Customer Management

Our monthly roundup of insight, analysis and new thinking on customer management and customer relationships

How the march of the robots will change customer management forever. “I’ve never seen a technology with so much potential.”

AIChatbots, preparing your business for automation, design in the age of Artificial Intelligence, RPA, machine learning, how customers respond to AI conversations, Augmented Agents, AI ‘celebrities', the world’s top ten robots and more

Intelligence Special

SECTOR DIGEST

Customer Management

Pronto or precise? What matters most to customers, fast delivery or precise delivery? Argos is following the current industry mood that it’s accuracy of delivery time that is the greater driver of customer satisfaction. They’ve introduced a tracking tool on all two-man home delivery vans so contact centres can give customers accurate information on exactly where their deliveries are, and how far away.

Logistics

Online Shopping

Around three quarters of online e-tailers are looking to build VR into their customer experience by 2020, reveals a report by Oracle into how virtual reality will come to replace…well… real reality. Online shoppers are growing less loyal, and increasingly looking for an interesting shopping experience to retain their attention. E-tailers are banking on VR and AR techniques to help them do so.

Can VR make customers stick?

Cathisophobia – the fear of sitting down. Sales staff standing around chatting, especially near the showroom entrances (because customers love to be judged by a gaggle of sales people as they walk through the door).

Galeophobia – the fear of sharks. They may power through a lot of deals but sharks’ attitudes to customers can be culturally bad for a dealership in the long run and contribute to high staff turnover.

Technophobia – well, you know this one. A reluctance to use new CRM systems because ‘the old paper way just works better’ may be hiding a lack of skills, and holding the dealership back.

Chronophobia – the fear of passing time. Sales staff complain of there never being enough time, yet spend hours in sales meeting after sales meeting (largely to put off actually doing anything).

Hypengyophobia – the fear of taking responsibility. Sales teams wary of adopting new systems or trying fresh techniques for fear of being held responsible if it goes wrong.

The showroom of fearFive only slightly tongue-in-cheek telltale signs of a sick sales floor, from Auto Retail’s Brent Wees.

Automotive

And you thought Millennials were toughBrands are struggling to meet the demands of Generation Z (post-Millennials). An IBM study of 15,000 Generation Z shoppers (with $44bn buying power behind them) found that despite living digital lives, 67% prefer to shop in brick and mortar stores. However they bring their digitally-informed demand for highly personalised interactions with them.

Which is a problem because IBM’s polling of 500 brands across 24 countries revealed that…

� Only 19% can provide a highly personalised service � Only 17% can deliver anything more than in stock/

out of stock information � 84% have nothing in the way of in store mobile

services

In other words, the gap between what Generation Z expects, and what retail is delivering, appears considerable.

Personalisation

And in other survey news, telephone answering service Callcare set out to find people’s biggest gripe when dealing with customer service teams and discovered it was…cue shocked face… being kept on hold! Who knew? Perhaps more interestingly, advisors who were over familiar were more complained about than advisors who showed no personal interaction.

100%9:41 AM

How many weeks us Brits will spend calling customer support across the course of our lives. Research among 2,000 consumers by customer experience business Genesys calculated we spend a surprising average of 17 minutes a week on the phone to contact centres.

When it came to holding time, government departments and agencies were easily the worst (above 15 minutes), telecoms, TV and local councils came in around 12, utilities at 11 and retailers at six.

5.5

X? Y? Z?

The promise of B2B What should telcos do to survive in an environment where digital dominates? According to McKinsey’s latest look at the industry, creating ‘super slim’ and efficient core businesses is key. One new business opportunity they could capture lies in the B2B space as this is just beginning to develop. Leveraging their infrastructure advantage and combining it with state-of-the-art digital technology would position them as the backbone of fast-growing digital ecosystems, especially around IoT, security and Industry 4.0. However, competition will be intense, warns McKinsey, as players from other sectors will also be chasing these fresh revenue opportunities.

Telcos

Utilities

Don’t get beaten to personalisationWhen the UK water market opens up to more competition, new brands will quickly use personalisation to target customers with the largest bills and lowest risk profiles. That will leave behind-the-curve incumbents with a higher concentration of vulnerable customers and those with small bill values and bad payment histories. So warned Utility Week’s Tom Webb, analysing the dangers water companies face by sitting on their hands over big data analysis.

Customer management

How much bad customer service cost UK companies in 2016 Research from Ombudsman Services (providers of independent dispute resolution) shows that 28% of people report losing loyalty for a brand after bad service but when a complaint is handled badly that leaps to nearly 80% vowing not to shop with that brand again. With a few quick sums, the researchers equated that to a £37bn loss in sales.

£37bn

What does a KFC face look like?Beijing may be about to find out thanks to a pilot programme that effectively lets customers order by expression. Facial recognition tech installed at the Chinese capital’s KFC Original store will recommend menu items based on a customer’s estimated age and mood, and use it to store their favourite orders. Do you want fries on that?

Technology

Happy New Year?? In a busy start to 2017 for EE, January saw it become the first UK mobile operator to bring all customer service calls to the UK… and also net a £2.7m fine from Ofcom for overcharging nearly 40,000 customers a total of around £250,000. This was mostly through applying US rates for contacting the 150 customer service number rather than EU ones - £1.20 per minute instead of 19p. Meanwhile BT collected The Daily Mail’s

Money Mail Wooden Spoon award for the worst customer service – second year running. Perhaps unsurprisingly the Mail was also irked by BT Chief Executive Gavin Patterson turning down their offer to collect it in favour of attending some other event… the World Economic Forum at Davos or something.

Telcos

Technological inhibitionsAlmost half of UK retailers are not reaping the benefits of new technology, because they’re struggling to implement it. Their ‘tech stack’ is too high. As a result, hoped for improvements in website stability, customer experience and analytics are not coming through, often because project cycles to make even simple changes are taking too long. (Source: Coleman Parkes Research for eCommera.)

