Page 33 - Food & Drink March 2020

“When an organisation pays attention to the these, AI is just one component of an overall solution. In essence, it’s a bunch of scientists writing an algorithm for a specific problem.”
THE IDEAL AI CANDIDATE
For clients wanting an AI solution, Complexica looks at: the number of customers they have; the number of products they sell; and the complexity around how they make and/or deliver those products to customers.
Michalewicz explains it is the complexity of a business not its size that is key.
“A company that sells 5000 products to 40,000 customers, with different customers needing different pricing and tailored service benefits from AI.
“It is a well suited solution when the complexity of a business’s operations begins to overpower the people working in it and they struggle to manage and stay on top of what needs to be done.
“So, our definition of a large organisation means complexity of operation in the number of products and customers.”
Michalewicz cites the work they did with PRW and Costa Group. For PRW, Complexica digitalised and globally optimised its supply chain across Australian, American, Spanish and New Zealand wineries.
“Optimising supply chain decisions across multiple operating sites and time zones was difficult and complex but it is driving significant benefits for business. Asset utilisation, volume throughput, consistent quality and working capital requirements are all benefiting as the company moves away from paper, whiteboard and spreadsheets,” he says.
With Costa Group, Complexica’s initial roadmap was digitalisation of its business processes. “You can’t add AI to a whiteboard,” Michalewicz says.
It has been a big project, with thousands of spreadsheets and disparate processes.
“We spent a couple of months researching what other companies were doing in the agricultural space, interviewed their customers and looked at what technology was available.
“It was comprehensive with close client engagement. Once you reach sign-off, the digitalisation process timeline can vary from two to four months out to a year, depending on the company and its business.
“Costa has a very physical business process in the fruit and vegetable trading markets. There are no digital components in that process. You physically came to the market, you physically handled the produce, you physically took receipt, delivery and you went on your way.
“What they wanted to do was digitalise the process completely so you could still do the traditional business process of attending the market, but also have a multi- or omni-channel experience, where digitally you could have as close to that experience in the market.”
“People have to stay at the centre of whatever we’re working on, we want to augment our own intelligence not replace it.”
Augmenting intelligence means giving a winemaker, or a fresh produce seller, a machine that can look at the things they don’t have time to look at and make suggestions.
Michalewicz says, “The applications we build for sales reps – in some companies there are hundreds of them – in companies like PFD, Dulux or Bunzl, have a system that analyses every single customer in the business.”
That can be 50,000 customers. Every night it looks at external data, social media profiles, buying patterns, deliveries and complaints. It takes in all the data sources available to it and applies the algorithm.
different customers, workforce change management challenges have far outweighed the technical ones. It is very difficult to change someone’s behaviour, how they do their job and to ‘trust the system’. Generally, people are apprehensive of change, some people are actually quite afraid of it.”
He says he has seen the full spectrum of human nature – the early adopters through to those with ingrained thinking who refuse to believe a machine could make a better decision. For Michalewicz, there are two ways of easing apprehension.
“Firstly, we all use AI every day without knowing it. AI in credit cards has been there since the 1980s. Car transmissions, camera lenses, GPS, Siri, satellites – it is used in all these things, but it has been behind the scenes. For those thinking it is new and uncomfortable, when they realise they have been using the technology all along much of their apprehension disappears.
“Secondly, as AI gains more visibility and is embedded in more business processes, people become more comfortable. Recommendation engines in Amazon or Netflix can be based on deep learning or collaborative filtering but regardless, we are beginning to trust those recommendations.
“What’s happening now is we are seeing that applied in a business context.
“When people are comfortable with it as a consumer, they become more comfortable with it as a business user.”
For Michalewicz, AI is simply a natural progression of our tech and science capabilities helping people make better decisions. ✷
(l-r): Matt Michalewicz and Costa Group South Australia state manager Andrew Christophides.
“ In all our years of experience we found that to deliver AI-based technology successfully, the first rule is to make sure AI is really the right technology to solve the problem you’re trying to solve.”
MATERIALS HANDLING & LOGISTICS
The project will mean everyone has access to a desktop, laptop, and mobile app to enable staff to look at inventories in the market, the different grades of produce and everything the Group is selling.
“Once you have the digital capability, omni-channel capability, then the second phase becomes the introduction of intelligence into that digitalisation process.
“That intelligence can come in the form of a personalised recommendation or pricing, dynamic pricing, identifying customers, what is trending and what sells well together.
“You’re introducing intelligence into a workload that is now digital, and when some things are no longer manual, they can even be fully automated.
“The way to look at complex business processes through the lens of AI is first to look at digitalisation, then move to embed AI and intelligence.”
Michalewicz also prefers the term augmented intelligence.
“Then overnight, it sends an SMS or email to each individual salesperson with an action list for the next day, covering things like problems, opportunities and which clients to focus on. The time that saves those sales reps is invaluable.”
Machine learning from embedded AI is the optimisation of decision support, he says.
For a PRW winemaker AI can provide valuable suggestive intelligence on which tank to use for blending or how many litres to bottle or pull from the tank.
“These are all decisions people would be making. AI is making that decision based on data and whatever KPI has been set.
“When a winemaker is trying to achieve the best KPI outcomes, having a system making based on myriad data sources is quite appealing,” Michalewicz says.
MANAGING THE CHANGE
Michalewicz says that in his
25 years in the industry, working in different continents, countries, businesses and with
www.foodanddrinkbusiness.com.au | March 2020 | Food&Drink business | 33