In this episode of BTG Insights on Demand, Vivek Garg—an independent expert in intelligent automation and artificial intelligence—joins BTG's Rachel Halversen to discuss the art of intelligent automation and how advances in artificial intelligence are helping top companies build and scale efficiency-boosting business solutions. Together, they'll explore the evolution of intelligent automation, sectors that are ripe for implementing intelligent automation, and solutions to common challenges companies may face on their intelligent automation journey. Listen to the episode, read our lightly edited transcript, or jump right to a specific section of the chat below:
Interview Highlights:
- Intelligent automation vs traditional automation
- Benefits of intelligent automation solutions
- Adjusting culture to adopt intelligent automation
- A look at the intelligent automation landscape
- Best practices for implementation
- Understanding technicalities and challenges
- Ensuring data quality
- Identifying KPIs
- Scaling intelligent automation
- Innovation strategies and recommendations
- Industry examples
- Innovation breakthroughs
- What to watch out for
Rachel Halversen (01:10):
Hi Vivek. Thank you so much for joining us today.
Vivek Garg (01:13):
It's a pleasure to be here, Rachel.
Rachel Halversen (01:15):
So, by way of introduction to our audience, would you mind providing a quick overview of your background and how you've helped companies with intelligent automation and data science and AI initiatives?
Vivek Garg (01:25):
Sure thing. I have straddled business tech, data, and automation across many different verticals. Most recently in the healthcare, biosciences and retail sectors. Going back 20 years to my graduate work, which was actually on what we know of as self-driving cars today used to be called Automated Highway Systems. So, you can see there's a common thread of automation throughout.
Most recently, I've been focused mostly in very large companies, so Fortune 100 companies in the strategy and governance and analytics of automation and intelligent automation. And as you can imagine, large companies are where the problems are the largest. The scope is the largest and the complexities and the process, the silos are the most. So, that's where most of my focus has been.
Rachel Halversen (02:27):
Thank you for doing that. So, would you be able to define intelligent automation for us and explain how it differs from traditional automation?
Vivek Garg (02:36):
We associate automation, especially in the 20th century, mostly with manufacturing processes, with auto plants, chemical plants, et cetera, et cetera. And then with the advent of computers and the internet, we started to see processes being developed around applications and software as a service. And as processes got wrapped around them, there continued to be a lot of human touch involved in these processes. And that's why large companies, this is so much more relevant for large companies.
And so, in the beginning of 21st century, we started to see a lot more automation of knowledge work. Some might call it white collar automation as opposed to blue collar back in the 20th century. And so, you can see it's already evolving into the software realm. And a lot of our economy today, 70, 80% of our economy is actually service based or nonmanufacturing. And so, it makes sense that that's where the automation happened. But the kind of automation that happened was still focused on human work and knowledge work.
In the last 10 years, more and more we've started to see machine learning and AI. And of course, over the last two or three years, we are seeing a complete explosion of large language models and the impact of that. So, there's automation and then there's intelligent automation. The differentiation I would make there is the intelligence on top of execution. So, there's the execution of robotic automation, et cetera, and then the intelligence that drives or tunes or parameterizes the automation, the execution of the automation.
Rachel Halversen (04:38):
That makes a lot of sense. Thank you. So, what are some of the key benefits of implementing these intelligent automation solutions in business processes?
Vivek Garg (04:48):
So, the most straightforward thing to say is cost reduction. But there are so many more nuances to implementing automation beyond just reduction because there's one part which is cost reduction, another part is enabling innovation. So, staying with cost reduction, there is reduction of labor, there is reliability of the work that automations provide, the ability of them to work 24/7 as opposed to humans who have variability in performance and expertise.
Along with that, you have to train them. You have to train them over time, not just when you're bringing them on board. And that's what speaks to the reliability of the automation. And speaking about enabling innovation or how automation frees up people it's freeing labor to do higher level work. And typically, in organizations, this is a little bit sidestepped. We want to not talk about attrition that automation might bring, instead it's about reskilling. But then reskilling should be addressed in a more comprehensive way. And part of it should come from leadership, not just saying that we will reskill, but it's really painting the picture going forward with what reskilling actually means.
Rachel Halversen (6:30):
So, going back for a second, one of the things you mentioned was the importance of leadership leading the charge on reskilling and that calls to the mind the need for a real culture shift within the organization to successfully adopt intelligent automation. Could you speak to that and how companies drive that culture shift?
