Simon Wakeman podcast cover

Episode 4

Simon Wakeman - Automation and conversational AI

Posted on: 06 Nov 2023

Simon Wakeman is the Chief Operating Officer at the UK technology group The Panoply and former Chief Growth Officer at GreenShoot Labs, a digital agency specializing in chatbots and conversational AI.

In this episode, we talk about the role of automation in digital transformation, with a focus on conversational AI and similar technologies. We explore how they can help humans perform more efficiently, discuss some of the challenges that hinder widespread adoption of automation, and take a look back through the years at how much AI and chatbots have improved since their early iterations.

Links and mentions:

Transcript

“We have to think about; are we asking the AI the right questions early on. If we ask AI the wrong questions we will get the wrong results based on the data we put into it and one of the challenges is actually making sure that we don't just ask-- we don't ask the wrong questions to the systems we build.”

Intro:Welcome to the Agile Digital Transformation podcast, where we explore different aspects of digital transformation and digital experience with your host Tim Butara, content and community manager at Agiledrop.

Tim Butara: Hello everyone! Thank you for tuning in. I’m joined today by Simon Wakeman, chief operating officer at the UK technology group The Panoply. In this episode, Simon and I will be talking about the role of automation in driving digital transformation. Welcome Simon, it's great to have you here.

Simon Wakeman: Thanks, Tim. It’s great to be with you.

Tim Butara: Okay, I think it makes the most sense to begin with something more general and basic and because automation is quite a broad term so not all of our listeners may know exactly what we'll be covering today. So, can you give us kind of a brief introduction or a short overview of automation.

Simon Wakeman: Absolutely. Sure. So artificial intelligence, AI, has been around a long time, lots of people talk about AI, but what we now see is the maturing of artificial intelligence. So, it actually means that automation as part of a digital transformation program is a reality for organizations in lots of different sectors.

If I think about the kind of work we do at GreenShoot Labs, we work in two particular areas. We think about automation as making someone's life easier. We're either making life easier for our users so they could be our customers or our members or our donors if we're a charity or we could be making life easier through automation for organizations. So people that work in a company or a charity or a government organization, how can we automate the work they do day to day to make their lives easier and to make things more efficient for the organization?

Tim Butara: And also in the context of making things easier for your organization, not only easier but also leaving more space for the creative tasks you know because it's usually the things that can be automated that take up a significant chunk of an employee's time so being able to automate that kind of gives so much power and agency to that employee.

Simon Wakeman: Absolutely. And yeah we can… really our focus with automation should be about how do we automate the things that machines are good at and humans actually aren't that good at and how do we, therefore, make space for humans to do the things that only humans can do.

Tim Butara: So I guess we kind of answered my next question at least partially now and that is why automation? Like what is… Is there anything else that we haven't covered under the business value of automation and how it can help drive DT?

Simon Wakeman: I think-- I mean if you look at it from that lens of users and organizations, for users really that means that we can drive up satisfaction, so we can have happier customers. We can have less customers leaving us so we can reduce customer churn potentially. And when we're trying to sell additional products, there’s cross-selling to existing customers, we can drive value per customer up as well. So there's a number of different metrics that actually demonstrate the value of automation in an existing customer or to a customer setting.

Within an organization we can think about, how can we use automation, artificial intelligence, conversational AI to drive down the cost per serve. So they cost us less to serve our customers and actually get a better experience and also how can we scale. So one of the some of the work we've done at GreenShoot Labs has been helping small organizations actually serve far more customers than they could ever do through traditional means purely by using conversational AI.

Tim Butara: So you'd definitely say that there's a lot of potential here and kind of it's a very broad… the benefits are very far-reaching?

Simon Wakeman: Yes. Yeah. And I think that's the interesting thing and, I mean we've, GreenShield labs has been in existence as a standalone business for about 3 years now and even in those 3 years we've seen the technology come on leaps and bounds and that means that really the possibilities of what we can do now will be small compared to what we can do in 2-5 years’ time.

