| If you don't have an account, Register here |
May 26, 2023
Mihir Shukla, CEO and Co-Founder at Automation Anywhere talks with Phil Fersht, CEO and Chief Analyst of HFS Research, at the HFS Horizons Summit-NYC on May 17, 2023.
You can listen above or watch this HFS Videocast here:
You can also listen on:
• Apple Podcasts
• Spotify
• Google Podcasts
• Anchor
This transcript was auto-generated from the original recording and lightly edited for readability. We've done our best to catch errors, but names, technical terms, and company references may be misspelled or imperfectly captured. For the definitive version, please refer to the original audio or video. Views expressed are the speakers' own.
We are back together on the stage again. I think I should have brought out a photograph — we were together just before the pandemic, back in the heyday of RPA. I think I had you and Alastair Bathgate at the time — remember Alastair, who was the CEO of Blue Prism — and I had to try and get you two to say nice things about each other. I remember that you succeeded, almost. Well, he’s retired now, so he drives a minivan around Manchester. So anyway, it’s great to have you back. So how have things evolved? I mean, it’s been a crazy few years, but how have things evolved for you, maybe in the business that you’ve been in, in the last maybe four years?
First of all, great to be here. Good to see you and so many familiar faces — always great to be in person. It feels very, very different than four years ago, in many ways. If you go back four years ago, we all were trying to prove that what we were doing was worthy of CEO and board attention. We may have succeeded in a few places, but at large, I don’t think four years ago most CEOs and boards cared about automation. It was a means to an end. Today it is entirely different. I was at the World Economic Forum in Davos earlier this year, and every CEO stopped me to talk about automation and AI. It is a board-level topic. Heads of state talked to me about automation and AI. And as we speak, the European Parliament is debating the AI Act. It is front and center. Who would have thought? It is different if you look at it just from an automation perspective. Back then we had three different offerings; now we have nine. RPA is just one of them — a very important one, but there are another eight. So it has expanded into a much, much bigger platform. I am surprised that today we run 150 million bots on our infrastructure. If you had asked me four years ago, I would have guessed 30 to 50 million at best — 150 million bots in 16 data centers. It feels very different. And if anything, gen AI changes that completely. We know about a few gen AI models, but I think there are at least 100 purpose-built generative AI models coming to the market that are designed for very different outcomes. There is a rumor that Automation Anywhere is announcing one soon that is purpose-built for automation. I think it’s a new ballgame.
That’s good — you’re starting your own rumors. I love it. You know, it’s interesting, you talked a bit about Davos and the attention this is getting. I remember 10 years ago, we had to watch Mark Zuckerberg in front of Congress talking about, you know, ‘you sell our data.’ Now it’s almost the opposite, where we’re suddenly seeing tremendous potential value from emerging, especially generative, AI, and it’s getting this crazy scrutiny from the general public who don’t quite understand it. Do you think we’re going to go into an era where we’re going to have more government intervention and oversight, and that’s going to actually pull back progress? Or do you think there’s going to be a broader acceptance of where things are shifting?
I see it in two different ways — and again, none of us has a crystal ball. On the enterprise side, I think it is fair to say that out of 5,000 of our customers, at least 70% are talking to us about generative AI. I’m personally aware of 200 use cases already working on generative AI, so it’s very real. We work with multiple partners there; some are public and some are not yet. The combination of generative AI and the other nine pieces of automation we have is just amazing — it supercharges each other. So on the enterprise side, there will be a lot of guardrails that we put in place, and I can share more later about what the guardrails are, but we will get that about right. There may be a few hiccups, but within enterprises we will have a transformative experience. I think the challenge is more on the society side, and that is where government will have to do something about it. I don’t mean to scare people, but think about it: the last time we used AI at scale was in social engagement models. We decided to engage in social media, so there was AI at scale. What did we get out of it? We had a mental health crisis with teenagers. We had democratic interference, election interference. In my view, it accelerated the polarization of society. This was AI just engaging us. Who would have thought a like button would cause that? Now we know. Now imagine AI generates content, and AI engages you. You have to be careful. This is about unintended outcomes in society. What do we do about that? And as technologists, this time around we can’t say we didn’t know it, like we did in the social media era. We now know it. So that is where governments will have to do something.
It’s a challenging thing. You don’t know what’s going to happen, so how do you regulate what you don’t know is going to happen? It’s tough, isn’t it? So we talk about a productivity crisis. I guess we’ve been going through this for a long time, but do you think we’ve come close to solving it?
