Salesforce: Combining AI and automation can give us superpowers and make us more productive
At Dreamforce 2020, Salesforce unveiled Einstein Automate, an automation platform designed to help customers automate workflows and connect applications using the low-code or no-code tools. Robotic process automation is a key part of enterprise digital transformation strategies and I had a chance to speak with John Kucera, SVP of Product Management for Automation at Salesforce, about Einstein Automate, how the company is using AI as part of its automation solution, the kinds of tasks Salesforce customers are automating, whether he believes automation will lead to fewer jobs and when we’ll be able to talk to Tableau like we do a smart speaker. The following is a transcript of our interview, edited for readability.
SEE: Managing the multicloud (ZDNet/TechRepublic special feature) | Download the free PDF version (TechRepublic)
COVID-19 effect: Rapid digital transformation means more automation
Bill Detwiler: Automation, process automation has been part of the digital transformation efforts and plans of companies for a while now. I’d love to get your take since you talk to Salesforce customers, and you’re talking to companies about those plans, where does automation figure in the mix right now? Is it work that they’re still doing? Is it work that is more prevalent now than it maybe was a few years ago? Talk a little about that.
John Kucera: Sure. And of course COVID is the big subtext. We found all of our customers, they had to do two years of digital transformation in two weeks or two months when this entire new way of working was thrust upon so many different people. And so that created this massive need and demand for automation. You had things like the paycheck protection program. It didn’t exist. It wasn’t on a bank’s backlog. It wasn’t in the IT team’s group to say, “We need a mortgage or a loan application process next week for thousands of new loans.”
And so this created this huge, huge need for our customers to think about how can we adapt to this new way to work? And so we’re so happy and thankful that we’ve been able to help them with this transformation due to the power of all these tools. And so fundamentally, I think of automation as taking away the tedious, the drudgerous, the inaccuracies from our processes, which frees people up to do the work that we really like to do. Nobody wants to be copy and pasting things on a form a 100 times a day. And so we’ve actually found a surprising insight is that automation has helped with employee satisfaction and retention. Not what some people have thought might be the impact.
SEE: Digital Transformation: A CXO’s Guide (ZDNet special feature)
Automation is about augmenting human workers and eliminating tedious tasks
Bill Detwiler: Well, let’s drill down on that because I think that’s a great place to start. That’s one of the fears that workers have with automation, whether it was robotics in the early days in the auto industry, or whether it’s automated manufacturing or it’s automation like we’re talking about more in process and behind the scenes. Talk a little bit about, expand on that, how automation is really augmenting companies, existing workforces, and is it really leading to just a reduction of staff?
John Kucera: Sure. So fundamentally we people are really good at what we do. People are the only things that can build these relationships, that can delight customers and make judgment decisions. That’s what we do every day. We figure out, okay, which direction should we go? Should we prioritize this? Or should we do that? I wanted to help make this customer satisfied and successful. I need to build that relationship so that we can have a mutual understanding of the problem, strategize, form my hypothesis of what to do and then work towards it.
And so automation, I see, as really a digital assistant that gives us superpowers. It takes away those tedious, those silly things that we don’t want to have to do. Nobody wants to go between the two different systems and move data. They don’t want to have to look between places and do those updates.
What it really does is frees us up to what we are uniquely gifted at, making hard judgment decisions and building relationships. And so we’ve seen that this has been a huge boon for our customers and that a lot of the fears that I’ve seen out there in the press are not really justified. As much as technology software people might say that it’s going to improve productivity and it does. Fundamentally, it doesn’t replace people, it frees them up to do these higher level tasks that also they like doing. So they’re happier in their jobs and retention rates go up.
SEE: Salesforce rolls out permanent remote work plans (ZDNet)
Einstein Automate: AI + low-code workflow tools + MuleSoft integration
Bill Detwiler: So let’s talk about Einstein Automate then. Let’s get into the nitty gritty of what it is, what it does and how the latest evolution of automation within Salesforce platforms. It was announced at Dreamforce in December in 2020. So give me a rundown on Einstein Automate.
John Kucera: Sure. And we chose the name because we’re bringing the best of AI, Einstein, and the best of automation. And basically they go together like peanut butter and chocolate. they’re the best bunch. And so fundamentally there’s two sets of things that we’re really, really focused on. How can we take all of those repetitive, tedious things that people are doing and have software free them up?
And to do that, you need to do this in a very cost-effective way. You need to have, basically, non-developers be able to do things that usually only developers could do, which drastically reduces the cost of it. You need to be able to integrate across all of those systems so you can have all the data in one place to make that happen. And then you need to unify these processes and move the work efficiently between people so that they can all have this concerted way to make these judgment calls, making customers and employees happy. And so that’s fundamentally our founding thesis is, how can we take away the tedious? And then how can we also unify and make people smarter in the decisions they are making to make for a more harmonious way to do work?
SEE: Highlights: Salesforce TrailheaDX 2020 (free PDF)
Bill Detwiler: And so tell me about some of the functionality that it has, that it allows customers to do. How can customers integrate it into and use it in their existing either Salesforce, architecture or find new ways to use it with their existing systems?