And actually, for lots of organizations, it's about how do we get on the journey? How do we start experimenting and learning about conversational AI? Because once you've started that the possibilities will grow in the future as the technology continues to evolve very rapidly.

Tim Butara: Yeah, it's definitely one of the fields that I’ve seen grow the most rapidly. I think we're primed for even more exponential growth of AI and similar technologies.

Simon Wakeman: Yeah I would agree with that, definitely. Tim Butara: And what are some types of automation that we're seeing being employed the most frequently? Are there some trends that you expect to see rising enormously in the coming years?

Simon Wakeman: I think the ones that we're seeing right now are the start of automating tasks that are relatively low level, relatively high volume. So if we think about the complexity of a task for someone who maybe is contacting an organization to receive a service, to sign up for something, or to change their details if they're moving house or something like that. Those low level repetitive tasks are straightforward to automate and actually, those are the kind of things where customers are getting a far faster service. Within organizations, we're working with a number of organizations about their HR and their IT help desks. So for example resetting a user's password within a corporate organization is actually a really time-consuming task because it happens so often.

How can we automate that and we can save time for those users within the organization? So I think at the moment it is those low-level tasks that are, low-level single or short sequences of tasks that are easily, most easily automated. The growth in the future will be about the multi-stage conversations that will happen. So it won't just be a case of a few short interactions, it will be actually a long sequence of interactions, open-ending interactions where it's far more complex and we may be dispensing for example more complex advice rather than fulfilling a simple task - it will be about listening in the conversation, the artificial intelligence understanding what's going on in that conversation and then providing appropriate advice within the bounds of the way the system has been configured. And those are the kind of things where we're still at a relatively early stage for market maturity.

Tim Butara: Okay I want to just return to something, to the example that you gave earlier about automating the updating of a password and the thought that crossed my mind when you said that - are there any additional risks, especially in this example, that you need to be aware of if you do it in an automated way rather than doing it manually?

Simon Wakeman: Yeah, there's certainly security considerations to think of. I mean, within an enterprise it's relatively easy to provide, to undertake some of those kind of things, but we see for example financial institutions using voice recognition as a tool to authenticate as well. So the combination of automating an audio-based chat-bot and voice analysis actually can provide really high levels of security. So there's different options but it's a… yeah, it's definitely something to think about.

Tim Butara: That's a very good example, yeah. Can you think of some further, maybe even more specific examples of automation being put to really great use like solving really particular pains or producing very unexpected gains?

Simon Wakeman: Yeah. I mean, so one of them is one of my favorite examples is a project we did for a UK charity called Cyber Helpline. So this is a very small charity set up by people who are cyber security experts and they set up this charity to help people who've been the victims of cyber-crime. So for example, if you have a virus on your laptop or ransomware or anything like that because that's a very-- that's a major and growing problem but it's also very difficult to get help because it's a very specialist area and they wanted to grow… they have volunteers who provide telephone support but they wanted to see how can a chat-bot help them extend the reach of what they do. So what we did was we did a discovery phase and understood how they provide the service at the moment and then we thought how can we use a chat-bot to provide that service.

So what we've developed is a chat-bot that anyone can use and you go on to it and you just describe what's happened. So you may say, “I’m being asked to pay 100 bitcoin to access my files” or something like that, and then we've developed a knowledge graph working with the volunteers. So we basically codified the knowledge they have and then built a system that analyzes what people tell it and then does a scoring so it will basically say, I think this is 90 likely to be ransomware or 75 likely to be a virus or whatever the type of attack is, and then it will provide some questions that allow it to clarify its thinking. So is it right? Is it definitely ransomware? And then it will provide advice on the back of that. It will also escalate a more complex case. So if it's not sure or if it hasn't seen something like this before because the threat is always changing it will escalate that to a human.