Vivek Garg (06:47):
Yeah. So, culture shift is a very dense problem. Companies think, rightly so, that center of excellence or governance program and councils that involve leadership will drive embedding of cultural change, intelligent automation into the individual employee. But as you can imagine, this happens at many different levels because process automation, processes themselves are of very many scales. There could be a desktop process that only involves one employee or it could be across different functions.
And so, embedding the right fit of automation or cultural change has to be thought of a lot by the governance program as well as by leadership. I would start from the bottom. when we will think intelligent automation has really sunk into the culture of the company is when at an individual level, one is thinking about how much time am I wasting on something that I shouldn’t that could potentially be done by something that's automated?
So, on an individual level, if you are doing what we call process mining for processes. First, you try to figure out how you would automate your pain point and that frees you up to focus on something that's more important, more higher value. And of course, from the top level, the governance programs setting up KPIs… Having large singular KPIs often don't tend to be very effective because what you don't understand is the segments of it. And so, when I think of KPIs, I think of a network of KPIs. So, KPI trees.
And more and more, since AI is really transforming how we work, we should imagine a world in which we are involved in less and less execution. We are involved a lot more in the strategy and the design. And leadership will also have to become more data savvy and ask deeper questions and not just take KPIs for what they are, but ask questions that traverse some of that KPI tree that we talked about.
And on the other hand, as I said earlier, painting the vision of where we go from here, driving a culture throughout the organization so that managers, senior leaders are driving this into their organizations [and asking] what can they do next? So, just to give you an example, somebody who is doing marketing analytics can now, instead of having to go search for data, put it together, do data prep, et cetera, do the analysis, now imagine the analysis is at their desktop. And now they say, "Okay. So, which other markets now become available? Which other customer segments should we focus on?"
The job of leadership, the responsibility of leadership is to paint that vision and make them comfortable and actually excited and inspired to move to the next realm.
Rachel Halversen (10:24):
Thank you. That's great. From your perspective, how has the landscape of automation and AI evolved over time, and particularly in the last few years since AI has taken a quantum leap forward?
Vivek Garg (10:36):
Right. Absolutely. In the last 10 years, a lot of automation is still robotic automation, because ML, AI applications tend not to be connected to automation as much. However, now there's a convergence happening of both connectivity between processes and AI, ML experiments that may be happening in pilots or they may be use cases that are being implemented, but they're not connected to the entire end-to-end processes that exist in companies.
And often, there are a lot of processes that have already been automated with robotic automation, but they haven't leveraged AI and machine learning appropriately or at least at scale and sometimes not at all. So, going from the last 10 years to the last two, three years now suddenly, we've been forced to become aware of large language models and the impact and the potential impact they will have on every aspect of our lives as individuals as well as in enterprises.
Immediately you can imagine large companies--there are scores of pilots already happening where they're trying to figure out whether there should be one monolith large language model for the enterprise, or there should be a few large language models that drive different functionality that are sort of ring-fenced. This is what will drive connectivity between the intelligent automation that we today know of to the AI that is exploding right now.
Rachel Halversen (12:29):
That makes a lot of sense. So, beyond the culture shift, what other best practices do you recommend for companies seeking to implement or improve intelligent automation within their business?
Vivek Garg (12:40):
I think structurally, as we touched upon a little bit, governance is very important and governance not in the sense of project management. Typically, governance ends up being project management. However, I think governance works well when people who are driving governance are actually domain experts.
They understand automation, intelligent automation. They understand the technologies. And analytics is going to be very important, not just people who do analysis, not just people who are managing products, but leadership as well as project management.
And so, the more governance itself is data driven in ensuring that we are focusing on the right things and we can actually track the performance of it going forward. And that you cannot do if you're not knowledgeable enough of the domain. So, having this center of excellence with domain knowledge--domain expertise--is important.
The other thing is, companies oscillate between a federated model where the functions are independent and they're driving their own ship--which they should--as well as a very centralized model where the governance—and whether that's governance or it's leadership—that's driving a lot of the decisions for the functions. However, in my experience and probably what makes the most sense is where it's a semi-federated model, where there is ownership within the functions and of course, execution, decision making.
