As many in the enterprise IT community will remember, technology suppliers succeeded in roundly confusing buyers in the early part of the millennium by “greenwashing” their products and services – or in other words, exaggerating the true extent of their environmentally-friendly credentials – thereby shooting themselves in the foot and, arguably, putting the brakes on the market.
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
But it seems that many have learned little from the experience. According to Gartner, the IT industry is now pursuing an equally self-destructive strategy of “AI [artificial intelligence] washing” – by applying the AI label too indiscriminately, suppliers are once again bamboozling potential customers, who are putting off making buying decisions as a result.
So just how true is this contention and, if it is valid, what impact is it having on the market to date? Nick Patience, research vice-president at 451 Research, believes that AI in the enterprise software space is certainly overhyped, and adoption has lagged behind uptake in the consumer market.
“A lot of startups are claiming to do AI when they’re using rules-based automation,” he says. “Suppliers also say they have AI systems, but it’s actually much more narrowly defined machine learning software that does image recognition or lead scoring. There’s nothing wrong with that, but it’s never going to be a robot that can do many of the things humans can do, so you have to cut through the hype to know what you’re getting.”
Emma Kendrew, AI lead for Accenture Technology, agrees that the hype cycle is reaching a peak, driven by busy corporate marketing machines hoping to take advantage of the possibilities opened up by big data and the cloud, as well as burgeoning customer interest.
“AI is often associated with futuristic sci-fi films and so has an air of deep complexity. It also has an air of mystery, and even of being sinister,” she says. “But there’s a sense of urgency driven by marketing and the media too.”
The upshot of this situation is that, while many organisations – and in many instances, their senior business executives – are aware of the technology and feel a pressing need to engage with it, they simply have no idea where to begin.
Pilots and proof of concepts
“There’s a lack of confidence in where to invest and where to start,” says Kendrew. “So there’s a lot of awareness of AI, but, in many ways, that’s driving the confusion and uncertainty as to where to begin.”
As a result, most market activity is currently focused on pilots and proof-of-concept work, with predominantly large corporates in sectors such as financial services and customer-facing industries such as retail starting to dip their respective toes in the water as they experiment to understand the technology’s potential.
For instance, financial services firms are using AI to help with everything from fraud analysis to processing customer account applications, while customer service agents are being employed by business-to-consumer companies of all stripes to understand customer needs more effectively and personalise interactions.
Emma Kendrew, Accenture Technology
“We’re starting to see adoption in areas where there are large volumes of data – especially if it’s unstructured, so text, audio and video – and where people have to analyse it manually,” says 451 Research’s Patience. “So a lot of adoption now is about taking the low-hanging fruit of tedious processes where there’s no value in having a human doing it. Helpdesk ticketing automation, for example, is quite a popular early use case.”
But he agrees with Gartner that a key inhibitor to the sector’s further expansion at the moment is simply the lack of AI skills in most organisations due to its nascent stage of development. Although not an unusual state of affairs with emerging technology, the situation is particularly acute here due to the fast pace of change – although Patience acknowledges that things will undoubtedly sort themselves out over time.
Another issue that is holding companies back in adoption terms, meanwhile, is the belief that they do not have the large datasets required to make AI work. “Where I see a lot of CIOs expressing scepticism is in relation to data volumes,” says Patience. “Large suppliers are buying in sources of data to build and train models to make them more effective, but mid-sized organisations don’t feel they have enough.”
But again he expects this scenario to change as the internet of things (IoT) starts to take off over the next two to three years. “The growth of AI will be tightly linked to the growth of IoT, and also social media-based data, as people express their views about their preferences in ways they couldn’t before. It’s a big trove of data that can be acted upon by everyone, and it’ll make a huge difference,” Patience concludes.
Case study: FreestyleXtreme
“We didn’t specifically decide to introduce AI,” says Shaun Loughlin, managing director of online action sports retailer FreestyleXtreme. “It was more of an extension to our existing automation strategy – we don’t differentiate between the two as they go hand-in-hand, so we see AI as simply the next stage in the automation battle.”
The company, which was set up in 2003, first began its automation journey six years later, but introduced machine learning-based marketing applications from Emarsys in 2015.
The cloud-based system collects information about each individual customer based on their demographic profile, purchase history, website browsing activity and the like, and personalises email marketing messages accordingly in real time. It also ensures that the company’s website homepage carries the goods that customers are most likely to be looking for when they visit it.