So what we've allowed them to do is reach far more people than they could with just human volunteers and actually, we're really proud that the detection rate is about 73% so about three-quarters of every problem that you present to it, it correctly identifies and then gives the advice that's relevant to that. The client is also able to update that and so the knowledge graph is always changing as cyber criminals change their attack vectors. We can-- we've allowed the client to update new… put in new attacks, tell it what to do, how to identify the kind of words that are being used. So it really is helping people who are victims of a new form of crime help for themselves.

Tim Butara: Yeah, that does sound like a very practical and very useful use of automation. Especially as you said, a kind of a new form of crime which is probably only going to grow probably also exponentially. Like, we see in the fields such as, for example, WordPress has a huge market share and that's why WordPress websites are so often the target of attacks just because it's such a large percentage of the web and I guess it's much easier. I think I had a conversation about this once that, of course you'd be more worried about somebody breaking into your house, but you'd be much less concerned about it happening than somebody hacking into your website, stealing your personal information stuff like that.

Simon Wakeman: Absolutely. Yeah, and it's a fascination and I like it because you can basically… if you think what they're doing there, they're taking advice that's provided by humans and turning it into a machine.

Tim Butara: Yeah, exactly.

Simon Wakeman: And that you could apply that to for example charities that may be helping people with medical problems or charities that are helping people who want to go to university, who want to understand the process. So you can extend that model to all sorts of different use cases relatively easily.

Tim Butara: Yeah it’s actionable real-life help that helps in all kinds of situations, not just something cyber or digital-related and also-

Simon Wakeman: And people want to check that one out. If you Google Cyber Helpline then the chatbot is there in live so you can go and use it at any time. Hopefully for not-- hopefully you're not a victim of crime but it's there to have a look at if you want to see it in real-life, in action.

Tim Butara: Awesome. I’ll link that in the additional resources for the podcast episode. And also, what you said has got me thinking; this is a perfect example of how automation and humans have to work together. Because this wouldn't be possible simply with automation and it also it would be much harder to do it on such a scale and such a high rate of success if it were only coming from the human side. So this is a prime example of maybe what the future holds in store for kind of the merging of automation of AI and human input.

Simon Wakeman: Yeah, definitely.

Tim Butara: Awesome. We kind of covered some of the values of automation already. What about some of the main challenges and concerns with automations apart from maybe the security concerns that we've already exposed a bit?

Simon Wakeman: Yeah it's a good question. I think there are probably two areas which maybe hold back adoption or are challenges of adopting automation and conversational artificial intelligence. So the first one really is about some poor historic experiences. So if you think about new technologies when they come on stream and they start being used, typically there's a lot of hype early on. Lots of people are very excited about how a particular new technology is going to change the world. So expectations are really raised and then really what happens is the early iterations of that technology are quite disappointing.So a lot of people's early experience with chat-bots has been quite poor. Everyone's used a chat-bot where they've not managed to get an answer or the chat box just talked gibberish back to them. So we're at the stage now where we're probably coming out of that dip and actually, the technology and the implementation of the technology is now up to a standard where we can deliver against expectations but those historic problems people have had kind of hold them back a little bit in their thinking. So that's one of the things we come up with, which is why we really recommend a very iterative approach. So start small, lots of small experiments and learn and build rather than thinking whether a chat-bot is going to save the world, a or chat-bot is going to deliver every answer to every problem that your organization faces, because it won't. So, start small and learn and actually you can learn as the technology learns and that will be in the future a competitive advantage for lots of businesses because they've got in there early rather than they're waiting until someone else has used it better.

Tim Butara: Yeah, because you have to do the work regardless, so if you start doing it early you have an advantage over the people who start doing it late because they'll have to do the same thing basically. If they're not super pioneers or kind of standing on the shoulders of giants and already wait for everybody else to kind of innovate and produce the best possible solutions, yeah. As you said, it's kind of the first impression that makes the biggest impact and that's probably why, one of the major reasons why we're seeing maybe not as much adoption of automation as the initial hype would have us believe.

Simon Wakeman: Yeah, it's definitely… in our minds it's definitely better to do a small number of things very well and try and do a lot of things and not do them very well.

Tim Butara: Yeah. Good is better than perfect, right?

