Originally running in our 7th Edition Report, ex-Editor, reporter and contributor to WhichPLM, Ben Hanson, discusses the theme of ‘intelligence’. This piece accompanies a dedicated series from Ben around intelligence, A.I, and data-driven design and development in retail – all of which you can find in our 7th Edition.
[For any reference to timelines throughout this piece it should be noted that the publication date for our 7th Edition Report was August 2017. Similarly, you will find references to other ‘features’, which denote to the other editorial pieces in our 7th Edition Report.]
Just as WhichPLM did for both of our previous special editorial examinations (covering 3D in 2015, and the Internet of Things in 2016) the last exclusive feature in our 7th Edition acts as the final piece of the puzzle, collecting guidance, food for thought, and practical recommendations for retailers and brands who may be looking to lay the long-term groundwork for their own A.I. initiatives, or to embark on a particular, more pressing project.
The clearest question for prospective customers of A.I. solutions: are these viable products, with clear return on investment potential? Broadly speaking, the answer is yes. While general intelligence – a single machine to run everything, with mental capacities far in excess of our own, across essentially all of human endeavour – remains a pipe dream, more focused applications of narrow, specialised A.I. are limited only by customers’ ability to find the right technology partner and to gain access to their own information and broader market data in sufficient volume to deliver results.
But even if A.I. was more limited – its capabilities confined to being a better analytics platform or Business Intelligence tool, for instance – I believe it would still rank as an essential investment for many retailers and brands. As the first feature in this publication explains, the volume of information now available to us is rapidly beginning to exceed our ability to comprehend even its outlines – let alone the sheer volume of raw, real-time data that makes up its specifics. Indeed, many experts believe that we have already passed the tipping point after which human capacity will never regain the ability to interpret the international, cross channel flow of information without A.I. assistance.
“We have already passed that point, absolutely,” said Julia Fowler of EDITED, whose solution is expressly designed to shrink large sets of current information down to more digestible chunks of market insight and competitive analysis. “Bear in mind that retailers like Zappos and Saks Fifth Avenue are bringing around 1,800 to 2,500 new products to market each week, and I’d go as far as to say it’s impossible for a human to keep track of. The purpose of our software is to whittle that information down into targeted, real-time insights in seconds – as opposed to having retailers’ buying and merchandising teams browsing websites fruitlessly, then periodically visiting competitor’s stores. Working that way, retailers would have a lot of people spending a lot of time not accomplishing very much, while EDITED would instead provide more structured, objective information – as much as they could gather in an entire month otherwise.”
Fowler’s perspective on the explosion of data is also supported by IDC forecasts, which suggest that, by 2025, the aggregate total of data generated by all industries will reach 163 trillion gigabytes – a tenfold increase on the most recent available figure, for 2016. Among that data, close to 30% is expected to be classified as “critical” or “hypercritical” to our daily lives, and while the RFA industry does not deal in life-or-death outcomes the way the medical sector does, we can nevertheless see the vital importance of having the right information at hand, and the ability to effectively analyse it within a meaningful timeframe.
As you will have gathered from the previous features, however, A.I. is much, much more than a new iteration of analytics or BI. From machine learning serving as a designer’s information-gathering assistant, to a computer vision model generating images of entirely new products before they physically exist. From algorithms powering dynamically-adjusted pricing, responding in real-time to changes in demand, to A.I. technologies revolutionising recommendations, personalisation, and other customer capture and engagement initiatives.
It is no wonder, then, that one major school of thought – and it’s one that I subscribe to, with the likes of Vladimir Putin as company – hold that A.I. is a revolution on the scale we have not seen since the invention of the Internet. In that light, A.I. proponents argue, it is absolute folly to wait to take action – particularly when we consider that many retailers’ business models are already balanced on a knife edge, and every individual competitive advantage counts.
“Since retail is typically low-margin, it requires even more focus on efficiencies than other industries,” said Alexander Grey, who serves as Head of A.I. Research at Infosys (formerly Skytree,) which offers ‘turnkey predictive insights’ at fixed rates, using machine learning. “On the positive side, that means that the impact of business improvements brought about by machine learning (marketing, personalisation, customer loyalty, inventory allocation, hiring, fraud prevention and more) is probably larger than it is for other industries.