Retail

Marketing

Perks WorkOne in five consumers choose everyday service providers – eg mobile, utilities – because of the benefits and perks on offer. So say customer journey specialists Echo Managed Services. Their survey of 1000 consumers said a quarter claim to be significantly impacted by perks, and a fifth confess to remaining loyal to a brand largely because they didn’t want to lose them.

Quick fixes, however dynamic, often just create new habits that don’t address the underlying problems

What to do? Recognise habituation exists in your business and resist the instinct to jump into fix mode until you’ve explored the problem exhaustively

These remain, perhaps never properly understood, while the new habits render invisible any further issues

Why it never quite worksIf plans to improve your organisation’s customer experience never seem to deliver quite what you’d hoped, it could be down to ‘habituation’, says Forbes customer service contributor Adrian Swinscoe. Why?

Customer experience

Customer services lead UK outsourcing marketThey accounted for 17% of UK outsourcing spend in 2016, a total value of £1.04 billion, a dramatic increase of 132% on 2015’s £449 million. The majority of last year’s deals were signed around multi-channel delivery, while the percentage going offshore fell from six to four percent. IT application and network management were the next biggest spends, at £906m and £503m. (Source: Arvato UK Outsourcing Index).

Outsourcing

AI SPECIALThis month’s Intelligence dedicates itself to the subject of Artificial

Intelligence and what it means to customer management.

Read on for what chatbots can and can’t do, the questions any organisation should ask itself when exploring automation, the world’s top ten robots, and how AI and Agent can work together.

Host: Mike Barnard – Executive Director – Capita Customer ManagementMike Havard - Director - EmberChris McKibbin – Future Solutions and Innovations Lead - CapitaSimon Hunt – Director of Products and Solutions – Capita Customer ManagementAnish Shah – Director of Automation and AI - CapitaCatherine Howe – Director of Digital Transformation – CapitaDavid Naylor – Head of Analytics – Ember

Look out for coverage of Capita Innovation breakfast event. See what happens when a panel of Capita experts meets a room full of clients to discuss the future of robotics, automation and the impact they’ll have on the UK’s businesses and customers.

Coming Soon

I robot. You worker.

Where will automation take the world of business in the next 50 years? Are we all out of a job or do the robots still have challenges on their shiny metal hands?

Questions around automation and its impact on the work we do (and customers we serve) are hardly new. Back in the 1960s an apprehensive US President Lyndon B Johnson established a commission to explore its influence on jobs and the economy.

Even the word ‘robot’ was originally coined in a 1920s play about androids doing the factory work of humans at a fraction of the cost (taken from the Slavonic word ‘robota’ meaning drudgery and servitude, if you were wondering).

However the difference between automation’s production-line focus back then and its relevance today is founded on breadth. There is barely a role existing in modern day employment that could not be automated in some way, so says the McKinsey Global Institute in its new report A Future That Works: Automation, Employment and Productivity.

It will reach “from miners and landscape gardeners to commercial bankers, fashion designers, welders - and CEOs.” (We’re betting that was written with a smile.)

The whole job, or just a small part?The question is how quickly those technologies will become a reality, and where the impact will be felt first and most significantly.

McKinsey has spent the last two years studying that issue, though not, as has been done before, by looking at whole occupations that might be replaced. Instead it examined a raft of individual activities carried out by workers everywhere.

The institute analysed 18 different performance capabilities (you can see the list in a box nearby) and where they sat in the level of difficult or easy for a human or machine to do. Armed with those categories, it then set about exploring the automation potential of more than 2,000 work activities across 800 occupations, basing it on data from the US Department of Labor.

And the results? At a global level, around 1.2billion employees would be affected by the technically automatable activities McKinsey predicts, and those workers represent a wage bill of $14.6 trillion. More than half of that number comes from just four countries: China, India, Japan and the US. (China and India’s large populations mean that those two alone represent 700 million full-time equivalent jobs between them.)

In Europe the figures are 62 million FTEs and around $1.9 trillion in wages.

Which roles will be most affected?Generally it’s the lower paid jobs with the lowest skill requirements that offer up the biggest potential for automation, but McKinsey is quick to warn that within those averages lay dramatic variations.

As mentioned earlier, even CEOs could find robots seeking the keys to the executive washrooms…”we estimate about 25 percent of their work could potentially be automated, primarily such tasks as analysing reports and data to inform decisions, reviewing status reports, preparing staff assignments, and so on.”

Automation could create a ‘barbell’ economy where mid-sized organisations lose out.

Within sectors, too, there is considerable divergence. Manufacturing’s reliance on the physical activities of welders, cutters, and solderers offers an automation potential above 90 percent, whereas in customer service that comes down to 30 percent.

In short McKinsey’s conclusion was that right now only a small percentage of occupations could be completely replaced using current technology. However some element of nearly every job could be automated, leading the firm to conclude that humans working alongside robots is a more likely near-future outcome than entire jobs disappearing to gilded C3POs and their like.

“Overall, we estimate that 50 percent of the activities that people are paid to do in the global economy have the potential to be automated by adapting currently demonstrated technology. While less than 5 percent of occupations can be fully automated, about 60 percent have at least 30 percent of activities that can technically be automated.”

But, with relevance to customer service, McKinsey continued thus: “While certain categories of activity, such as processing or collecting data, have a high technical potential for automation, the susceptibility is significantly lower for other activities including interfacing with stakeholders, applying expertise to decision making, planning, and creative tasks, or managing and developing people.”

When will it happen?Going back to McKinsey’s 50 percent figure for automation potential, at the current rate we could be seeing that in 2055. Or 2075. Or 2035. There are arguable scenarios for all three, says the company. It’s that hard to predict.