Simon Wakeman: Exactly.

Tim Butara: And you mentioned another area of concerns?

Simon Wakeman: Really it's about technology choices. So the maturity of the technology that we're dealing with. At the moment there are still lots of different technology choices. I mean I’ve worked in the web for 20 years. I remember the pre-CMS era so before the likes of WordPress or Drupal came along and actually people were building custom CMS’s. If you think about where we are with conversational AI, we're probably at that stage when things like Drupal and WordPress were starting to emerge. So there is some coalescence around technologies but actually, there aren't obvious choices for lots of the technologies that you're using.

There are plenty of obvious cloud services that we can plug together, but actually, some of the enterprise maturity isn't quite there yet. We’ve developed an open-source framework called ‘Open Dialogue’ which is actually our solution to the challenges of not having a bespoke system. So it's the start of something like a content management system but for conversational AI which allows you to plug in things like Google, Amazon, Microsoft technologies, but also manage them yourselves and integrate other systems. Because one of the challenges of delivering true customer service through conversational automation is about integrating with other systems, so maybe your CRM system or your payment system or your ticketing system and the integration there you need to… unless you build it bespoke, you need to get some tools, and Open Dialogue is our open-source solution for doing that.

Tim Butara: Yeah, integration is definitely one of the key things that your platform needs to take care of if you want to compete in the kind of multi-channel and the ubiquity of digital experience world.

Simon Wakeman: Absolutely. It’s impossible to deliver lots of the tasks that use good ones if you're not going back to the original system of record or the back-office system.

Tim Butara: Hey but you know what, I noticed that we haven't really talked much about the ethical concerns of automation and probably more specifically AI and machine learning. Do you have any thoughts of that maybe?

Simon Wakeman: I think it's a very hot topic for us at the moment. There's some challenges in the UK about the awarding of exam results recently on the basis of algorithms because of the cancellation of exams and one of the topics that we talk a lot about is actually, it's-- we have to think about are we asking the AI the right questions early on. If we ask AI the wrong questions we will get the wrong results based on the data we put into it and one of the challenges is actually making sure that we don't just ask-- we don't ask the wrong questions to the systems we build. We think very carefully about the questions we ask, the data that we're going to be training the tools on, and how that could introduce bias or inadvertent results into it. And, yes, there are lots of examples where people have got that wrong but again, our steer is, if we start small and don't ask big complex questions, then we can learn, we can make smaller mistakes before we go into bigger deployments.

Tim Butara: Would you say that this is one of the biggest challenges or kind of issues that AI needs to overcome before seeing widespread to worldwide adoption?

Simon Wakeman: I think it's overcoming I think. I think it's in progress. I think we're seeing people learn very rapidly and AI is behind lots of things that people use without even realizing it.

Tim Butara: Yeah, exactly.

Simon Wakeman: When people use the predictive text on their mobile phones which they have been for some years now, they are using artificial intelligence. But what we see is as the adoption broadens out beyond those really early adopters, there's another wave of learning to happen and I think that's where those, yeah, organizations need to be careful because there's reputational risk if they get it wrong, if they don't think about the ethical side of the questions they ask of AI early in their projects.

Tim Butara: Yep. Definitely. Okay, we're already kind of going more current with our discussion. I mean okay, it is in its essence the current discussion because it wouldn't be-- it wouldn't have been possible on such a scale maybe five or ten years ago. But, like, we are in 2020 after all so how would you say that the COVID crisis that we're going through right now is affecting the future of automation, like there are probably some huge shifts that we're noticing already in that sphere.

Simon Wakeman: I think it's—yeah, I mean I think the way that COVID has changed the way that organizations have to interact with their customers means that people have been far more open to automating technologies, because they realize actually automation is a way to deal with some of the challenges of people not being able to be face to face or being, having restricted capacity in existing channels, for example, because you can't have as many people work in a call center because of social distancing. So there's, I  mean, we have seen in the past three to four months an acceleration in the type of work that… sorry, and the number of projects we're being asked to take on the ambition of those projects as well and the pace at which we deliver those projects. So it's-- for us it's kind of a compression of a trend that was already there or already and it's yeah, it's an exciting time to be working with these kind of technologies.

