conversational AI

DRUID Talks Season 2 Ep #1 - Redefining Business with RPA and Conversational AI: Expert Perspectives

This episode explores the evolution of automation, conversational AI's role in the enterprise world and a future where individuals own specialized AI agents.

Season 2, Episode #1 of our DRUID Talks Podcast features Tony Nudd, a seasoned expert with over 25 years of experience in RPA and automation technologies. Tony's passion lies in giving robots a voice in the business world, and he's here to share his wealth of knowledge. He shares insights on the evolution of automation, the power of Conversational AI, and its role in the enterprise world, particularly in enhancing self-service for customers and employees. Together with Kieran Gilmurray, our podcast host, he foresees the future of Conversational AI and RPA in the context of Generative AI, envisioning a world where individuals can own and offer specialized AI agents. This episode is a deep dive into the potential of these technologies and their impact on businesses and individuals. 

Subscribe for the next episode!

Kieran Gilmurray:

Our guest today is Tony Nudd, an RPA and Conversational AI expert who is passionate about business and giving robots a voice. Tony has more than 25 years' experience with RPA and automation technologies, and he shares a tremendous amount of wonderful content online. And we're very grateful today to have Tony with us to share his expertise and experience.

Tony, welcome to DRUID Talks!

Tony Nudd:

Thank you, Kieran. I've seen the first series of these talks. They were really good, and I'm really privileged to be actually part of series two. So, thank you for inviting me.

Kieran Gilmurray:

Well, we couldn't imagine doing it without you.

Tony, we jump straight in! You share a lot of RPA, Conversational AI and voice assistant content online. What is your experience of those technologies?

Tony Nudd:

Thanks, Kieran.

So, I started off with Business Process Management a few years back, now probably turn of the millennium, and that was basically automated workflow and robotic process automation is the natural evolution of that technology set. And so, I've been spending a lot of my previous years automating business processes within organizations, both on personal productivity automation, but also on end-to-end process automations like HR onboarding and mortgage processing.

And what was really exciting about having the ability to do Conversational AI is that rather than use my keyboard and type commands, I can now either use my keyboard to have a conversation with my automations, but more excitingly, I can actually switch on voice and speak through a headset to my robots. And those robots can actually talk back to me and we can have a conversation with each other.

Now, I started doing that a year ago, and that in itself just blew my mind that I could do that. But then, with the explosion of ChatGPT and these larger language models, this has now added even more nuance to the fact because it's almost like I'm having a human conversation but with one of my robots. It's absolutely amazing what's happening.

Kieran Gilmurray:

It's phenomenal the pace of technology. I have to say I get inspired on a daily basis, and I love watching your content, Tony, because you do the personal piece where you look at it and drive your own productivity, and you also look at the business side of things.

If we were to focus on the business side of things for a moment, then where do you think Conversational AI's role is in the enterprise world?

Tony Nudd:

I think it's all around that whole self-service model, whether that's a self-service of one of your external customers or self-service for employees internally as well.

A tremendous amount of time is spent supporting your colleagues, say internally. So, my own team: I get lots of requests for information that they could probably find themselves if they spent a bit of time, but of course, time is something that's very valuable, and people don't have a lot of. So, it's easier to ask a colleague, or obviously, if you're a customer, it's just easier to ring the company and ask rather than go through pages of information.

The advantage of having a Conversational AI now is that we've gone beyond just the FAQ, answer a quick question, and of course, previously, that question had to be exactly correct. Otherwise, the answer would be, “I'm sorry, I don't understand.” Whereas now we've got this nuanced effect of being able actually to speak to the robot or speak to something, and it will actually have a conversation and guide me to what it is I need and then provide the answer in natural language. And this has been the shift we've been able to emulate simple conversation, but been very stifled, very logical, a bit like, you know, Mr. Spock on Star Trek. And it was a little bit, you know, stilted, whereas now it really is conversational.

And, you know, the power of this technology is not just in English. I can actually have a conversation in English with a robot that only understands French, and it will do the translation for me, and it will translate back. The fact that we can have this sort of polyglot ability is just amazing and mind-blowing, and it's only just starting now. Organizations are only just seeing the power of this, and it's the cusp, and that's what's really exciting for me as a technologist.