If that seems a huge shift in not so many decades it’s not without precedent. In the US, farm employment fell from 40% of all jobs to just 2% of all jobs in 100 years. Manufacturing employment dropped from 25% to 10% in 60.

“In both cases, new activities and jobs were created that offset those that disappeared.” And importantly, as has also often been said about digitisation, “it was not possible to predict what those new activities and jobs would be while these shifts were occurring.”

At a global level, around 1.2billion employees would be affected by technically

automatable activities, representing a wage bill of

$14.6 trillion.

Businesses that could benefitSpeaking of digitisation, there could also be echoes in how automation disrupts classical business models, considers McKinsey. Large corporations have the capital to scale up, while smaller players and even individuals could begin to take on work previously the domain of larger businesses. “The growth of very small and very large companies could create a barbell-shaped economy, in which mid-sized companies lose out. In all sectors, automation could heighten competition, enabling firms to enter new areas outside their previous core businesses, and creating a growing divide between technological leaders and laggards in every sector.”

Finally the knotty question of putting people out of jobs…Overstated, thinks McKinsey and founded on a lack of understanding of the realities of population growth and job demand.

“While much of the current debate about automation has focused on the potential for mass unemployment, predicated on a surplus of human labour, the world’s economy will actually need every erg of human labour working, in addition to the robots, to overcome demographic aging trends in both developed and developing economies.

“In other words, a surplus of human labour is much less likely to occur than a deficit of human labour, unless automation is deployed widely. However, the nature of work will change. As processes are transformed by the automation of individual activities, people will perform activities that are complementary to the work that machines do (and vice versa). These shifts will change the organisation of companies, the structure and bases of competition of industries, and business models.”

Barriers to adoptionMcKinsey identified five factors that could hinder or encourage our robot colonisation.

Technical feasibilityPresently machines can match or outperform humans in some of the 18 capabilities included in McKinsey’s framework (mostly involving data retrieval, planning and gross motor skills). However that leaves a lot of technology waiting to be invented to handle the rest. “In particular, advancements in natural language understanding could unlock significantly more technical automation potential. Emotional and social reasoning capabilities will also need to become more sophisticated for many work activities.”

Cost of developing and deploying solutionsThe investment required to develop and deliver automation technologies cannot be underestimated. Even ‘virtual’ software based solutions, which tend to be adopted earlier because of their comparatively lower cost, still require real investments in engineering to create solutions.

Labour market dynamicsJust because a job involves skills that are easily automated, doesn’t mean it will be economic to do so. McKinsey gives the example of a restaurant cook where 75% of the skills could be replaced by technologies that exist right now. However, in the US, cooks works for $11 an hour and there’s a long queue of them ready to eggs-over-easy at that salary.

Automation will also mutate the labour market itself. “For example, if middle-income workers such as clerks and factory workers are displaced by the automation of data collection and processing and predictable physical activities, they could find themselves moving into lower paid occupations, increasing supply, and potentially putting downward pressure on wages.”

Economic benefitsAs well as labour cost savings, a business case for automation might also include performance gains, increased productivity, improved safety, higher margins and more. In some instances such advantages can actually dwarf the savings on salaries alone, say in an industry such as oil and gas production where not only can accidents be dangerous and expensive, but a high proportion are down to driver error.

Regulatory and social acceptanceEven if deploying automation makes business sense, adoption can be a long process, waiting on shifts in Government policy, supply chain reconfiguration or retraining of workers, even when they’re ready to embrace the change. And the emotional resistance to seeing automation within intimate or even life and death scenarios – from hospitals to self-driving cars – can be considerable.

Three examples…

Given by McKinsey of how automation is generating real value right now

Rio Tinto – automated haul trucks and drilling machines at its Pilbara mine in Australia have lifted utilisation and productivity by 10 to 20%.

Google – the machine learning of its DeepMind AI at data centres has reduced energy consumption by 40%

Financial services – ‘straight-through processing’, where transaction workflows are digitised end-to-end, has boosted throughput by 80% in some instances.

Capability requirements The 18 categories around which McKinsey based its research

Sensory perception � Sensory perception

Cognitive capabilities � Retrieving information � Recognising known patterns and

categories � Generating novel patterns and

categories � Logical reasoning and problem

solving � Optimisation and planning � Creativity � Articulating or displaying output � Coordination with multiple agents

Natural language processing � Natural language generation � Natural language understanding

Social and emotional capabilities � Social and emotional sensing � Social and emotional reasoning � Emotional and social output

Physical capabilities � Fine motor skills and dexterity � Gross motor skills � Navigation � Mobility

The Uncanny Valley

What is it and why does AI need to cross it to be fully embraced by customers?

The Uncanny Valley. So where’s that? Disneyland?No, it’s a dip in the graph of AI reality – or rather how humans interact with AI.

What does it measure?It’s the spot where our minds can’t quite tell if what we’re looking at or receiving responses from is real or not. Even though what we’re experiencing is so nearly genuine, there’s just enough of a difference for part of our intellect to sense something is wrong.

And what’s the reaction?A surprising one. Revulsion. The graph of our relationship to the ‘almost human’ takes a massive drop in comfort at this point. It’s why a robot with a very lifelike face often seems creepier and harder to relate to than one that’s clearly rubber over motors. It’s the ‘very nearly’ that truly spooks us out, more so than the plainly false.

So why is that a problem?Because it would appear that the closer we get to a human-like AI response, the more we dislike it. We want something that is either totally indistinguishable from the real or something that lets us make an instant judgment of ‘robot’. It’s the same problem currently facing Hollywood and CGI ‘actors’. The Star Wars movie Rogue One was a good example. The CGI recreation of Peter Cushing was an amazing technical achievement, but just enough off perfect for a section of the audience to feel creeped out (and vocally enough to begin a debate around studio boardrooms on whether this was really such a good idea).