John Kucera: Sure. And so Salesforce customers today already have Flow. Flow Builder is this fantastic tool for process automation. You can kick it off by a trigger. You can do a batch job, and then you can have it do all of this fancy stuff. You can update data in Salesforce. You can make actions in another system like Stripe for doing payments. You can then also have it do pretty complex logic. And then it also integrates with the number one integration platform in the world MuleSoft, so that you can connect to any system. It’s not just the popular web apps out there. One of the things that is MuleSoft’s secret power is doing all of those arcane legacy protocols. So that if you have this old system, it can do it.
And then further we’re partnering with the best in breed RPA vendors like Automation Anywhere, like Blue Prism. So that if you have disconnected systems, you can further have this integration, this automation take action in those, too. So really we’re unifying it all the way across all of those. And then we’re also introducing Flow Orchestrator, which is to help organize the people processes where we could have this person make the decision on, okay, the underwriter has to decide, is this an appropriate risk? Then they pass it off to the loan officer who has to make a step and review it.
We want to make that visible, easy to monitor and unified with all of the rich integrations so that people can automate end to end processes in an easy to use way through all these tools. And last but not least of course is MuleSoft Composer, which is the non-developer way to integrate across systems. And so we’re making all of these work well together with the best in breed systems to have this really unique value proposition in the market.
Point-and-click automation tools non-developers can use
Bill Detwiler: So let’s talk about that portion of it, which is the low code, no code, clicks not code mantra that Salesforce has always had, but that we see more and more today, which is, as you pointed out, people who maybe don’t have a development background or they aren’t coders by trade, but they understand how the systems work. Talk about the interface maybe just to a little bit or about how those people can use Einstein Automate to accelerate or to accomplish those digital transformation plans.
John Kucera: Yeah. We want to meet all of the personas where they are. So you have, let’s say a sales manager. Hey, I just want to create something simple. I click the save button. If it’s a big deal, I want to alert my team. And so inside a flow, we make it so that they can do that easily. That happens by the way, literally a trillion times a month that people will click the save button and it kicks off those basic triggers that then send emails and do field updates.
Then you have your ops person, your person that is comfortable with Excel, they can build a macro. They can basically do a pivot table. These people then can make more complicated types of solutions. So they can then automate solutions for the entire department. They can use Omni Studio to put together a guided web form that you can integrate to a website. They can create a self-service chat bot to do customer self-service. They can do more of these guided workflows for a call center rep to do, say, password resets and a super fast way.
And so, fundamentally, this is transforming how people build because it’s reducing the expertise and therefore increasing productivity and allowing all these processes that before couldn’t be automated to be cost-effectively. But even better, these work with all those pro code building blocks. So the developers can avoid writing a wizard. Nobody as a developer wants to write like a wizard to move people through things. So they can put a best in class web components, like a custom UI component or an apex piece of logic, a pro code building block inside of these tools and extend them without limits. And so it’s really this unique value prop of helping each person do what they need to make their jobs, their department, their company, more effective.
Bill Detwiler: So you’ve given me a few examples. You talked about a call center. We talked about creating a flow for reps, help desk reps, to be able to change passwords, redo password resets. You talked about some payment options. Give me a really good, maybe example, or how one of the customers is using Einstein Automate now to change one workflow that people who are watching and listening might identify with and say, “You know what, I could use that in my work.”
John Kucera: Sure. I’ll try to spit off probably five or so. One is just creating a support ticket. So usually there’s a bunch of questions in a call center that are coming in and you need to jump between places. Using a guided flow for creating a support ticket is really important. The number one request that comes in for a lot of software vendors is reset my password. Making it easy to do that, either self service or in a guided way, because often you can’t log in to actually get to that UI is really, really important for salespeople. It’s renewables management. I need to make sure I have a structured process so that when I’m 30 days or 60 days out from people buying more, something expiring, I can have that process in place. I can guide people through it.
And then it’s also things like in government. So we had all the unemployment insurance requests. And so a lot of these governments had to basically scale about a 100 acts in the course of a week or two. And so they use Einstein chatbots, help people get the solutions they needed really quickly. Then you also have things in finance like mortgage processes. This is a really, really hairy, gnarly process with about a hundred separate sub ones.
We have our customers using this solution to move the work between the right people, making the customers aware of where things are at in the individual processes and more. And so there’s so many use cases across so many people that one of our problems is making people aware of these are all the great things you can do through this great suite.
AI and automation go together like “peanut butter and chocolate”
Bill Detwiler: So I love those examples and I think people, at a fundamental level, if they’ve been involved in designing a system, whether from a coding or from an admin perspective, or even from a product perspective or an owner perspective, right? Where does AI come into that process? Where does the Einstein component come into the automation part? What role does it play?
John Kucera: So fundamentally what I think about what AI does and I’m going to undersell it a little bit, is it helps make predictions and recommendations. And fundamentally it’s like an assistant to people. So people are really good at the judgment decisions and building relationships. AI gives us recommendations to help make those decisions. And then if we’re confident enough in those predictions or recommendations, we then say, okay, AI, great, just do that.