Kieran Gilmurray:

Think it and say it, then Generative AI and ChatGPT can create, and I do think that, much like yourself, what we will be able to do will only be limited by our own imagination. So, it's exciting times in technology world!

Also, you've only been showcasing fantastic examples of combining a technology that I love, which is robotic process automation with a Conversational AI platform with DRUID’s.

What are the lessons you've learned as you've been doing this, what has your experience been, and would you mind sharing those with the audience as well?

Tony Nudd:

I think, first of all, obviously, you start with the conversation, and then you build the automations to complement that. Or you start with the automation, and you build your conversational flow around the automations.

I obviously start with automations. That's my space. And so, what I found was I had a library of reusable business robots, some of them, as I mentioned before, personal productivity. So, create an email, create a calendar entry, update a spreadsheet, and update the CRM. And previously I would have had to manually start these robots and I would also have to programmatically get data to and from them if they didn't have any systems then they could easily read.

By combining Conversational AI, it means that the robot will wait for a set of data. It will then say to the Conversational AI, “This is the data I need,” the conversational layer will then guide me into collecting the data it needs; it will say, “Tony, you want to create an email; here's the email too,” and then the robot will find that person in my address list and confirm that it's the right person. Then what's the subject? And we can start building this up.

Now, I don't know about you, but I'm very much somebody that likes to sort of think things through; I talk aloud when I'm thinking; I actually like to walk up and down a little bit, so I'm really annoying like that. And maybe over a year ago, I was unable to do that. I was tied to a keyboard, and I had to use my fingers to type. I do this as a keyboard, not one of these; you know, my daughter's brilliant at it. Her thumbs fly around as well. I'm not; I'm a plotter with the keyboard. I've now been freed up. And it really has revolutionized the way I do my own sort of work and, you know, work for the company.

And if it's helping me, I'm sure it's going to help thousands of people as well. It really is a revolution that's just about to start.

Kieran Gilmurray:

And what I wonder about is, Tony, that keyboards are not the most natural thing. We weren't born tethered to one, although we've probably spent our careers tethered to one. I often think that language is the most natural thing in the world. And therefore, now we've reached an evolution with computers where we can actually have that natural conversation, and it feels more human, and it just feels like a better way of doing things.

What are your thoughts?

Tony Nudd:

100%! Although I've been in technology all my life, my degree was actually theology. But I'm a technologist. I did electronics before theology, and I've done computing afterwards. And I had to switch my brain from logical, you know, math's type thinking to something totally different. And what I found was my creativity when addressing theological problems really accelerated when I was talking to my student colleagues about subjects, and that's really where the creativity came from.

A lot of people are saying creativity is now dead because of these large language models. We don't have to think any more. There's no reason to create. But that's because, I think, we've been limited by the use of the keyboard, and now we are free from the keyboard. We can start using voice. I mean, remember that now I'm not a historian or a psychologist, but I'm thinking that language developed over at least, you know what, maybe 50,000 years. Keyboards have been around for, what, 50 years?

So, like you said, Kieran, conversation is embedded in our brains, it's in our DNA, it's how we communicate, it's how we got from, you know, inventing the wheel all the way up to inventing the Internet or creating the Internet. So, voice and language are key to creativity. And I truly believe that if we can now ditch the keyboard and we can actually start conversing with the AI, we can actually start really creating and beating AI at its own game, because that has taken us to a creativity level that we might use the AI for. And then we can stand on the shoulders of that creativity and create even more.

And I believe conversation is key to that.

Kieran Gilmurray:

And Conversational AI is enabling it, which is fantastic. And tell me, where and how should companies get started with Conversational AI technologies like DRUID?

Tony Nudd:

So, there are a number of different ways: If it's very new and you've never had any experience of chatbots or any of that sort of technology, my personal recommendation is to start with something very simple: It may be an FAQ chatbot. It may not be as sexy as a Conversational AI deployment, but it's a start and it will help you get a feel for what's capable of. You've then set the architectural foundations and I guarantee within six months to a year, you will have moved on from that simple FAQ chatbot into more conversational flow designs.