What does that mean to AI developers?Unfortunately a lot of AI development currently sits plumb centre in the uncanny valley – it hasn’t crossed the divide to the emphatic real. So which way should developers jump if they want people to interact comfortably with their creations right now? Dumb down, or push on? Also our ability to tell fake from real would appear to grow more sophisticated as the technology advances. For example 50 years ago audiences might have classed waxwork models as eerie residents of the Uncanny Valley, whereas today our higher developed perceptions would instantly dismiss them as fake. Neither is the Uncanny Valley new. It was first mapped by Japanese robotics engineer Masahiro Mori 40 years ago.

And how about customer facing AI?The lesson would appear to be that while customer service AI has huge potential, organisations need to be wary of its creep into ever more complex contact centre problems, just because of its offer on cost savings. Problems it’s not yet ready for. It’s in exactly these complex interactions, where there is uncertainty and ambiguity, that today’s AI will demonstrate its Uncanny flaws – and customers could end up not just frustrated, but thoroughly creeped out.

While customer service AI has huge potential, organisations need to be wary of its creep

into ever more complex contact centre problems - areas it’s not ready for, where its ‘Uncanny’

flaws will show up.

Real, CGI or spooky? A ‘teenage’ Robert Downey Jr in Avengers: Age of Ultron, a “de-aged” Michael Douglas in Ant Man and a resurrected Peter Cushing in Rogue One.

The science of conversation

Conversational Interfaces. Ok, not as catchy as ‘AI’ or ‘chatbot’, but in terms of contact centre environments it’s a more accurate description of the category these aspects of technology sit in, and the principal use to which they are put.

Essentially they’re pieces of software that can extract both relevant data and customer intent from text-based conversations. However, significant advances in speech recognition mean that the same underlying technologies and techniques could soon be applied to voice interactions as well.

These interfaces may exist to handle very specific and limited tasks, or they can take a broader stab at standing in for a human when the customer calls.

As Capita’s Lead on Future Solutions and Innovation, Chris McKibbin spends his time hunting down the technologies and ideas that can advance how we serve our clients and improve the lives of their customers.

Which makes him the ideal person to ask about conversational interfaces and where they stand today. What they’re good at, where they still need work and how they benefit the customer experience.

So then, chatbots…“Chatbots are an example of a conversational interface, and currently they’re pretty good at transactional enquiries, where a customer has a single, easily understandable enquiry,” says Chris. “For example, if I ask what my bill is, that’s pretty straightforward. But if I say can you tell me what my bill is, oh and by the way I’m thinking of upgrading, that’s a lot more confusing. Furthermore, can the bot handle a string of queries that maintains context from earlier in the conversation?”

The ability of chatbots to respond intelligently and discerningly to customer enquiries is paramount. If customers have to change the way they interact in order to get the best out of an AI solution, then that solution has plainly failed.

“Clients are rightly concerned about how their brand experience translates in an AI-powered conversational interface. They want to know how their customer experience can be best brought across and, absolutely, if the customer is perceiving something different, then it hasn’t worked.”

Regulatory and data protection wrinkles are also beginning to emerge within the chatbot universe. For example, when a customer is party to an automated decision making process, do they have the right to know that? Is there a legal imperative to announce that the advice you’re being given is from an artificial intelligence?

If that’s the case, wouldn’t that undermine the efforts of software developers labouring to make AI indistinguishable from human? The rules here are still being written and likely to develop rapidly.

How good are we at talking to robots?Right now it’s too early in the chatbot timeline to know for sure how humans want to interact with them. In some instances, reckons Chris, we may end up feeling more

comfortable with a robot than a human.“Maybe I’ll get on better with a bot because it can provide me information just like that. I don’t have to wait for the adviser to look it up, and I don’t have to worry about being polite and saying thank you. Though that does raise questions about a whole generation being taught that saying thank you isn’t required anymore…

“However, strangely, there’s some evidence to say people are more polite with bots than humans. They realise there’s no point in getting stroppy, because the bot can’t have an argument with you, so they tend to focus on what they need to get done, and just get on and do it.”

Or perhaps we just won’t be able to resist the temptation to break the bot. That’s a lesson Microsoft learned with its ill-fated Tay learning chatbot when, in the space of 24 hours, hundreds of sabotaging messages ‘taught’ it to be racist. Perhaps the first recorded case of chatbot abuse.

And should chatbots be heard or seen as well? That’s an option being explored by IPsoft’s Amelia AI, billed as the first digital employee. As well as answering questions via speech synthesis, Amelia is partnered by a visual avatar whose expressions and facial movements attempt to mimic what’s being said.

“There’s a lot of effort going into making her look as human as possible,” says Chris. “Mapping different emotions onto her image to try to give people something visual to interact with. It’s hugely interesting, it will get better and better, but it isn’t there yet. In fact it just seems…er… a bit weird at the moment...” [See our feature elsewhere in this report on The Uncanny Valley.]

When a customer is party to an

automated decision making process, do they have the right

to know that?

McKibbin

Chris

Can you teach an Artificial Intelligence to be ‘onbrand’?“There’s also a lot of interesting work being done around conversational interfaces and brand differentiation. How can you tell a John Lewis chatbot from a Debenhams chatbot? How much better and more polished would John Lewis like their chatbot to be compared to, say, a more transactional one at Currys, where all I want to do is buy a TV as quickly as possible?

“From a brand’s standpoint I’d be very keen to have that ‘John Lewisness’ come across. But how do you do that?”

As conversational AI evolves, we may find small armies of chatbot scriptwriters and linguists, with ears trained for dialogue, working behind the scenes to capture the fine shadings of a brand’s language and feed it to the software. That’s a project beyond the scope of simply writing a contact centre script. It would need to be more detailed, more engaging and be something that could continually learn and progress.

