Liz Bearce: Thank you for joining our webinar, co-hosted with UiPath and Lydonia, to discuss the future of work, the power of agentic automation.
Logan Emerson: And I’ll hand it over to Vinny to talk about some specific use cases.
Vinny LaRocca: Yes, I think we’re getting asked pretty frequently, you know, what the use cases look like for this, this new agent capability. And I think it is better thought of as a capability. And what I mean by that is the two processes I’m going to show today of where you would use agents within these processes are things that we’ve already automated, the things we’re already doing. And there are also things that we can actually go back in time now and improve upon by deploying agents to the things that have already been automated. So I’m going to kind of show where within these process flows we would use agents, and I think that’ll help clarify, you know, where people might use this capability.
So, the first one is invoice processing. You know, a typical invoice process flow would look something like what you’re looking at, where we’re going to get an invoice in. We’re usually going to throw that to document understanding, and we’re going to try to extract the information out of it. And then we have to go program all those business rules in to say, “Is there any discrepancies here? Is there anything that, you know, doesn’t make sense?” And if so, I need to flag it to a human. I need to know, to catch that. And then I need to let a person actually triage that, whether it’s going back and forth in an email with a customer, and then the robot can pick up the rest of the data from there, enter it into your ERP, and kind of finish the solution off.
Yeah. And so, this kind of shows where we would plug in the agents for this kind of capability. So, you know, if I’m having those discrepancies, this is a big area where we’ll have processes that kind of fall apart when we sort of leave the flow. And we have to go into an email and start going back and forth with the customer. Things get lost in email, and then I’m also going back to manual. I’ve got to take that email in, I’ve got to look at it, figure out what information I need to change, and then actually go do that and pick up the process where it left off.
And so, you know, what these agents can do is now have the ability to not only go and look up records within your database, do communication via email, but more importantly, it can figure out which of those steps it actually needs to go do. So, I’m not saying, “Go do this thing and look up data.” I’m saying, “Hey, I have a discrepancy. Here’s the data,” and it will be able to go try to figure out what next steps it needs to take in order to triage that without flagging it to a human. So, you can imagine we could take a process that even if you’ve automated invoice processing, you might be 80% automated. And we can take it that last 20% all the way to the finish line.
Now, another common use case that we see all the time is order entry. So, you know, we get an order in a couple of places in here where this is a really problematic solution to do with just robots is really upfront. I’m going to usually get some sort of email into a shared mailbox that comes in. We see a lot that that email can be a combination of information on the order within the body of the email itself, in a document that’s attached to the email, or a combination of both, and also trying to figure out if that email that came into that shared mailbox is an order at all.
And then, once we’ve been able to do that, which traditionally, we’ve done with pretty large lookup tables that have taken significant amount of development effort to build out and we’ve extracted that data, then we’re going to look up the SKUs in the system, and we’re going to try to complete the order unless we find out that that SKU, or part, or whatever was ordered, is actually not in stock, in which case we’re going to go back and forth with the customer. We’re probably going to want to suggest similar parts if there’s maybe a replacement that they can make to get something that’s in line with what they were ordering. Figure out, you know, potentially what my supply chain looks like so I can tell them what the back order time looks like, and then do that back and forth communication to swap out if they want to, before I actually complete my order.
And so, there are two areas here as well where we can plug in agents. You know, the first one is triaging that email that comes in. So again, traditionally, really difficult to figure out, where is all the data? Is all the data even here? Do I need to go back and forth with the customer if it doesn’t exist? And is it in the body of the email, or is it in a document that’s attached? So, being able to have an agent now orchestrate all of those pieces to make sure it’s gathered all the information it needs first, and then, if not, going back and forth with whoever sent that email in the first place, to get any missing information, takes a huge amount of development effort out of trying to build that process.
And then as well, if we find out that, “Oh, this thing that was trying to get ordered is either back ordered or not in stock at the moment,” we can actually go look at automatically. Look at those databases, you know, leverage those as skills or tools within the agent, figure out, you know, what the supply chain looks like, and then do the back and forth with the customer to acquire whatever new thing that they want to order and then finish out the process like it would have happened prior. So again, now we’re talking about something that was probably closer to 50 to 60% automated, potentially going all the way to like 90 to 100%.
Liz Bearce: Logan, Vinny, thank you both so much for your time, and again, thank you to everyone who joined today.
