How businesses can get maximum value from chatbots and their smarter cousins, digital assistants
Synthetic were interviewed as part of an OUG article on chatbots. You can see the full magazine article here
Chatbots are everywhere. They are becoming a genuinely valuable tool, for both internal and external interactions, in almost every conceivable organisation. Oracle’s own research confirms this, claiming that some 80% of businesses are planning to deploy a bot by 2020. So, love them or loathe them, it’s unlikely you’ll ever be able to ignore them.
Debra Lilley, an associate director at Accenture, has worked on the development of an HR bot using Oracle
HCM, and has been struck by how the rate of chatbot- adoption is currently outpacing many other new technologies. “In my experience, there’s always been a big gap between technology becoming available and it being picked up by end-users,” she says. “In other words, a gap between the marketing and the reality. But that hasn’t been true with chatbots.”
The reasons for this are simple, she believes. “I think it’s because the chatbots themselves are just another channel, just a way of getting to data quickly. The public is no longer worried about ‘chatting’ with a machine, as long as they get their answer.”
That’s a view shared by Roy Murphy, founder of Synthetic, which has developed chatbots for a wide range of clients, including Oracle (see page 22).
“Conversation is so easy – we all understand it, right?”
he says. “So the discussions we’re having with clients are more about the ROI and business benefits rather than, ‘What the hell is this thing and why we are talking about it?’”
He also believes speed to market – and the associated low costs – are prime reasons for the rapid uptake. “To get a pilot up and running, in most cases, takes under a month – and then it’s not particularly difficult, through bot analytics, to quickly ascertain if it’s a channel that’s worth pursuing,” he says.
The benefits of bots
|
Chris Ashworth, CIO of parcel delivery company Hermes, has seen his business’s first chatbot deliver a considerable ROI very quickly. In December 2018, Hermes deployed an Oracle bot to improve its online customer service, which until then was outsourced as a basic webchat function. “ That was a tool with a person behind it and at peak times there could be a delay before someone got back to you, which wasn’t great from a customer experience perspective,” he says. “So we believed the bot would have a strong use case for instant customer query handling, and we were right. We saw immediately that it was handling 20% of all bot/webchat queries without needing any human interaction – and that has now risen to 38%.” This efficiency gain has led to several direct benefits
|
On top of that, there has also been a huge uplift in customer satisfaction levels. “We’re getting a positive response,” says Chris. “Our CSAT, net promoter and TrustPilot scores all went up. We’ve found that if you can give the right information, speedily – and it’s rich information – customers really like it. And, of course, if the bot can’t do that, customers can still always break out and communicate with a real person if that’s their preference.”
Debra Lilley sees similar benefits within the enterprise for HR bots. “Most organisations have a helpdesk for HR, which is unlikely to be open 24 hours a day,” she says. “But a bot is a channel that allows you to get answers when it suits you and in a way that suits you. This leads to a reduction in the number of people required to run the helpdesk. So, if you get it right, the number of queries answered should go up, your employee engagement should go up, but the cost of servicing it should go down.”
Both Chris and Debra believe that starting simply is the best place to begin. “It’s about picking the right challenges to tackle in the first place,” says Debra. “For example, over the past five to 10 years, the number one thing that has reduced the quantity of support calls a helpdesk receives is self-service for changing your password.
So I think it’s about capturing those kinds of things, the ones that you get asked lots of times that are simple to answer without the need for any decision-making, like ‘I’ve lost my payslip’ or ‘I need a letter for my building society’.”
“We just automated one customer journey at first,” says Chris. “So if someone comes onto our portal, we can give them real-time, up-to-date tracking information through the bot, including signatures, a photo of the safe place we’ve left an item in, and a map of where it was left.”
Any organisation that’s required to interact with customers in some way should find the technology can bring real benefits, argues Grant Ronald, Director of Product Management at Oracle and a specialist in chatbots and digital assistants.
“A good starting point is: Can we make this process more efficient?”
Can we make information more accessible and easier to find for our customers?” he says. “For example, at a call centre, how many actions are low-fidelity and relatively straightforward to action? Things like: ‘Can you reset my password?’ Couldn’t we save the real humans for where they can add value?”