Linked to that is the work being done by the likes of Cogito, a Boston-based company seeking to deliver real-time analysis of customers’ emotional states just by listening to their voices. It goes beyond a contextual study of the words being said and examines how they’re being said, measuring stress, inflection and so on. Cogito can then prompt advisors to be more empathetic or speak more slowly, or predict if a caller might be about to get angry or is perhaps more open to a sell than they seem on the surface.

Such technologies could be an important step in the development of more intuitive conversational interfaces because, frankly, they don’t do sarcasm. An irritated caller’s comment of “yeah, thanks, you’re doing just brilliantly” would be interpreted by most present day bots as the purest of praise, when to a human ear it’s clearly a fed up customer resorting to sarcasm in the face of overwhelming frustration. And that interaction is not going to go well.

Where chatbots outgun their human counterpartsSo, problems to solve, yet there are undeniably areas where AI-driven conversational interfaces have huge advantages over human agents, one being their ability to process data. This opens up new possibilities on how much a chatbot could know about customers before it delivers an answer.

For example, if you call an advisor working for an airline, they should be able to quickly surface information about whether you’re a club member or where you last flew. But in that same space of time an AI could just as easily be interrogating your Facebook or LinkedIn pages, and scanning your last few hundred Tweets.

There’s some evidence to say people are more

polite with chatbots than humans. They

realise there’s no point in getting stroppy.

“

Have you ever expressed an interest in seeing a particular country? Are you the sort of person who complains a lot, or loves everything they see? What’s your tone of voice, serious or humorous?

Which brings us to ‘Augmented Agents’ – occasionally given the grisly epithet of ‘Centaurs’, (presumably humans with horsepower). Here the AI is doing a mass of data mining in the background, and serving up material that isn’t on the organisation’s own system to the human advisor taking the call. Like a coach whispering strategy in a footballer’s ear.

“Suddenly that interaction is not just limited to the company’s own engagement with you, it’s about the information available on you from different sources. Should that dictate or change how the organisation is talking to you?”

AIs can also do one thing at the press of a button that can take humans days.

They can scale up.

A hundred more ‘agents’ at the flick of a switch“For example, look at the work we do for William Hill on Grand National Day,” says Chris. “We might take on an extra 600 people just for the day to deal with the spike in volumes, and then they go away that same night.

“A lot of the time what they’re doing is essential but fairly basic stuff, checking ID, making sure important messages about gambling responsibly are delivered. But really a lot of that is transactional and quite limited: ‘I want to place a bet on the 2.30. Can you tell me X, Y and Z’.

The obvious advantage AIs bring is the massive difference in cost between training up 600 people to handle a task for one day, or switching on a few more banks of servers.

Clients are rightly interested in how their brand experience translates in an AI-powered

conversation. They want to know how their customer experience

can be best brought across.

“

There is more to that scenario.

“If certain messages have to be delivered and certain words always need to be said – like the warning about gambling responsibly or even a cross selling opportunity – you can programme a conversational interface to always, always say it. Whereas an agent, perhaps because they were distracted in some way by the customer, might forget.”

What we don’t yet know is whether customers will actually like the idea of always, ever and unfailingly being sold to by a robot when a human agent might read the conversation and rightly decide, perhaps not today.

It’s an area of huge and fascinating potential, but with a long way to go yet.

When AI is operating at its best, it’s working quietly in the background, both informing us and giving us control.

And finally it should remove complexity from life And certainly not add a further layer of it.

For a more detailed look at what Béhar had to say, and examples of some of the products he’s been behind, check out the Fast Company coverage:

Designing the age of artificial intelligence

What makes a successful AI product or interface? In such a new and rapidly changing field, how can we tell what good looks like? And for that matter, will customers even know?

Yves Béhar is a Swiss designer distinguished for his work where smart technology meets the ordinary person. (For example the wearables company Jawbone, where he is Chief Creative Officer)

As guest speaker at January’s A/D/O Design Festival in Brooklyn, he laid down 10 guidelines for great design in the AI era.

Well designed AI should…

Solve an important human problemToo many current AI applications verge on being gimmicks with no true and deep purpose, says Béhar. What they should be doing is finding ways of genuinely improving people’s (and customers’) lives. Béhar cites as an example the smart crib his firm has developed that helps parents with young children get more sleep. (Who would want that…)

Avoid historical clichéAI doesn’t have to look, be or act like a robot just because all our old sci-fi stories have painted it that way.

Enhance not replace the humanGood AI should be about helping us do more, not replacing us as workers (or agents).

Work for everyoneWell designed AI ought to be immediately easy to operate for all, not just the well informed techie.

Be discreet When AI is operating at its best, it’s working quietly in the background, “giving us subtle signals that allow us to be both informed and control the environments that we’re in,” says Béhar. For example, a lock might be dubbed ‘smart’ if it doesn’t need a key but opens when you wave your phone at it. But how smart is that really? Wouldn’t it be better if the lock opened automatically when it sensed you were there, and made your phone vibrate in your pocket to confirm it had happened?

Be an adaptable platformSomething that can grow and learn. Good AI products should be updateable, like software, not one-offs frozen in their current form.

Deliver services that can build long-term relationshipsCreating loyalty by constantly improving - without being told to or how to.

Learn, predict and encourage human behavioursRather than waiting for instructions from customers and consumers, good AI should be delivering prompts of its own, e.g. recommending fitness activities for the elderly instead of remaining inert until they request something. (A proposition with clear significance for customer service.)

Accelerate new ideasIncluding bringing new products to market more rapidly.

The Robot Top Ten

The closest we currently have to the sci-fi dream of a two legged, human-sized mechanical man that can do the heavy lifting for us (literally) and pick itself up if it falls over. Funded by DARPA, with the ultimate aim of helping soldiers on the battlefield, and developed by leading robotics engineers Boston Dynamics. Interestingly, the company was acquired by Google in 2013 but is now up for sale, reportedly because, remarkable as Atlas is, the road to a high volume, commercial version was seen as too long. Less friendly commentators have said that Atlas is also just too damn frightening – both in appearance and the negative publicity surrounding its potential to do humans out of a job.