In other words, adopting machine learning seems especially urgent and critical for retail – and that is in the context that I don’t believe any industry can afford to wait.” Then there is the counterpoint – and it’s a valid one, depending on your perspective and business objectives – that if it has taken me, a researcher and writer whose job it is to cover new technologies in fashion, 15,000 words and a raft of expert interviews just to articulate what it all means, maybe the smarter money waits until more concrete ROIs are established, or until more accessible A.I. solutions reach the market.
“A.I. is the latest buzzword, and that’s led many retailers to explore the different ways that the technologies can benefit their businesses,” continued Julia Fowler of EDITED. “However, only a few actually have the capacity to take advantage of all the available data and use them to their full potential by themselves. A.I. algorithms require large curated datasets, top notch servers and infrastructure, and a team of experienced data scientists – all of which can become extremely costly. Not to mention the complications of managing something so different from a retailer’s core business. Having said that, if applied effectively, technology can bring many positive transformations to fashion businesses.”
Despite my optimism around A.I. in general, I agree with the contention that A.I. is so fundamentally different from any retailer or brand’s business that asking them to refocus and really understand all the umbrella technologies would be like asking them to quit making blouses and go into running backend banking systems tomorrow.
This has been a difficult subject to analyse and present. Far more so than 3D working, or even the Internet of Things, which at least have agreed definitions. When it comes to A.I., tenured professors and technology experts alike are prone to coming up short in truly explaining core concepts. Even the famous Turing test – named after British mathematician, hero codebreaker, and father of modern computing Alan Turing – stops short of deciding whether a machine can think, and instead gauges how well it can fool a human into believing it is thinking the way they are.
If these experts cannot readily define what intelligence is, I do not believe a brand or retailer should have to concern themselves with how their A.I. assistant reached the conclusions it did from the information it was fed. And neither do I believe that they should blanket-hire programmers and data scientists out of a blind need to understand and adopt the latest technology. Because, like most enterprise technology before it, A.I. vendors are already working out how to take the sting out of its tail.
“Most businesses are not currently staffed to actually implement machine learning in-house, so they will need to use third party service companies to realise their vision,” continues Alexander Grey of Infosys, who, among others, offer machine learning as a service, with the need for customers to invest in infrastructure and large numbers of in-house resources. “There are roughly three kinds of human activity needed in projects like these: starting with translation of the business problem to machine learning and evaluating the results, moving on to core data science work like modelling, and finally the usual I.T. issues associated with bringing in any new software. We focus on providing the first two pieces, with our expert data scientists, but this also has the effect of minimising the human time and skill needed for the actual execution.”
And while governments are clearly investing in pushing the envelope of A.I. and fostering this kind of talent, some are also interested in setting down regulations that will provide brands and retailers with a better understanding of A.I. processes without having to investigate them in-house. From summer 2018, the European Union may begin to ask companies that create automated decision-making systems to provide end users with an explanation for how those decisions were made.
As for the deeper philosophical questions, for now we can leave those to the philosophers, while we instead ponder how to put theory into practice.
Start behind the scenes
Like many other technology trends, the most immediately compelling applications of A.I. are the ones that promise to reshape the retail experience for consumers. But while there may be clear ROI potential in the likes of personalised in-store advertising, perhaps the most practical place for many businesses to begin is with more mundane use cases that nevertheless offer similar, if not greater, monetary value.
“The opportunity to make sense of unstructured data and build use cases from it is tremendous, but you need to understand which of those use cases is going to move the needle in your particular business,” said Steve Laughlin of IBM. “The customer-facing side of A.I. is the shiny object with the most obvious appeal, but there’s a lot going on in other parts of our clients businesses. They can use A.I. to track attributes like local events, local weather forecasts, and the actions of local competitors to accurately produce demand for hyper-local markets. Then, if they know that the retail performance of a particular SKU is highly affected by temperature and precipitation, they can better plan their inventory allocation for different stores based on accurate forecasting. And for businesses that have a lot of inventory tied up in supply chains and distribution, that can represent a significant financial improvement.”