However, they work with a center of excellence, of governance organization, that is collating best practices across the organization, but understands the domain well enough to map into what the functional needs are. And when that connection is done well, that's when you can see culturally the organization changes as well as better and better prioritization happens of automation projects across the board.
Rachel Halversen (15:22):
So, from a more technical standpoint, how does the integration of AI and automation technologies work? What challenges can arise during the integration process?
Vivek Garg (15:30):
Depending on how large a company is, there may be hundreds, thousands of such initiatives--automation and intelligent automation initiatives--across the company. You would start with the scope of the current automation and the underlying process that it's automating. Where does it make sense for the automation to be extended beyond? So, whether it's just limited or it has potential to connect to other processes to become a lot more end to end.
Any process that has a potential for automation has underlying data about the process. And if it has underlying data, it has a potential for machine learning, it has a potential for AI, because you always want to understand your processes, you want to characterize them well, predict things that are important so you can plan against it. And so, almost every automation has a potential of AI.
Now, the integration challenge really happens where you've built automation on a set of data, whereas the AI models, or if we’re talking about a large language model, the large language model may be working with a different data set. There may be overlap, but they're still not completely consistent in talking to each other.
So, the challenges would lie in the cross inferencing that would be done from the AI part of the automation with the execution. And that's where things like human-in-the-loop, becomes important because you don't want to take your AI inferences blindly. You want to validate them, you want to align them to your application, and see if they're effective. And this is a continuous experimentation effort .
And so, yeah, the challenges will be in the data itself, the underlying data, whether the data itself is co-located, they're talking to each other, is the data relevant for the automation, for the AI. And what level of availability of data is to the various processes?
Rachel Halversen (18:20):
That goes into the next question really well, which was what role does data play in intelligent automation and how do you ensure data quality and availability for successful implementations?
Vivek Garg (18:30):
Yeah. Great question. Yes. Data is really the foundation of automation. We talk about AI, but data is the blood running through the body. It gives connectivity to all of your processes. If we start to trust the execution and the models, then we're dependent on the quality of data being high.
And that goes to governance of data, not just of intelligent automation. The objective of data governance is that you want to have someone who is a domain expert. And that domain expert could be an analyst, could be a manager, could be a robot. It could be a bot.
So, you want to make the data that's most relevant available to the domain expert when they need it, in the form that they need it, in the quality that they need it. And for this to happen, the underlying data strategy should be, well thought out and built up. The foundation of it should be built up well to empower both automation as well as the machine learning and AI models. So, data is critical to the automation.
Rachel Halversen (19:58):
And earlier on, you were talking about KPIs. So, how are some ways that you can measure the success of an intelligent automation implementation?
Vivek Garg (20:07):
This goes back to all the benefits that we were talking about of automation. You obviously want to measure the benefits. And the first one that one would think about is the value that automations provide. But as I also said that the benefits of automation are all multifaceted. They're not just cost reduction. They're innovation enabling. They're user experience improving. They are consistency improving, reliability.
As organizations become a data-driven culture, there should be tracking off these benefits and not just stop there. As I said, going back to the KPI trees, each one of them should have their own tree underneath so that we can get to the root causes of what is driving the KPI to be high or low and take corrective action.
So, there'll be KPIs within a process and across the process. So, for example, within a function, take a look at the important processes that the function is built upon and what percentage of them have been automated. We could look at external benchmarks where in the same vertical, across verticals by function and look at what level of automation has happened outside and what's here. And then you could look at—speaking of data and AI and connectivity between automation and AI, to what level have our processes been automated? Are they using intelligence? Which is the machine learning and the AI part. And to what level is our data itself being leveraged for automation?
And so, designing KPIs is important. Tracking them and executing against them, adapting our strategies to the KPIs is important. And also, what's important is understanding and alignment of the definitions of KPI, not just the KPI. Because as the KPIs become more involved, what the conditions are, what the context of the KPIs are need to be well understood and aligned on by the people who need to be aligned, which again connects to the overall governance program, data governance as well as automation governance.
Rachel Halversen (22:54):
A lot of these KPIs seem to speak to the scalability of intelligent automation. So, how do you approach the question of scaling this automation as businesses grow and evolve?
Vivek Garg (23:04):
Yeah, exactly, connecting to what I was just talking about adoption, scalability is again, very automation dependent. Sometimes you could see a connectivity between different processes and they become end-to-end and the total may be larger than the sum of the parts.