And so some examples of that, are Einstein chatbots. we have the best in class AI that can figure out what you typed and say, what did you actually mean? You’re saying, okay, I need to check on the order for this. We can figure out, okay, you needed to get the order status. Then we’re going to transfer that over and then go fetch the order status, bring it back to you.
So this is really that peanut butter and chocolate story where the AI can figure out whether it’s extracting the text, whether it’s looking at an image and figuring out the form fields on it, or whether it’s helping make a recommendation to somebody of what to do, based on all of this other stuff that we have in Einstein next best action, that we have in call coaching and more. So really the AI is helping us make those better decisions. And when we have real good confidence in it, as we do in chatbots, actually taking those actions to make these better experiences.
When can I talk to Tableau? The future of AI, automation and voice
Bill Detwiler: So, I mean, I love the analogy of the peanut butter and chocolate there. The two things being complimentary. I’m curious how, and I like the examples too, about how AI is manifesting with the chat bots and actually trying to gauge user intent and not just actually what the words are, but what they mean. What do you think the next step for AI and automation is? What’s that next place we’ll be moving to with this combination?
John Kucera: There’s so much more. So as the models get smarter, as it gets easier, one of the things that’s really transformative is when you don’t have to be a data scientist yet you can use all of these insights. Like collecting all the data into more places or into a single place to run that against, that unlocks all of these possibilities. So there’s really unique solutions that are popping up there as our data is consolidated in one place and the predictions get smarter.
So maybe 10 years ago reading a form with confidence wasn’t good. The technology just wasn’t really there. Now we’re able to introduce things like form reader, where you can put a driver’s license or a 1099, or any type of different PDF that you have and say with confidence, this is exactly what the people have written in there even if it’s with a hand scribble. And what’s transformative about that is that’s a tedious job somebody had to do before. They had to manually retype that looking at the form. And so then it frees those people up to making those judgment calls and making those decisions.
I think voice is another frontier that we’ve talked a lot about. And one of the things with service cloud voice is transcription in real time. We now have high enough accuracy to figure out what you are saying, make text out of that and then further have more AI on top of the text, kind of like bots to figure out the intent behind what people are saying. Then that lets us say, hey, you might want to suggest saying this. They might be objecting to this concern. Or they sound kind of upset, you might want to be careful about what you say next. And so it’s really helping superpower people and have AI be the digital assistant for building these relationships and making people more productive.
Bill Detwiler: So, eventually you’re telling me I’ll be able to talk to Tableau and ask it for a certain report and it’ll give it to me without me having to understand how to do query language and how to build that.
John Kucera: Exactly. It’s your natural language search for everything. I think the ultimate that all the smart people in the movies figure out is we want our Jarvis from Ironman. We want to just be able to ask questions to this really smart assistant, have them figure out what we’re saying, do the legwork to collect and organize that info and bring it back to us so that we can make decisions. That’s where the future is and that’s the vision that we keep marching towards.
Bill Detwiler: And I think what some people don’t realize is how difficult that is to do. I mean, especially with different languages and with different meanings, different pronunciations of words. It is not a simple task for a machine to understand human intent. I mean, it’s not a simple task for humans to understand human intent sometimes.
I’d love to close things out and we’ve touched on this a little bit, but talk about where you see process automation and automation in general, maybe going in the next few years. What are your customers asking you to automate for them that maybe we can’t do now because of limitation in technology or we just aren’t there yet.
John Kucera: One of the things that a big focus is integration. You’re starting to see this where all of the different systems, how can we talk to all of those? How can we get data from one, bring it to this system and vice versa, how can we trigger off of changes there? There’s been a huge amount of innovation in this area, and there’s a lot of progress, but there’s still so many different protocols, so many different places to unify. And so just getting all of the data into one place is really hard.
And so that’s one of the things that I think is going to be really transformative over the next few years is you’re going to increasingly see all of these different systems brought together. And what this is unlocking is a way to automate across those systems, automate across teams, across departments, do these company-wide types of processes in a unified way where you can monitor, where are the bottlenecks? You can analyze what’s working and what’s not? And then you can really do that productivity gain at the next level.
So I think integration is a massive, massive piece there. And then of course, on the AI side, it’s going to keep getting smarter. One of the funny things people don’t know, we actually only ourselves when we’re talking to a human get 95% of the words that are spoken. And so we need the AI to get that high and higher. We’re basically at that cusp of understanding what people are saying and almost as good as a human, I believe. And soon I think we’ll surpass it, which will be really interesting.
And we might have aides to basically have the AI tell us what we missed when we were talking to people like when your Zoom goes down and it’s kind of fuzzy. So I think like the increase of AI getting above those certain accuracy thresholds is really, really interesting because it opens up all these different use cases, whether it’s in voice and talking to people and the text transcription which you have for bots and call coaching and so many more use cases.
ZDNET’S MONDAY MORNING OPENER
The Monday Morning Opener is our opening salvo for the week in tech. Since we run a global site, this editorial publishes on Monday at 8:00am AEST in Sydney, Australia, which is 6:00pm Eastern Time on Sunday in the US. It is written by a member of ZDNet’s global editorial board, which is comprised of our lead editors across Asia, Australia, Europe, and North America.