I think the motivation is the highest I have ever seen. Let me define the productivity crisis. The bigger question is, why do we care about all of this technology? Why do we care about generative AI and automation and better, cheaper, faster? It is about a lot more than that. If you think about the entire world, all institutions operate on growth. And there are only two ways to grow: more people working, and existing people producing more — otherwise called a productivity increase. In most Western countries there are no more people working, and that’s true for the next 20 years, so you no longer have growth happening through population increase. So your entire growth has to come from productivity increase. The way the math works today, we have to produce 50% more productivity out of existing people just to hit our GDP targets. Now let me tell you how challenging the problem is. It took us computers, mobile devices, the internet, and every piece of software ever written to produce the productivity we have today. Does anybody have an idea how to produce 50% more than that? And if you don’t, then there is not enough to go around, and there are more riots on the street. When I talk to heads of state, they come and tell us, ‘do something, because there is only so much to go around; whoever does more riots gets more money.’ So what we are talking about is far more than just automation and AI and productivity. It is about operating equitable societies. And if you don’t do that right, we will have a challenging time. So it’s very important that we get that right. I think AI and generative AI and automation hold great potential. I’m sure there will be other ways — new metals and energy sources and whatnot — but right now, at Davos, when I was there, there was nothing else on the table except AI and automation. That’s what we have. Nothing else.
You know, when we started looking at RPA, we started looking at different types of technologies, machine learning around it. Obviously, we’ve talked about this very rapid emergence of generative AI — the launch of GPT-4 in particular has made a huge difference. Is this a continuum, or is this more tools in the box, in that digital toolbox of productivity and value-creation capability?
I think it is a continuum, but from an executive perspective, it has gotten people’s attention. As you and I know, not many CEOs or CXOs know all the processes. They can’t tell the difference between 50 processes automated and 500 — that happens somewhere in the back office or some other part of the world. Generative AI has captured their attention. It is the self-driving car of our world. And CEOs are able to tell — I’ll speak for myself — out of the many things I do, I’m probably averaging eight or nine out of ten. Generative AI will beat me on that. It will produce better PowerPoint presentations — not particularly my great skill. So CEOs themselves are able to see that, hey, if it can replace some of the five things I’m doing, this must be true. So now they understand the new operating model that automation and AI will drive. This is our opportunity to grab their attention and tell a cohesive story that is end to end. And that is where the power is. Let me take a call center — we were talking about call centers earlier in the session. If you apply generative AI to a call center, external studies show that you get about 14 to 15% productivity gain, because you can generate emails automatically and so forth. If you do the rest of the automation toolkits we and others have, you get about 20% productivity gain. But when you combine them together, it supercharges the operating model. We have a customer working today — take an airline example, everybody can relate to it. When a request comes in to, let’s say, cancel or change something, generative AI will process the email, and the rest of the automation technology, including RPA, will operationalize it across six systems. And then generative AI will take over and send a warm email back — all of this round trip in two minutes. 80% of all call volume, end to end, automated in the next-generation call center. This is 14% here, 20% here, coming together to produce 80%. This is the power of supercharging two technologies. So it is a continuum, but — let’s grab the attention of the leadership. This is the moment.
Yeah, I mean, we completely agree. The impact on CX and the call center is probably going to be the most impacted, because of the capabilities and the sheer cost of customer service for businesses, which is very high. When we got all excited about automation back in the day — and we still are — this was about taking humans out of loops they shouldn’t be in, getting people out of stuff they really shouldn’t be doing, so they can focus on more interesting things. Gen AI is about bringing humans back into loops where we can take advantage of their decision-making skills and their prowess and their judgment. So where does this ultimately land, and how does this ultimately affect people and the skills that we’re looking for for the future?
There is a definite need to have a human in the loop. Some of you have seen this in a recent news cycle — that generative AI tried to convince somebody last month that it was the year 2022. The generative AI was convinced. So there’s obviously a need for a human in the loop, and there are many such examples. A validation, a guardrail, is essential, and most of our implementations are heavily focused on guardrails today. So that is true on one end. On the other end, it is also true that I have personally seen a case of a junior banker in a large bank, where this person is excited to be on Wall Street’s ideal investment-banking route, and is seeing 70 to 80% of the job getting automated. I have seen it in Silicon Valley with programmers. I’ve seen a non-programmer use generative AI to produce code, and a 20-year veteran programmer said, ‘how about you do this?’ and the answer was, ‘yeah, I already did this.’ ‘How did you know? It took me 20 years to learn.’ ‘Yeah — generative AI.’ And I’m looking at this person’s face after 30 minutes, thinking, what happened in six months? I had 20 years of experience programming things. So that’s also real. It has a need for human beings in the loop; on the other hand, it will change many of the jobs as we have seen them before. One thing I’m certain about: you always overestimate technology in the short term and underestimate it in the long run. In the RPA era, and other times, people had fears about jobs going away, and people have similar fears now. But one thing is certain: a human will definitely lose a job against another human working with automation and generative AI. That much is certain. And that applies to every knowledge worker, including a CEO. If another CEO and his team are working with generative AI and automation, this CEO will lose his job. That is clear. One more thought — based on things I have seen, I think it is real enough. Do we know everything? No. But I think there’s a cost to knowing everything, which is that you’re too late. So in my opinion, working with multiple generative AI models and reimagining a new operating model is a no-risk decision at this point. It is in the right direction. Although we don’t know everything, we know enough to do something there.