Special skill - can give advice in multiple languages, with lifelike facial features Developed by Toshiba and currently delivering information to tourists on Tokyo’s waterfront. Speech synthesis is realistic, movements not so… Third generation version Chihira Kanae furthers Toshiba’s ambitions to build in speech recognition as well as speech synthesis. Prompted UK robotics scientist Professor Noel Sharkey to note an intriguing cultural divide in our attitude towards robots; Japanese audiences want them to be as lifelike as possible, while Westerners seem to prefer robots that look like robots so we know where we stand. He did also describe Chihira Kanae as having “the slight look of a psycho killer” and you sort of see his point.

Special skill – cuteness Though not as advanced as Atlas, Asimo has the benefit of not resembling a terrifying assemblage of steely rods and slicey parts. It was developed by Honda, who may have been overtaken in this field, but deserve credit for being one of the first to take it seriously. Asimo – Advanced Step in Innovative Mobility - was introduced in 2000. Back then it was dazzling enough that a friendly, four foot tall piece of self-animated white plastic could wave at you and climb steps. See it hop on one foot, see it take a penalty, see it dance, admittedly like a tiny plastic granddad.

Most of our present encounters with AI are via screens or over the phone, but what about actual robots? Who’s taking machine learning and putting it in, well, a machine? Here are ten of the most advanced robotic projects currently under development across the world.

Special skill – partnership

ATLAS

CHIHIRA JUNKO

ASIMO

There’s an intriguing cultural divide emerging in our attitude towards

robots – as lifelike as possible, or a robot that

looks like a robot.

1

2

3

Special skill - agile and human-like body movementsUsed to test military clothing under stress, in conditions where you wouldn’t want an actual human inside. Another Boston Dynamics creation, as is….

Special skill - robot dog that can take a kickingFour-legged mechanical friend famed for its sense of balance. Able to negotiate remarkably tricky terrain on its motorised legs, but more impressively still, can regain its balance and stay on its…er… paws even after a hefty shove. 2016’s SpotMini is even smarter with a gripping jaw and able to duck under tables as it pads round your home.

Special skill - highly advanced prosthetic armIn fact the first to use electrical signals from a patient’s muscles to generate complex movement. Importantly LUKE’s return-force technology allows users to tell how strongly they are gripping things, as in this video of a patient picking up eggs.

SPOT

LUKE

5

6PETMAN4

Special skill - world’s first robotic kitchenTwo robot arms that can record a chef’s movements – whisking, mixing, chopping and more – then play them back to whip up dishes within its own glass-panelled suite of cooker, sink, tools and ingredients. Comes with an iTunes-style library of recipes. Currently prototyped with plans for a 2017 launch and an estimated price of £50k.

Special skill - spookily real facial expressionsBuilt to resemble an actual person – robotics engineer Henrik Scharfe – as an exercise in testing our emotional responses to a super lifelike robot. Developed by Hiroshi Ishiguro of Osaka University’s Intelligent Robotics Laboratory who had already taken a stab at recreating his own features with his earlier Geminoid Hi-4.

MOLEY

GEMINOID DK

7

8

Special skill - Just like a real boy All three iterations, at various sizes, are the work of SoftBank Robotics who produce “kindly and interactive humanoid robots”. They exist in a blurring line between toy, educational aid and personal assistant, particularly for the elderly or those lacking mobility. While the comic-like eyes and three fingered hands don’t do much to inspire confidence or seriousness, the French-founded Romeo project combines the expertise of some of Europe’s top robotic labs and institutions.

NAO, PEPPER AND ROMEO9

Special skill - Talks to pets Released by LG in 2016, it boasts little of the sophistication featured elsewhere in this list. However, Rolling Bot is notable because it more closely represents the first wave of robotics we’re actually likely to see in our homes. It’s essentially a rolling ball with camera and speaker, which you steer via a smart phone app. LG thinks it might be useful for people wanting to check up on or chat to their pets while out of the house. In truth no one really knows what it’s for, other than an early marker in a very new market.

LG ROLLING BOT10

AI and Agent working together

When KLM Royal Dutch Airlines opened up Facebook Messenger as a service channel in March 2016, the result was a massive spike in volumes that threatened to overwhelm advisors.

Approximately one million messages arrived in the first three months – a leap of around 40% on the day of its activation.

To lift the pressure, the airline opted for an AI solution. They partnered with customer service automators DigitalGenius to create a machine learning system that would save advisors time by surfacing the most relevant responses.

The software chewed through around 60,000 questions and answers operators had encountered in recent years. Now when a customer message comes in, the AI sifts that database to isolate answers that offer a general resolution, and the agent picks the best.

You can’t keep adding advisorsToday the airline receives around 100,000 messages a week in 12 different languages. They’re handled by 235 agents and it’s just not feasible to keep increasing that number as volumes rise, admits KLM. AI was the most promising solution.

But why go for this hybrid approach rather than a standard chatbot? Because, as the firm’s Social Media Manager Karlijn Vogel-Meijer says: “KLM wants to deliver a personal feel and we truly believe bots are not able to do so at the moment”.

Supporting agents with AI suggested answers is the best of all worlds “a timely answer, a correct answer and a personal answer.”

Perhaps not surprisingly. DigitalGenius president Mikhail Naumov agrees. Chatbots’ construction around basic linguistics means they are too reliant on keywords, he argues. “They will work well in very defined areas – marketing or weather for example – but when it comes to customer service… they can break.”

Essentially machines should take care of machine-level tasks, while humans should “do the stuff they do better than robots,” he believes.