With a really good interface and at DRUID, this is really drag-drop and configure; there's very little coding if any at all and it's very visual. And so to be able to create these conversational flows and the DRUID platform, change them together as well. So you create these intents, which are basically what to do if a certain phrase is found, and then by pressing the train button as if by magic, this conversational ability takes over. And when it recognizes a particular phrase, it will jump to whatever flow that that phrase belongs to. And it really is creating conversations with technology, and if you've got RPA robots at the back of that as well, even more powerful.

Kieran Gilmurray:

Yeah, that's a bit I like: the combination of the two where Conversational AI extends the functionality and the bolts, and the bots extend the functionality of Conversational AI.

Tony, you've mentioned a couple of things there, you know, designers and flows and dragging and linking items together.

What should companies look out for when they're choosing a conversational platform? What are the things that need to tick the box to allow them to create these great conversational and productivity experiences over time?

Tony Nudd:

Yeah, I mean, certainly, a graphical design interface is needed. Some chatbots still require scripting, and it's very, you know, lots of, I wouldn't say coding, but lots of scripting line by line, it's not very visual, there are no diagrams, it's very sequentially written and not that powerful but you would have to have a lot of experience to get the best out of them.

Really what you want to look for is a graphical interface, preferably something that supports a library, so you're not having to redesign all the time, you can leverage what you've previously built.

And also, organizations like DRUID provide readymade frameworks, so that if you are going, you know, I need something for HR, there's a ready-made, and it may be 60 to 80% fit for purpose, plug it in and then you just spend your time just tweaking it and configuring it just to meet your particular needs.

This is why with the right platform, you can actually get going within literally days rather than weeks or months with a more traditional coding style approach.

Kieran Gilmurray:

It is a difference. I'm delighted to see low-code – no-code tooling because technology shouldn't be complicated. We've all the technology we'll ever need in the world to some degree, although I'm constantly surprised, particularly with Generative AI since it's come out. But we've all the technology we need in the world to get things done.

With that in mind, Tony, what should companies do or what would you advise in order to get an ROI out of the conversational platform? Because ultimately that's kind of what they're meant to be doing.

Tony Nudd:

Let's start with something simple, and obviously, often, the simple things are the most effective. So we have a very large supermarket chain in Australia, for example, and during COVID, they went from bricks and mortar to having a very strong online presence. And the first thing they had was an inundation of calls coming through, going, “Where's my order?” And also “When is the store open?” “How long does an order take to deliver?” etc. These are general FAQ-style questions, and that's what they started with, and that reduced the calls by about 60%, just by having an online FAQ chatbot that could answer basic questions.

And then phase two is, you know, “Where's my order?” That's where they linked it then to robots that went actually to the back-end ordering system to find their order and would tell them what state is in, and if it had already gone to the courier, the robot would jump to the courier's website, figure out where the package is, and then bring this back to the chat conversation as well.

So, it's bringing a very powerful capability of self-serve, and because they are also present outside of Australia, across South Asia, they are now looking at actually doing multi-language and supporting some of the Asian languages that they work in.

So, again, they started reasonably small, but the ROI was almost instantaneous on the savings of allowing their call center staff to actually concentrate on the real heavy-duty problems that customers have. And not the simple, you know, “Is the store open down the road?”

Kieran Gilmurray:

Yeah. It's interesting, isn't it? Because it's a fantastic use case, and I go back to what I said earlier on, the use cases for Conversational AI and RPA are just limited by our own imagination, that shows you what you can do.

And Tony, we mentioned earlier on ChatGPT and Generative AI, we cannot have a conversation without mentioning it these days, and I think that is going to be the case going forward.

How do you see ChatGPT in the context of Conversational AI for businesses?

Tony Nudd:

As I mentioned before, I think it's fantastic that we are able to have a technology that's able to create natural language responses, and I see that as a plus.

There is also a danger that - let's just take ChatGPT, the open source publicly available version for now- the problem with that is that it has been trained with everything on the internet. That means it's been trained with everything true on the Internet, and it's also been trained on everything that's not true on the Internet. So, the responses that you get, it's hit and miss whether the data it's getting is correct.