Interesting. I was talking last night — someone was mentioning there was this machine driven by generative AI that had been taught to program itself. So we started talking about Skynet.
So, about that: one of the things we were discussing with European governing bodies on the AI Act — and one of the things we suggested — is to tag every piece of generative AI content with a tag that says it was generated by generative AI. Because imagine, there are 100 more generative AIs coming, each one generating a gazillion amount of content, and each one consuming each other’s content — you can make it believe anything. So maybe one approach is to tell other generative AI not to consume other generative AI’s content, because it has a tag in it. So there you go — it’s one way of doing it.
And I mean, with GPT-4 they can tag where the content comes from, which is important. But then there’s also stuff like — someone the other day was doing a search on me, and it turned out HFS was founded in 2006 by one of my former employees. So he must have had it on his CV somewhere and it was floating around.
So, but what you’re pointing out — like it says Automation Anywhere raised capital in 2015, when we raised it in 2018. Now, this is not a hard thing to get right, and so there are still obvious things that are just some glitch in there.
Absolutely. You know, you talked about a rumor — things may be changing in a couple of months — and I’d love to know, who are going to be the winners and losers in all of this? You’ve founded a number of software companies, I think, prior to Automation Anywhere, and Automation Anywhere has been a major driver — I think it’s been the most downloaded automation product in the world, ever. But who’s going to be the winners and losers as we look at where things are shifting now, where the money is coming from and the investments are coming in?
I think these vertical, purpose-built generative AI models are going to be an exciting arena. I mentioned lots of things that we are doing already in generative AI, but the future of generative AI is doing things, not just chatting. We’re going to announce it on the 1st of June, and I won’t reveal everything, but what we started thinking about 6 to 8 months ago is: is it possible to train generative AI models to tell generative AI to locate delivery receipts, match them against invoices, and if it is less than $1,000, pay it? Could generative AI execute this statement? Now, this is far more complex than generating Python code, because what I just described goes across three systems, and this kind of thing is at the heart of what we call enterprise processes. And as you can imagine, we are in a unique position with 150 million bots running and working with about 15,000 applications. So, arguably, we know more about doing things than anybody on the planet. As we started feeding some of these things in, it is exciting to see how these models are emerging. You can leave it at that, but those kinds of models — and again, they evolve fast; imagine the difference between 3.5 and 4, they evolve at an exponential rate — those, in different industries, are a very exciting arena to watch. I think in the partner community — our view is, and you probably have an even more informed view, Phil — partners who are doing large-scale operations but have very low domain knowledge will have to rethink their operating models, in my view. And partners with vertical domain expertise have a significant advantage if they can leverage this. On the customer side, customers who are reimagining their operating models are going to be winners. Let’s look at the entire IT industry: we spend about $4.5 trillion on IT worldwide. Most of it does not increase any productivity — sorry if you’re in IT, but it does not. We keep the lights on and keep upgrading systems and running projects, lots of IT projects, but it just keeps the world in the same place. So what businesses are doing is asking, can I take 20% of that budget — that’s about $1 trillion — and put it in this top layer, this experience layer that we talked about, what we call a system of work? A system of work is a set of generative AI, automation technology, process mining, ServiceNow, and many other things — not just us. But that entire system-of-work template actually produces a productivity gain, so now you can take that money and put it somewhere else. Business customers who are reevaluating operating models that way are the winners. If you’re number three, you’ll become number one; if you’re number one, you’ll outperform your competition. And every single software company — I think we’ll definitely have more competition — will be in the space they have to be.
I mean, when you think about the ability to automate generative AI, it’s incredible. You could, for example, have an automated query into GPT every hour around stock performance or industry performance, that sort of thing. You’ve got that balance between humans in loops, humans out of loops, and automated loops of intelligence that you can actually control, because you’re using maybe specific large language models. You can have regular automated feeds of information that are piled in. It’s very, very exciting where this is all heading. In every industry — I was watching it in pharma, there is an opportunity of $50 billion in pharma drugs just on generative AI models. You can almost set your own agenda, not just constantly posing questions but having it automated and done for you on a constant basis. So the direction of travel is very interesting. Now, maybe to finish with: you’ve spent a lot of your life in the world of venture capital and private equity, and you know where the money comes from and where it’s going. We had a bit of a wreckage of the PE market last year, and now we’ve got this incredible traction around gen AI. Where are you seeing the next round of VC money coming from, and where do you think it’s going to go?
