MUST come first is a business intelligence platform that ensures any automation and prediction is based on data that is cleansed and accurate.