So, it's very easy for an employee or if you've got a customer that's kind of linked to ChatGPT as well, to either go down a rabbit hole, that they don't really need to -because it's so open and varied on what it can do- or it could actually be providing information that may not be sensible or correct for that user.

And this is where the conversational flows really come in, having that first line Conversational AI layer before we get to ChatGPT. And I would call that the Conversational AI, the guardrails.

So you can actually have your conversational flows making sure that: a, the user, whether that's a customer or an employee, is asking the right questions and not deviating and going down some sort of rabbit hole you don't want them to, and also we can do a check of the information coming back that's going to be either spoken or displayed to the to the employee or the customer. And just do a final check to make sure it's relevant information, and maybe we need to do a quick check to make sure it's true information as well.

So, while we have the power of these large language models, whether it's ChatGPT or Bard or Claude or whatever you want to plug in, I really see Conversational AI being the guardrail or the guardian of either keeping the conversation short, but also, you know, being the guardian of truth as well. And I think that's very important.

Kieran Gilmurray:

I really like that. I really do, because I think we shouldn't believe everything the computer tells you, that that's an age-old part. But if you can protect people by providing those guardrails, fantastic as well. There is a wonderfully secure version: Microsoft Azure Open AI; It is a really secure version that would block some of the stuff that comes out on the Internet as well.

And I love what you're saying there. It's not about replacing people. It's about putting humans in the loop, and putting the points where they can add value and allowing technology, the moments where it can add value and take away some of those more mundane stuff.

Tony, what does the future look like for Conversational AI and RPA in this wonderful and exciting Generative AI world?

Tony Nudd:

So, something that I've been sort of honing up on is the concept of what's called the multi-agent framework. Now this is where you can basically have a single sentence, a single phrase, and the example that the group that created the first framework basically said that, I think that it was an asteroids game on my computer, and that was great. Now, what happens is that that single phrase is then sent to this multi-agent framework.

One of the agents could be the coder. And so it's using, you know, ChatGPT actually to create the code for this simple asteroid game. Another agent in the mix could be the finance agent, and that will be telling you possibly what your sales could be for this game or it could be what the costs are to create the game to a production level. Another agent might be your marketing agent and that would be then analyzing who your, you know, target market might be. Another agent might be, you know, advertising. And so by just having that one phrase, you are then able to call on these agents and their unique focus on a particular subject.

Now, this is good for today. I mean, you can see that happening, how that would be good today. But imagine in the future where these agents are not owned by the company or a particular conglomerate, but actually individuals own that agent. And I could be an independent consultant and my specialty could be, I don't know, Lego- building things with Lego. And for whatever reason, Company X needs a Lego building expert, I could then provide my learning model that I've trained with my brain and put all my information into that model and it could be plugged in and I could be hired for that model for the period of time they need it, then I can unplug it and move it on.

So, it could be a very interesting concept, where actually we're bringing back ownership of our IP and we're not just losing it to this sort of big mash of Microsoft and Google and losing control over what it is unique about our offering.

Kieran Gilmurray:

Oh, wow! Wow! That sounds to me like almost like an equal system of value where the end user, in other words –us - obtain or abstract some of this value as opposed to big companies maybe making billions and billions Which just shows you the art of the possible, and goes back to what we were saying earlier on the beginning of the talk, in the middle of the talk, Tony.

Our life ahead with Conversational AI and RPA will be limited by our imaginations only.

There are two technologies that I adore. The fact that both of those technologies can combine to be even more effective in a Generative AI world just brings me unending joy and it makes me excited to see what the future is.

I see, and you and I have been in these technologies for the last number of years, when you see the progress that has happened in the last two, three and four years and you can imagine the progress that is going to happen in the next two, three and four years ahead; We live in truly exciting times and I can't wait to see what unfolds.

Tony, thank you so much indeed for sharing your expertise and knowledge with our DRUID talk audience. We really appreciate it. Thank you so much!

Tony Nudd:

Thank you, Karen. It's been an absolute pleasure, and I look forward to watching your next guests in the second series. Thank you!