I think VC money is chasing, as everybody’s heard, generative AI now, like it chased other things. The problem with VC money is that everybody bets on one or two players. The problem is there are hundreds of them, so they end up betting on 200 — there’s no way for all 200 to succeed. So that’s the betting game. A part of the market always gets overhyped, but as somebody said on the first day, hype is good, because it’s about hope. So as long as you understand that, that’s a good place for it to be. I think the new operating models around generative AI are leading Wall Street investors to change their valuation models. The Rule of 40 — which is 20% margin, 20% growth — is now being talked about as the Rule of 50 or the Rule of 60. If that changes, none of us has a choice but to do something about it, because once the investors start comparing you to somebody else with a Rule of 50 — so that is where the investors will push everybody. And especially after Silicon Valley Bank and many other bank crises, the entire investment community has shifted to this higher-margin model. So that is another reason why everybody’s pushing for automation and generative AI. I can share, without naming names, that many large public companies are talking about a world called extreme automation — which is to automate 35 to 50% of all work. I don’t know how much of that will happen; we’ll only know when we go about it. But the fact that CEOs and boards are dreaming about it is saying something. It’s all coming together.
Yeah — the intersection of generative AI and automation. It really is. This has been most fascinating. Do we have time for maybe one question? Let me put it to you on behalf of the room: Mihir, thank you for the excellent comments. As you think about your platform and bringing more AI automation as part of the platform, would you be using generative AI models in your platform, or offering that as a service for clients — or are you still not sure that’s the right place to go?
It’s a good question. We already use that today — we are a partner to OpenAI, Microsoft, Google Vertex AI, and Amazon Bedrock, although Bedrock is not generally available yet, and a few others. So internally, we have a chance to see which one works better, and in certain places one works better than the other, and everything is evolving very fast. That’s all available today — live, functioning, real-world customer examples, hundreds of them are possible. What I was talking about is the next revision of this, where we have our own generative AI model that’s able to — you can give it a request and say, ‘I want a process executed across applications.’ How much of it can generative AI execute? Not just tell me how to do it — can you do it? That is early days, I’ll set the right expectation, but you will see very soon how much of that is real and working.
Let’s take one more from the floor. Which industries are on the cutting edge — BFSI, healthcare? Of these two industries, which one is further ahead?
I think it’s about the leaders, is how I see it. There are leaders who are taking charge of extreme automation and a new operating model. I’ve seen it in supply chain, where large companies with a $50 billion supply chain are reimagining the operating model and having a $3 billion impact. I have seen oil companies do that — saving 30 cents on a barrel, which saves $5 billion, with generative AI. And BFSI, for many reasons that we know, has a huge, huge opportunity. I think it’s a leadership moment for us and for every leader — about who can navigate through this. I think on the first day, maybe it was Sanders who said that 50% of employees think that their career is in the hands of the company, which is not — my take is that it’s in their hands. But we have a responsibility to lead our teams through it, and some will and some may struggle. So it depends on the management. And I think in healthcare there’s an amazing opportunity. On the clinical side, generative AI is not reliable enough yet to risk lives with it, in my view, based on what I’ve seen — you know, my front-page New York Times article on one wrong clinical assessment. But there’s an amazing opportunity there. On the administrative side, there are a gazillion things in healthcare — claims and all kinds of administrative things — a huge, huge opportunity.
Yeah, and on the research side, I’m seeing quite a lot of interesting things happening. I was looking at an app the other day where you can actually ask for stock advice, and you could even ask about presentations on Wall Street and how they’re impacting stock — it was incredible, the amount of information going backwards and forwards that you get immediately around stock movements and things like that. We’re also launching an LLM for HFS next month, where we have all our research tied into a specific search model so you can actually ask it questions. We talked a bit about it yesterday, but we’re going to launch it very soon. That’ll change things — if you’ve ever gone to a research website and tried to find anything, it’s a bloody nightmare, right? We hope we can do this a little faster than Gartner can. Anyway, this has been wonderful. It’s been great catching up and great to talk to you. I look forward to seeing what’s going on. Thank you.
Your account has been created. You can continue exploring free AI insights while you verify your email. Please check your inbox for the verification link to activate full access.
With the exception of our Horizons reports, most of our research is available for free on our website. Sign up for a free account and start realizing the power of insights now.
Our premium subscription gives enterprise clients access to our complete library of proprietary research, direct access to our industry analysts, and other benefits.
Contact us at [email protected] for more information on premium access.