Changing customers’ views on what makes a fast responseWhile it’s an AI success story, KLM may also have unleashed a monster of their own making. Their customer experience goal is to get back to every enquiry within an hour, something KLM have received praise for in the past. But now customers using social media messaging as a channel are complaining if they haven’t heard back within 20 minutes.

“With chat apps the one-hour response isn’t good enough,” conceded KLM Social Media Hub manager Gert Wim ter Haar at last year’s Digital and Social Marketing Conference in Amsterdam. “We’re seeing a strong shift in customers preferring chat apps rather than sending that public tweet or message. It’s more personal.”

But when they chat similarly with friends they get an answer within minutes, says Wim ter Haar, and that expectation rubs off. “You have to find ways to make it easy to have a personal interaction but cope with the huge volumes.”

Supporting agents with AI suggested answers is the

best of all worlds: “a timely answer, a correct answer and a personal answer.”

The impact of empathy

At the D&SM conference, Wim ter Haar echoed the point about chatbots only being able to go so far, illustrating it with KLM’s own satisfaction score with customers when they were in talks over compensation. When customers were shown empathy, they gave the airline a score of 7/10, even if KLM didn’t pay out. When no empathy was shown the score was 1.5/10 – despite KLM awarding the highest level of compensation available. Empathy is essential, but it’s eluding chatbots and the AIs behind them.

CASE STUDY

Automating your organisation? Ask yourself this.

Anish Shah heads Capita’s Robotic Process Automation (RPA) team, helping embed RPA and Artificial Intelligence solutions across Capita’s businesses to drive process efficiency and effectiveness. It’s an important part of Capita’s belief in automation as a powerful element of overall transformation. Prior to Capita, Anish advised on RPA for a global consultancy and has established SSC in India and Poland.

We asked him to address some of the questions that most often run round organisations’ minds when considering automation - including the most common mistakes they need to avoid.

1 - How can I tell if my business will benefit from automation?

“Generally automation itself is a bit of an eye opener for most organisations. They may have attempted it before, but with mixed success, and often their perception of the level of benefits automation and RPA can provide are… quite far fetched.

“Organisations generally embark on RPA as a cost play, but they continue it because of the governance and control benefits it brings. It’s about the quality and effectiveness of activities and processes, and the level of customer service you can offer to your clients. These are things that can be harder to do through more traditional transformations.

“The key thing for organisations to understand is how to identify processes that are conducive to automation.

“I think the biggest piece of advice I would give businesses is don’t consider your world as a set of processes, consider it as a set of activities… many activities make a process.

“So I would say to clients, look at your operational footprint across back, middle and front office, and identify all the activities that are transactional, rule-based, repetitive and high volume. Once you find activities that tick these boxes, you then have a long list of good opportunities you can prioritise.

“Automation can drive efficiencies in a number of activities within a process so you will get a more efficient function overall. And don’t be surprised if you come across tasks that are being done due to historical reasons or through inherited practices from existing teams. Robotics allows you to challenge that and support a standardisation agenda.

“The other thing I talk about is how to get the maximum benefit. Automation is not just about replacing the activities you are currently doing with humans, but about how you can better equip humans to interact with a customer.

“So for example HMRC uses assisted automation, and that’s not about automating the process end-to-end, but about helping their customer service agents provide a better customer experience to their clients.”

Automation is not just about replacing the

activities you are currently doing with humans, but

about how you can better equip humans to interact

with a customer.

Shah

Anish

2 - What are the most common mistakes I need to avoid?

“The one I’ve seen most often in the market is: ‘oh, so that means I can remove my back office completely.’ No, that’s not the case. Part of the back office, yes, but the entire back office, definitely not.

“The second misconception I see is that this is a panacea; it’s going to solve all of my problems and make everything faster and simpler. No it’s not. It’s one of the many solutions you have. Bring it together with a transformational toolkit that encompasses people, process, technology and governance, and it can deliver a step change in your operating model.

“What often catches organisations out is scaling a successful RPA pilot in order to commercialise the opportunity. It’s not unusual to see inertia in standing up a team and programme to drive automation across an organisation. In some cases it’s the organisational structure that prevents the drive, or it’s the proverbial ‘make or buy’ decision. There can also be a debate about whether this is a business or an IT initiative. For me it’s business led, supported by IT.

“Whatever option you choose will depend on the organisation you are, the level of change you’re looking for and the degree of commitment from leadership.

“To date, I’ve never seen an organisation that’s had a successful pilot turn down the opportunity to do more. The case for change is very compelling and the benefits so significant that it’s hard to ignore.”

3 - How do I go about successfully embedding automation within my organisation?

“Where organisations can struggle here is in finding the right operating model of how they’re going to implement automation; what’s the ideal RPA blueprint for their business? It’s all very new and they don’t know how to manage it. Should they build capabilities, buy capabilities or do a hybrid?

“My advice would be, once a pilot is successful, then depending on the size and scale of the opportunity, you should always evaluate a Centre of Excellence (CoE) model. It should include two key roles of Business Analyst and Technical Analyst, with an option for Hosting and Robot Management. This gives a focus that can drive automation opportunities across the organisation.

“For example, I remember in one global bank there were several teams across the organisation, each piloting RPA, each having a successful outcome. There were pockets of teams around the world all experimenting and saving money. But because it was not centrally governed and tracked, the benefits were not visible. The other challenge to this model is that you are potentially under utilising the capacity of your robots. “Drive RPA centrally, with the right governance, and you have control over what’s being implemented, what’s being prioritised and how the benefits are being tracked.”

4 - What are the major developments I should be keeping on top of in the next few years?

“Over time, organisations have attempted to centralise and standardise, they’ve looked at shared services, then at outsourcing, then at GBS (Global or Group Business Services) all with the aim of reducing costs and improving the quality of processing.