And so, that's where again, a center of excellence becomes critical in understanding the domain because some group may make a decision that if we connect this process to another process, there will be too much friction in between. And hence, we need to have two different automations that address each of these different processes.
Another view of that and a more abstracted view of it could be: you look at your functionality and see what percentage of the overall sum of the two processes has an end-to-end aspect? And so, that could be one single automation. And then you could have scalability for the leftover.
So, getting a little bit into the nitty-gritty here, you can splice automations such that there are these long pipes of end-to-end automations that really enhance the value of automation throughout the enterprise and will benefit from connectivity, will benefit from AI, from large language models because now, you have data that's being leveraged end-to-end in the automation.
Now you are able to connect a section of the envisioned automation end to end, and for the rest, you can scale as needed or as the potential exists.
Rachel Halversen (25:09):
Makes sense. Do you have any strategies that you recommend for organizations looking to continuously innovate and optimize their intelligent automation initiatives?
Vivek Garg (25:19):
One of the things we didn't touch upon so far is what AI is enabling us to do. And we talked about it on a level that is the execution part and the intelligence part. When we think about automation today, we think about what is the common case, what is the highest value place to place our automation? So, in other words, a lot of cycles, a lot of time spent somewhere, you focus your automation there.
But really what AI will start to enable us to do is to actually focus on the long tail. And the long tail can be really long, but it is, of course, small enough that you don't want to put targeted solutions there, targeted automations. However, if the large language models are based on a foundation that has all of the data, a lot of the custom work that we think we still have to have a human-in-the-loop for, now may be done by AI. Of course, I'm not saying all of it goes away, but you can see 40% can become 70% very quickly.
That speaks to driving innovation in the organization. It's the part of intelligent automation being an enabler is now since you're really hitting the long tail and you can build custom-targeted solutions for things that you wouldn't have targeted earlier because doing machine learning experiments, building implementations, that's expensive and it takes a lot of resources, takes time. But if large language models are informed with all of the data, you'll start to see that 40% climb up.
That's where a lot of innovation will happen; you can fine tune your product using data that is local as well as semi-local and global. So, the amount of innovation that even a very traditional product can make enabled by AI is going to be tremendous.
Rachel Halversen (27:39):
When and how do you identify when humans should be involved?
Vivek Garg (27:42):
Well, again, it is automation dependent. It's based on the amount of experimentation the reliability or the trust that we have in the models that have been developed. Think about intelligent document processing. All of us have gone to an ATM, tried to deposit a check, and then we've been prompted that it couldn’t decipher, it couldn't figure out what this number is, enter it. That's human-in-the-loop.
The parallel of that in an enterprise is when the automation accuracy or the confidence is not high, as long as it's not a complete hallucination as far as the AI model is concerned, if it knows when it is high confidence versus not, then you prompt a human to make decision in between. And sometimes, it may not be simply a decision, it could be connecting different parts of a process just because it is not integrated currently, the different applications are not integrated today, or they cannot be integrated or they're different platforms, et cetera.
So, there is always going to be places where connectivity is required. There's places where trust in the algorithms is lower. There is, of course, the whole part of exploration and experimentation that has to be done by humans in even deciding on an algorithm or a strategy. So, yeah, humans are going to continue to be in the loop.
Rachel Halversen (29:34):
That's a very good thing. Do you have any examples of industries and sectors that can benefit the most from intelligent automation? Just to zoom out for a second.
Vivek Garg (29:44):
Yeah. So, let's zoom out even more, even beyond sectors. If we were to simply look at large companies and small companies irrespective of the sector, there are functions within large companies that may be perfect for automation.
I'll say backend processes of any company, even if product development and R&D is very up to the mark with automation and AI, other supporting functions are often not. You could simply look at the number of employees in each of those functions and have a sense of how much human touch there is and what's the potential of automation there.
So, irrespective of sectors, large companies--especially companies that have been in existence for more than 20 years--they'll have processes that have been established over long time, sometimes created a lot of silos across the board. So, that's already sort of in my mind low hanging fruit for execution level automation, so robotic automation there. And of course, everything else you can layer AI on top of that.