‘We’ve found that if you give the right information speedily – and it’s rich information – customers really like it’
Alongside these benefits, Roy Murphy believes there can also be a considerable upside in terms of enhanced customer insights, thanks to the data that bots are able to capture. “Within the parameters of GDPR and data privacy, there is a lot you can learn from conversation flows, from what people are asking and where they’re getting stuck – this is information you can see and therefore act on. It’s gold dust: people are actually telling you what they think.”
Chatbot challenges
Despite the undeniable business benefits, installing and running bots is not all plain sailing.
At Hermes, Chris came up against a few teething problems when their bot first launched. “We use Oracle CRM for our contact centre, so our bot is integrated into that,” he explains. “But we found it a bit clunky at first. I’d say we were ‘early adopters’ rather than innovators, but it turned out that not a lot of other businesses were doing this already, so there was some bleeding-edge learning for us all.”
Managing a high volume of queries was one early challenge. “On our busiest day, we deliver 2 million parcels and one of our most popular models is to leave a parcel in a safe place,” explains Chris. “With 40-50% of our parcels being delivered that way, you can imagine the number of queries that can come in from customers as to where that place might be. Lots come via webchat, and this caused some outages at the start. The scaling took Oracle some time, but now they’ve really hardened their infrastructure to become more resilient.”
Nonetheless, Chris is ensuring Hermes has an alternative available, should the bot service ever go down. “We will definitely always have another route into Oracle, so we can always have a solution available for our customers,” he states.
One of the challenges Debra Lilley encountered early on was with the Oracle HCM API that the chatbot used to access employee data. “The problem was that it was initially too open for our purposes. If you had access to the API, you had access to everything about everybody,” she explains. This meant that, even if a chatbot was able to respond to a simple query about salary, for example, the API would also grant access to other data that could end up in the wrong hands.
“Therefore, initially, we wrote another API to interface with the Oracle API in order to add an extra layer of security, ensuring appropriate levels of confidentiality,” she continues. But the good news, believes Debra, is that Oracle is now addressing this as well. Improved APIs, via a wide range of “skills” on offer through the Oracle Digital Assistant platform (ranging from salary and pay slip queries to job searches), should soon be made available to programmers and eventually become fully open.
What’s more, Debra is impressed overall with the Oracle technology. “I’m pleased they’ve moved to Digital
Assistant,” she adds. “That’s a good step that will give more control over the platform. It’s very simple to use and I believe it will be popular. I’ve also seen customers using it against non-Oracle applications, as the great thing about a digital assistant is that the data doesn’t have to come from a single source – so you could be pulling some data from Oracle HCM, for example, and other information from another external payroll provider such as SAP.”
Roy Murphy also believes the Oracle platform is a strong competitor in the market.
“It’s a platform where some technical understanding would be helpful,”
he says. “But for larger projects, if you need enterprise-level technology with robust security baked in, cloud infrastructure and so on, it definitely fits the bill.”
At the same time, Debra recommends that customers be mindful of Oracle’s pricing model. “It’s not the cost,” she says, “but the metric, which is based on the number of interactions with the system. So if I said to the chatbot ‘hello’, my organisation will pay for that hello and every subsequent interaction – and the bill-payer has no direct control over the end-user. We’re not talking a lot of money here, but customers should ensure they can limit the number of interactions to avoid costs running away from them.”
Chris also has a note of caution around expecting an immediate ROI from the cost-saving bots can bring. “We wanted to do things properly,” he says, “so we needed to invest a lot in our infrastructure to make everything real-time and deliver a proper micro-service. That meant we went from traditional on-premise into the cloud, we got rid of all our batch jobs, and so on – all of which had a related cost. That said, as a tool in itself, the bot is definitely paying for itself.”
Meanwhile, Roy advises organisations to ensure they have a robust data and privacy policy right from the start. “It’s crucial,” he says. “You’re collecting people’s data – that’s nothing new, and you should be following all the rules anyway, but that’s the first thing we always talk about with our clients. Even if you’re running a pilot, you must have that side of things nailed down.”
One of the other great challenges facing organisations who are starting to use chatbots and digital assistants is the question of how ‘human’ they ought to be, and whether the person interacting with them should be made aware of whether they are communicating with a real person or a machine.
“Our customers always know they’re interacting with a bot,” says Chris Ashworth. “We started off with a very vanilla one, but now we have a few and we’re trying to give them a playful character. But it can’t be anything too funny, as situations can be quite emotive with parcels, especially if one goes missing – it will mean something to somebody. We try to get things right all the time, but when you’re delivering 350 million parcels a year, you’re going to get the odd one wrong and we need to be sensitive when that happens.”