“Now, it’s about automation and simplification because organisations have ‘been there and done that’ with regards to low cost locations and driving process standardisation. Robotic Process Automation offers an opportunity for organisations to drive significant cost savings within a short time frame. Coupled with technical solutions such as chatbots, apps etc, they can also take customer service to the next level.

“In the next two to three years I see three main things happening…

� Robotic Process Automation software will embed Natural Language Processing and elements of AI. That will mean whole swathes of activities that are currently out of RPA’s scope will become part of how organisations will deliver service.

� The cost of implementing an Artificial Intelligence solution - which is the utopia for RPA – will become much more competitive. At the moment it’s prohibitive for many, but as the cost falls, there will be a shift towards adopting this technology - cognitive computing, reasoning, learning and so on.

� And there will be a fundamental shift in the skills required of your workforce. You’ll no longer need people who ‘know the rules’, you’ll need people who can interrogate data, understand it and spot trends that will help the business grow. Most other activities will be automated and generated by a AI or RPA tools.

“I’d have to say that I’ve been a shared services person for ten years, and I’ve never seen a technology with so much potential to fundamentally change how we support our clients and help them solve real business problems.”

AI Celebrities

If you were so inclined to produce a list of ‘famous’ AIs (real ones that is, rather than the world-ending SkyNet of Terminator…) at the top would likely be IBM’s Watson and Google’s DeepMind.

Watson, named after IBM’s first CEO, Thomas J Watson, is strictly speaking a QA (question answering) computing system that applies natural language processing, automated reasoning and machine learning to the vast amount of information at its disposal. It effectively ‘thinks’ about the question – in more than a hundred different ways, according to IBM - and returns an answer. This sets it apart from precursor systems that merely serve up the documents their searches have concluded should hold the answer somewhere, however ‘smart’ those searches may be.

Famously Watson ‘won’ America’s Jeopardy quiz show in 2011 with the aid of four terabytes of data (including Wikipedia in its entirety) but no Internet connection to fall back on.

Since then IBM has flexed Watson’s silicon sinews in ways as varied as cancer treatment, third world development, music recommendations, children’s talking toys, analysing aerial imagery and – of relevance to us – customer service. American customer experience business Genesys began using Watson to sift through its call data in 2014, initially to detect when a self-serving customer was becoming frustrated or about to skip a sales opportunity, and diverting them to an agent.

That year IBM also invested $1bn in creating a business unit around Watson, building on three different ‘personalities’ – Watson Discovery for R&D, Watson Explorer for enterprise insight, and Watson Engagement focusing on natural-language, self-serve applications.

Last year Alder Hey Children’s Hospital became the UK’s first ‘cognitive hospital’ using Watson to analyse and improve patient experience. At the other end of the scale, smaller businesses can rent a Watson Virtual Agent for a monthly fee, and customise it to deliver conversational chatbot answers.

Not that’s everything’s been rosy in Watson’s silicon garden. At the end of February its highly touted collaboration with the US MD Anderson Cancer Center was placed “on hold” following multiple issues around costs, strategy, and an audit saying Watson was simply “not ready” for use by doctors in the real world, impressed as they were by its talents. Forbes declared the audit “scathing” and saw it as a significant setback for AI medicine.

Coming up behind Watson (with medical issues of its own) is DeepMind. Originally a British start-up, it was acquired by Google in 2014 and takes a different slant on AI with the avowed aim of actually “solving intelligence.” Whereas Watson shares some of IBM’s cultural DNA – the methodical application of programming to solve problems one at a time – DeepMind has a little more of the big picture Google about it. If you can truly understand human intelligence, and shape machine learning to match, then you could build an intelligence beyond that of human, one that might one day solve problems outside our mental capacity.

Perhaps not surprisingly, that’s made a few people uneasy. This hasn’t been helped by Google establishing an ethics board to consider the work DeepMind does, but resolutely declining to say who sits on it or what they debate. And in 2016 DeepMind was caught up in a controversy over work with three London hospitals that appeared to give it access to highly sensitive patient data, without those patients’ consent. A data agreement with the NHS was reached, but industry commentators have continued to eye Google’s perhaps avaricious attitude towards that data with disquiet.

Nonetheless the DeepMind project is big, bold and ambitious. Founder Demis Hassabis has dubbed it ‘the Apollo programme’ of AI, and envisions contributions in everything from advanced medical treatments beyond current science to powering thinking, moving robots. If it does indeed ‘solve intelligence’, the results would be revolutionary.

So what if they had a fight? Interesting you should ask. In February DeepMind revealed that it had been exploring what would happen if different AIs were set competing tasks. A real world example might be one AI charged with speeding up a city’s traffic control, while another was given responsibility for reducing pollution. In a game test, two AIs were given the job of collecting dwindling resources with the ability to fire on each other and temporarily take the opposition out of the game. The cognitive capacity of the two AIs was then steadily increased. Would they learn to cooperate, or would they conclude aggression paid off?

Result? Sometimes they collaborated, but often the smarter they got, the more they fired...

Customer Management Intelligence will be published every month, bringing you the latest updates, news and insights in your sectors.

If you have feedback, questions or any thoughts on what you’d like us to cover, please contact: Jo Knight, Client Experience, Customer Managementjo.knight@capita.co.uk07710 384705

www.capitacustomermanagement.co.uk

linkedin.com/company/capita-customer-management

@capitacustmgt

Connect with us

As the UK's leading provider of business process management and integrated professional support service solutions, our 73,000 dedicated staff across the UK, Europe, South Africa and India help make processes smarter, organisations more efficient and customer experiences better. We unlock value by applying talent and technology for you, your organisation and our communities.

Capita is quoted on the London Stock Exchange (CPI.L), and is a constituent of the FTSE 250 with 2016 revenue of £4.9 billion.

Further information on Capita can be found at:www.capita.com