“The product recommendation feature has directly increased revenues by 8% because suggestions are more relevant,” says Loughlin. “But the system has also allowed us to operate more efficiently and consistently with a much smaller team than we’d need otherwise.”
Although the firm has only 70 staff, it runs 20 localised storefronts and ships products to more than 60 countries around the world. “Without automation and AI, it just wouldn’t be possible,” Loughlin points out. “In marketing, for example, we’d probably need between 200%-300% more staff to do what we do now.”
Pros and cons
The fact that email marketing campaigns can currently be conducted in two hours rather than the former two days means the team is able to focus on more creative and strategic activities rather than become bogged down in basic administrative tasks.
“We’re growing massively, but we’re not letting people go,” says Loughlin. “It just means that we don’t have to recruit people at the same level, or we can recruit in other areas that are most beneficial to the business.”
Shaun Loughlin, FreestyleXtreme
But he acknowledges that the concept of AI has been subject to a lot of hype that is putting many people off, particularly in the small to medium-sized enterprise (SME) space.
“Without realising it, e-commerce firms have been using AI for years with things like text ads in Google, but they’ve not called it that,” he says. “So the increased popularity of the term is causing confusion, especially among SMEs. It’s scary as it sounds so unobtainable.”
Other common barriers to adoption include cost and fact that the majority of AI systems are far from plug-and-play, requiring time and effort to implement.
But given that Loughlin’s aim is to expand the business into Japan and Eastern Europe over the next year or so, he believes increased levels of automation are an important way forward.
“We’re going to continue to focus on automating as many elements of the business as we can as a key way to boost efficiency. AI and automation go hand-in-hand, so it has a part to play in our continued development. Expansion was always part of our business plan – it’s just that AI makes it easier,” he says.
Case study: Hertfordshire Partnership NHS Foundation Trust
The Hertfordshire Partnership NHS Foundation Trust has just started a 12-month pilot project to establish whether a so-called CoachBot AI-based system can help clinical team members work together more effectively.
The trial will consist of four teams with about 65 members in total, and is intended to make it easier for participants to benefit from coaching interventions without needing to dedicate unrealistic amounts of time to the process.
“Recruitment and retention is a big challenge, especially given some of the NHS’s operational pressures, so we’d been talking about what we could do to support the workforce,” says Stacie Coburn, principal advisor at the Eastern Academic Health Science Network (EAHSN) – one of 15 similar organisations set up around the country to promote innovation. “The problem is that getting a full team to spend a day on face-to-face coaching simply doesn’t work in a world where it’s difficult to release staff from frontline duties.”
Stacie Coburn, Eastern Academic Health Science Network
The advantage of Saberr’s CoachBot system, on the other hand, is that it handles the time-consuming fact-finding portion of the process to help set an agenda and assist team members in understanding what issues they should get together to tackle.
The online app can be accessed at a time convenient to each individual team member, but comprises a series of questions that last about 20 minutes in total. These consist of a four-minute set of on-boarding questions to gain a contextual understanding of the team, followed by around 10 minutes of diagnostic questions to get a handle on the challenges it is facing.
The advantage of algorithms
Algorithms are then used to sort through the data and make suggestions on areas that require work, as well as providing participants with access to learning toolkits.
But Coburn points out that the system is not necessarily being seen as a “complete internal replacement for people-led learning”. As a result, one of the four teams will be provided with a coach on a quarterly basis to assess whether additional human intervention is of benefit or not.
In terms of measuring success, there will be quarterly checks on whether teams are meeting self-defined targets and whether any change has taken place in their Net Promoter Scores – the willingness of participants to recommend working in their team. Team productivity will likewise be measured, as will any impact on quality from a healthcare perspective, which includes improvements in patient outcomes.
But Coburn believes it is neither here nor there that the system EAHSN has bought is AI-based or not. “It wasn’t a factor in the purchasing decision,” she says. “ChatBot as a product looked like it could solve our need, but I didn’t know it was AI.”
Instead, what did appeal was the fact that the offering could learn on an ongoing basis. “Teams are always going to have challenges to work through and there’ll always be some element of improvement they could be working on. So there’s a huge value in having a product that can continue to learn,” she says.