Address the industry you have, not the industry you want
An important thing to remember with any emerging technology is to remain anchored in realism, rather than getting too carried away by less concrete possibilities. As thrilling as the idea of A.I.-led design no doubt is, at the brass tacks level shoppers still pay retailers’ salaries, and the solutions the latter implement should therefore be targeted at improving the value they can obtain from that relationship.
“Today’s shopper is price-sensitive and technologically savvy,” said Cheryl Sullivan of Revionics. “It’s an entirely different generation, accustomed to instant gratification, and a recent PWC survey revealed that 50% of shoppers will even buy outside their country if they can obtain a better price. In that context, it’s a huge challenge for a retailer to know what price to begin with for a given product, or to properly plan when to take an initial markdown, how significant to make it, and how many more should follow. We use machine learning to make optimal price recommendations that give retailers a good chance of reclaiming that 50% and more, but what it comes down to is them accepting that retail has been redefined and is now about art and science, rather than just art. I believe the retailers who will remain relevant in the face of Amazon and other disruptors will be the ones who embrace the potential of machine learning and use it to their advantage in the current retail market.”
Supplement your workforce, but keep them safe
One of the most prominent reasons that businesses turn to technology is to avoid increases in overhead and headcounts as consumer demand and market pressures place greater stress on processes and individuals. But A.I. is a different beast: we know it hits the headlines when it makes factory workers and financial analysts alike redundant, and in theory it has the potential to do the same for an alarming number of job functions in the not-too-distant future.
Before we tackle that issue head-on, though, let’s look at an instance where A.I. is being used to add expert-level capacity to an area of business where resource constraints can have a direct and material impact on international success and profitability.
“In most teams, the work of merchandising is done by a handful of people for a global audience,” said Andy Narayanan of Sentient. “We all know, though, that people in New York don’t dress the same way as people in Singapore, but there’s an element of limitations on human decision-making there. Without increasing our headcount, we, as brands and retailers, have had to accept that we simply cannot make localised merchandising decisions for all our markets at once. With A.I., though, we’re eliminating that bottleneck and letting the A.I. make decisions, engage with that shopper in New York, and understand what they want. Here, A.I. isn’t replacing the job of merchandisers, but it is decentralising decision making in the moment – the same way a local store manager or business associate would. I think that’s the huge potential of A.I. in the future – the idea that we can do this at scale for merchandising, assortment planning, product recommendations, returns management and son on. I believe that every one of these common business challenges can become an A.I. problem with an A.I. solution.”
As a senior figure within an A.I. business, Narayanan is bound to be optimistic. The reverse side, however, of solving so many business problems with A.I. is the possibility that human problem-solvers in these areas may become obsolete.
We have already discussed the sudden possibility for white collar workers and assembly line staff alike to be replaced by A.I. and robotics, but how far is this egalitarian approach to automation present in the RFA industry?
While the morbidly curious can investigate for themselves by visiting the appropriately-named www.willrobotstakemyjob.com (which pulls its data from a University of Oxford research publication written in 2013), we can say with some confidence that key creative roles are currently considered extremely safe occupations. Indeed, fabric and apparel patternmakers are predicted to be more in demand than ever by 2024, while roles in apparel and footwear manufacturing threaten to go the way of electronics assembly work conducted by Foxconn, the world’s largest consumer electronics contractor, which plans to replace 60,000 of its workers with automation and A.I. within the next three years.
For people who work on non-routine, cognitive tasks (which is to say ones that engage our brains and creative faculties), however, I can only see A.I. as good news for themselves and their businesses. Because far from replacing them, A.I. solutions will instead work to take weight off their shoulders, and provide them with the best possible starting point for more creative work.
When it comes to repetitive work, humans – fallible and expensive – are easily replaced by flawless A.I.s that demand nothing more than electricity and maintenance. For what we’ll refer to as “higher order” tasks, though, the evidence seems to suggest that we will retain our edge for the foreseeable future.
Until the time (which may remain distant until the emphasis of commercial research and development shifts to a more general approach) that creative tasks like design fall under the umbrella of A.I., then, the best applications will be those that support human beings by taking on the time sapping, innovation-light portions of their jobs, and allowing them to focus on more creative work. This, in turn, will allow businesses to do more – and be more creative, on balance – without dramatically expanding their creative teams.