Tim Butara: Yeah and the keyword here is I think compression because I’ve heard numerous times it being said that basically, the major thing the COVID crisis did was to condense the digital transformation that was supposed to happen in the span of maybe two years to just a few months.

Simon Wakeman: Yeah, absolutely. And I think the other interesting thing is for us is how automation is starting to-- automation and conversational interfaces are starting to become the norm. So the fact that more people have worked from home than ever before and more people are using things like Slack, Facebook at work, all those kind of tools that, means that people are far more comfortable doing-- using those tools for things that they weren't previously doing, which actually means that the conversational interface is just normal. So in a way, I mean it, doesn't take a big leap to say in an organization, why would an organization have an intranet if they're using a conversational interface, when actually you could just ask a bot any question and then the bot does all the hard work of retrieving that organizational information from data repositories all over the organization rather than trying to have an intranet layer that presents that information in a very traditional web interface. So that kind of adoption of the conversational interface I think is a real enabler for another stage of organizational transformation.

Tim Butara: It's like almost accidental adoption, right? Kind of spontaneous adoption.

Simon Wakeman: Yeah exactly. And it's the door and the fascinating thing is what doors does that open for things that we haven't even thought of yet.

Tim Butara: Yeah, because before, say, March most people would not have thought of that as you said.

Simon Wakeman: Yeah. And yeah so we're really excited to discover and do new things as those opportunities emerge.

Tim Butara: Yeah I guess every crisis truly is an opportunity if you know how to act on it.

Simon Wakeman: Indeed.

Tim Butara: And it's interesting, you mentioned Slack as something that you use frequently, daily. Like I remember the first time I used Slack I was really like impressed by the Slack bot. Like I just thought that, I mean the way that it responded to you even then two years ago was really on point. And just the whole way of how they set it up to kind of really brighten your day with inspirational greetings early in the morning that were original and usually also contain some puns or something. Like I remember one of my favorite ones being: “Be cool. But also, be warm.” Stuff like that. So yeah, it's definitely something that is facilitated through the digitalization spurred on by Covid-19.

I think I covered everything that I wanted to talk to you about that's automation specific. Is there something that you'd like to point out or cover that that we haven't really thought of earlier? Simon Wakeman: It's been a fascinating conversation and I’ve really enjoyed it. The thing I would… If people are really interested in this kind of area, there's a book written by our co-founder Dr. Ronald Ashri called ‘The AI-Powered Workplace’ published last year so still pretty up to date and definitely worth a read because it talks about lots of the trends that we've talked about today and provides a real practical handbook on how you can get started with this kind of thing. So yeah, I think we'll, hopefully share that link in the notes afterwards as well because it's a great book to have a look at.

Tim Butara: Yeah, I’ll do that. I always try to include all the things that might be relevant to the readers, to the listeners, you know, because the aim of this podcast is to educate, to provide info, to provide insights, and as much as possible is better than some lacking points.

Okay great. Yeah, as you said, it's been a great conversation. I'm really happy to have you here. Just, before we conclude, if people want to reach out to you or learn more about you what's the best way for them to do that?

Simon Wakeman: Well for GreenShoot Labs, if they check out the website at www.greenshootlabs.com and I’ve got a website at simonwakeman.com where I  talk about these kind of things and yeah Twitter is a good place to have a conversation. So I’m Simon Wakeman on Twitter as well.

Tim Butara: Great. Thanks, Simon. Thanks for the great chat and to our listeners, that's all for this episode. Have a great day everyone and stay safe.

Outro:Thanks for tuning in. If you’d like to check out other episodes, you can find all of them at agiledrop.com/podcast, as well as on all the most popular podcasting platforms.

Make sure to subscribe so you don’t miss any new episodes, and don’t forget to share the podcast with your friends and colleagues.