On the other hand, small companies, and I'd say, you know, even startups, a lot of challenges they face is resources. They may have a great product idea and they need to continue to work to improve the product, but they are undercut, shorthanded by the lack of resources. Even software development. Forget about social media and marketing and sales, et cetera, but even software development, you can see how AI is changing that because now you get a lot of help from AI with pilots, co-pilots for coding.
And so, the enabler part that we talked about earlier, not just cost reduction, the enabler part automations are going to be enabling small companies and also project that into small groups within a large company. So, even putting aside different sectors, you can see how if we just break them down by size, we can identify lots of places where there's opportunity for automation.
Speaking specifically about sectors, pharma, of course, healthcare is huge because there is a lot of moving parts, sectors where there's significant complexity of supply chain. Wherever there are lots of people in processes, intelligent automation becomes important. Retail, especially physical retail. Again, they're built on supply chains, people. Even industries like insurance, even though a lot of automation has happened there, but they're working with a lot of data and hence, especially the AI part of it will become more and more applicable.
Rachel Halversen (33:07):
Can you share any other examples of instances where intelligent automation not only improved efficiency, but also led to innovative breakthroughs in a company?
Vivek Garg (33:15):
In pharma, all the large companies today are using AI to analyze billions of compounds to speed up the process for chemists to do drug discovery. Healthcare services, there's one that comes to mind, Pager, which looks at the gap of healthcare treatments for individual patients beyond looking at clinical data, looking at claims data and connecting the dots for them for minor ailments for which you may not go to a physician.
That's just one example of healthcare, but there's innovation all around us. Everybody knows ChatGPT, et cetera.
Rachel Halversen (33:59):
What are some other trends that you foresee in the future of intelligent automation and AI? And how can businesses prepare for them?
Vivek Garg (34:06):
So, one major trend we already talked about is the long tail. Now, the word "custom" is not going to sound so scary. Because a lot of intelligence that we associate with custom projects, custom problems, corner cases is—when there is enough context that has been provided to the AI models—the hope going forward is that they will be solved by AI targeting custom problems across the board.
With large language models, what you'll see more and more is applications on top of it. So, applications that become a lot more--again, going back to the custom and the targeted part for an individual--having a personalized tutor for very many different aspects. If you inform them with your data, then you can expect to get a high level of advice. But that's just at an individual level. You can think about that within an enterprise in small projects, large projects, et cetera.
We all know large language models, generative AI is basically next word prediction with a lot of context--with a whole lot of context behind there. Imagine where that would go next: Instead of just next word in terms of text, companies are already working on the video version of that. But also, strategy generation. So, next steps as opposed to next word. And you can see how now we're not thinking about execution, we are thinking more about strategy and next steps and innovation on top of that.
I think the promise of intelligent automation for global economy is so huge because all the problems that we don't solve today—by freeing up capacity, now we can hope to solve. And I'm talking about things like, large things like poverty, like education.
Corporate training, that is a huge area of innovation. And not just corporate training, education in general, to educate employees within the company as well as myself, knowing if a model knows what I know, what my background is, and not just my resume, but a lot of other things, my life really to be able to provide what are the next steps, speaking of next steps, this is a next step. As a human being, what do I need to learn next to get to where I want to go?
And you can imagine what the role of education at the very, very basic level for people will be in terms of eradicating poverty and tapping into opportunities, climate change. And these are all large problems that I think automation is going to really be able to target and address over time done well. Yeah.
Rachel Halversen (37:40):
That all sounds very exciting, and it'll be really interesting to see where we go from here. So, thank you for joining us, Vivek. It has been really insightful speaking with you today.
Vivek Garg (37:49):
Yeah. This is fun. Glad we could take time out from our daily lives and pull up a bit. Yeah. Excellent talking with you.
Rachel Halversen (37:59):
As a reminder, our guest today has been intelligent automation and AI expert, Vivek Garg. And I'm Rachel Halversen for Business Talent Group. To start a project with Vivek or thousands of our other highly skilled independent consultants, visit businesstalentgroup.com or subscribe for more of our conversations with on-demand experts and future work thought leaders wherever you find your podcasts. Thanks for listening.
GET THE SKILLS YOU NEED
Thousands of independent consultants, subject matter experts, project managers, and interim executives are ready to help address your biggest business opportunities.
About the Author
Follow on Twitter Follow on Linkedin Visit Website More Content by Business Talent Group