A strong example of this approach in action comes from a Harvard Business Review feature titled “How Companies Are Already Using A.I.” Here, the author sets out the case of the centuries old Associated Press, whose team of just 65 business reporters were struggling to provide coverage of even a small percentage of the quarterly earnings stories their audience demanded. In 2013, the company contracted an A.I. provider, and began training the program to be able to write short earnings stories from data provided by financial research systems, without any human intervention. Two years later, of around 5,300 public companies in the USA, the Associated Press’s A.I. was covering the earnings of 3,700 on a quarterly basis. As a result, the company’s existing reporters were able to focus on producing detailed coverage of more complex stories, and not a single person lost their job.
This is a perfect case study of A.I. in use, because the program was trained by people who understood the business at hand, before being set loose to work on repetitive, low-value tasks that would otherwise have occupied huge amounts of human time. As a direct result, that same pool of human time is now spent on generating unique value for the business and its customers. And while the parallels between the worlds of publishing and apparel are limited, the same principles already do apply to cases where A.I. has been used in a similar way in our industry.
“I’ve seen this happen first-hand,” Raj De Datta from BloomReach told me. “Our insights product is used by visual merchants to lay out their product assortments on their websites, and before machine learning they would have to dig through sales data and Excel sheets to figure out where, for instance, a particular evening gown should appear on a page. That involved a considerable amount of work. Now, they work from actionable intelligence that tells them what kind of evening gowns are trending right now, and it’s then up to the merchandiser to decide whether to promote something that’s already selling well, or to take action with something new that may capitalise on a trend. These are the sorts of decisions that only humans can make, so machine learning and A.I. here are shifting the time humans spend away from basic work – which computers can accomplish far more quickly – and towards higher order tasks that actually harness their creativity.”
This drive for process transformation and efficiency in time-consuming manual processes will be familiar to readers of previous WhichPLM publications, where, even when they are tempted by exciting innovations like 3D working, most customers want PLM to get better at removing repetitive manual work like data re-entry.
Of course, actually designing or merchandising products is not the only kind of creativity occurring within a typical apparel business. The industry is waking up to the importance of connecting marketing departments to PLM – sharing common asset libraries and empowering advertising teams to better articulate the value of a product – and A.I. has already begun to support heightened creativity in this area as well.
“Working with A.I., my team is smarter, more efficient, and they’re able to spend more time doing what they love doing,” said Courtney Connell of Cosabella, whose A.I. marketing initiative delivered more than just improved KPIs. “When I hire someone out of college, I’m not burying them in busywork the way I would have done before. Instead, I can use their brains to their full capacity. I think everyone’s terrified about A.I. coming in and taking their jobs, but for me it’s a complete industrial revolution, so it’s obviously going to free up people’s time and challenge them to find better ways to use it, but I can only see that as a positive thing for society. I think that, as humans, we’ve become very machine-like over the past decade or so. There have always been repetitive tasks, sure, but we just were not meant to sit in front of a computer and do those tasks all day. That should be machine work.”
Even with this relationship whereby A.I. solutions support humans in their creative endeavours, it seems inevitable that some roles will still either disappear or become impossible for their current occupants to hold as they change. But this should not mean the overall workforce will shrink; rather the onus will be on governments and corporations to ensure that enough people are trained and hired to fit into a greater range and diversity of new higher-order careers.
“As many other technologies have done in the past, A.I. is going to take over jobs,” said Alexander Grey of Infosys. “But it’s equally going to create new ones – some that we can’t even conceive of today. The only real risk with A.I. is that it might progress more quickly than we’re able to diversify the workforce. As a comparison, robots came into car manufacturing and took over, but a lot of other job roles – car design, car sales, and industrial robotics – came into play. And although robotic automation is not pretty ubiquitous in a lot of industries, we now have more people on earth than ever before, and more of them are working than ever before. I’m not worried about A.I. itself, rather that our industries and governments may not be preparing fast enough to fill the new jobs that it will create.”
Track down talent
Assuming they do exist in your local market, you may very soon find yourself needing to hire precisely the kind of talent that Grey talks about. And while I remain convinced that hiring data scientists en masse is not the right direction for most brands and retailers, tracking down the right talent may be a harder-fought battle than many businesses realise.