Grant Ronald also believes it’s important to be up-front with customers about who – or what – they are interacting with.
“Don’t try to fool the user,”
he advises. “You’re not trying to trick anyone – you’re trying to help them. Giving human characteristics to your bot is fine, but without passing it off as human.”
He is also noticing that people interact differently with machines, making their expectations suddenly change depending on whether they’re talking to a bot or a human.
“Think about how you might ask Alexa to play some music, versus how you might ask a human,” he says. “If you know you’re not interacting with a human being, you’re going to be more concise and direct – and possibly more forgiving if it makes a mistake as you’ll be aware of its limitations.”
Roy Murphy, meanwhile, sees a clear advantage in the fact that bots are not human, which can be helpful when, for example, dealing with personal information.
“People are definitely more open with bots,”
“But as a business, you have to be very sensitive to that, as someone is trusting you to look after their data. What’s more, at the conversational design stage, you may need to ensure certain trigger words or phrases will alert the bot to hand over to a live agent – particularly in areas like healthcare.”
Therefore, businesses need to ensure they are paying attention to a lot more than just the technology – for example, areas like language, tone of voice, and so on. “There’s a whole bunch of new skills required here,” says Grant, “such as natural language processing and conversational design, which is massive.
For example, people don’t always say what they want, they might just say what the problem is – so you have to understand a customer’s intent. For example, ‘My phone is broken’ could signal intent either for wanting to have it fixed or for wanting a new phone. That’s the kind of thing bots have to learn to deal with.”
The bots of the future
So where next for conversational AI? Grant Ronald is excited about Oracle’s move to digital assistants. “It’s moving things onto the next level,” he says. “For example, you might ask the assistant: ‘Can you book a meeting room for me for next Tuesday? And for that meeting can we have the last three weeks of reports?’ That kind of engagement is taking it up a notch from a simple question and answer and is helping people in their day-to-day processes.
“And the great thing is that it will know the right process to kick off, as well as the context, and start to consider things related to that request,” he continues. “So I might say: ‘Bring the reports for the last three years.’ And if I’m the HR Director, it’ll know I mean HR reports because that’s what I always ask for.
But if the Finance Director asks the same question, you’ll get a different result, because that person always asks for financial reports.”
And that’s just the start, he says. “What if our digital assistant could be proactive? For example, if you always ask for reports on a Monday morning, it might ask if you’d like to have them emailed to you automatically at the same time every week.”
Meanwhile, Chris Ashworth is looking at rolling out the technology right across Hermes – possibly for internal-facing functions, but still primarily focusing on customer service. “At the moment, there are so many use cases in CX, and so much low-hanging fruit there, that we’re going to tackle those first,” he says.
‘People are definitely more open when speaking to bots – but as a business, you have to be very sensitive to that delivering greater personalisation, trialling the automation of proactive communications, and creating a bot on the WhatsApp platform.
But whatever your own experience of bots, and whatever applications you envisage within your own organisation, Chris advises that a key priority should always be to maintain a relentless focus on your users. “Everything we do these days, we take back to our customers, as customer experience is a big differentiator for us,” he says. “So you have to really understand your customer journeys and customer service. And then you must keep testing it, because often the journey you think you’ve got isn’t the journey you’re presenting.”
Always being able to give accurate and instant responses is also vital, he adds. “For this, you need to make sure you get your infrastructure right – and you can’t take shortcuts. If you’re going to do this, it’s got to be real-time, it’s got to be elegant and well thought-through, because you can’t kid a customer – they know if they’re getting something that works or not. But when you do it well,
it can be really, really good, both for you and for the customer. Those are always the best kinds of solutions!
Drop us a line at hello@thesynhetic.co to kickstart your AI journey — we run explainable AI workshops, run industry-specific roundtables to help power your customer engagement and ROI with conversational AI.
We partner with a range of forward-thinking businesses — from global healthcare corporates and enterprise technology giants to innovative charities and edu-tech startups, on conversational AI initiatives — chatbots, voice UI, digital assistants, and emerging technologies.
Get expert campaign planning, customer experience mapping, conversation design-build and optimisation.
Conversations Worth Millions.