In essence, what it means to work in fashion is changing. A decade ago, a design, marketing, patternmaking, or similar course would be the logical on-ramp to a career in RFA; today it is an equally viable path to study advanced mathematics and to then specialise after graduation.
“If we take a step back and think about what we and other businesses are doing in this space, we’re using a new class of technology to reinvent the entire value chain in retail,” concluded Raj De Datta from BloomReach. “For us the specific applications are focused on the digital experience; for others it might be supply chain optimisation, store layouts, or trend prediction. Wherever machine learning and A.I. technologies are applicable, there are just so many ways for business processes to be completely transformed. At the moment there’s a very clear shortage of technical resources who have fashion industry experience – and in fact specialised data scientists are sought-after in every industry. The net effect is that, if you’re coming out a good university with a PHD in machine learning, you may have a guaranteed job in retail, but you also have a guaranteed job at the best tech companies in Silicon Valley, and the best hedge funds in New York City. That makes machine learning a very competitive labour market, and retailers are going to need to partner as much as will need to recruit their own talent.”
From time saved to total transformation
Fittingly, for a subject that tackles what it means to be creative and to be human, there are no easy answers when it comes to A.I. There are, however, two cardinal rules that I feel prospective customers should attempt to abide by.
First, biding your time to make an investment in A.I. is perfectly valid, but businesses should take care not to underestimate the amount of progress that has already been made in this area. Anyone waiting for advanced general A.I. before jumping in is likely to be disappointed, either because AGI is eventually proven to be impossible, or because the value of a broad intelligence will be considerably less than the specialised systems we already have today.
Second, while it is possible to look at A.I. as a way of pursuing automation, and increasing head counts across the board, ours is a creative industry where A.I. is currently poorly-equipped to take over human jobs. Today, a far better application is to use all the tools under the A.I. umbrella to automate just those mundane tasks that are sapping time from your business, and to retain all or most of your current creative team, giving them a new opportunity to push their creativity further without compromise.
At the time we wrote this publication, these repetitive tasks are in market, retail, and inventory allocation, but it is unlikely that these will remain the only stages of the lifecycle where A.I. will make itself useful.
“I want to make a strong statement here,” continue Alexander Grey of Infosys. “I believe it’s just a matter of time until machine learning, A.I., and automation transform at least half of what every business does. They have applications in sales, marketing, merchandising, product recommendations, loyalty programs, fraud detection, human resources – the list goes on. You need only look at the finance industry, where A.I. is making the future look extremely uncertain for financial advisors, and consider how comprehensively Amazon has transformed the retail industry using technology. Companies that are not using machine learning in any way will find themselves left in the dust shockingly soon. I firmly believe that’s true.”
And let us not forget that while the RFA industry is forging its own path for A.I., the wider world is likely to be even more receptive to the technology and its applications. And some – again, myself included – see A.I. as an engine for change and social re-engineering the likes of which our lifetimes have never seen.
“As far as I’m concerned, A.I. could trigger the next renaissance in human evolution,” concludes Courtney Connell of Cosabella. “It’s all too easy to just keep ourselves in that repetitive loop, where we’re just doing the same things we’ve always done, and there’s so much work that it’s hard to stop and think about new approaches. As we use A.I. more and more, though, I see us valuing the human mind much more – valuing our own creativity, and valuing the time a human spends making something. I think that could be a really interesting shift in society because we don’t know how far our minds can go. And I think that as long as we keep holding onto menial work because it’s the only work we know, we’ll never find out.”
In the here and now, however, A.I. remains a tool. A tool with tremendous untapped potential, but, from a commercial perspective, a tool nonetheless. Its near-term applications will be supporting humans to not just keep their heads above water in the intelligence era, but also to make information decisions about the direction they want to swim in, and then to take more confident strokes towards their objectives.
“Our aim with A.I. isn’t to automate jobs or dictate creativity – it’s simply to give retail professionals the information they need to do their jobs to the best of their ability,” adds Julia Fowler of EDITED. “We’re not using A.I. to paint masterpieces, getting rid of painters in the process; we’re giving painters better paints and brushes to help them create their own masterpieces.”