Welcome to the Use Case Podcast, episode 260. Today we’ll be talking to Alex from Thoughtful Automation about the use case or business case for why his customers choose Thoughtful.
Thoughtful Automation gives human workers the freedom to pursue the creative, strategic work that builds companies, as well as their careers.
Give the show a listen and please let me know what you think. Thanks, William.
Show length: 25 minutes
Enjoy the podcast?
Announcer: 00:02 Welcome to Recruiting Daily’s Use Case podcast, a show dedicated to the storytelling that happens, or should happen, when practitioners purchase technology. Each episode is designed to inspire new ways and ideas to make your business better as we speak with the brightest minds in recruitment and HR tech. That’s what we do. Here’s your host, William Tincup.
William Tincup: 00:24 Ladies and gentlemen, this is William Tincup and you are listening to the Use Case podcast. Today we have Alex on from Thoughtful Automation. We’ll be learning about the business case of the use case for wise prospecting customers. Pick Thoughtful Automation. So Alex, would you do us a favor and introduce yourself and Thoughtful Automation?
Alex: 00:45 Yeah, absolutely. Thanks, William for the generous intro. Sure. Alex Zekoff, Co-founder, CEO of Thoughtful Automation. Started my career back in 2009, Deloitte Consulting and Tech Group, building large ERP systems for aerospace and defense companies. From there, I actually moved to Japan, helped a large aerospace company set up a company there. From there I got to go to business school, sponsored by Deloitte, went to Berkeley for a couple years, stayed in the Bay Area for an extra three after that. I was there actually I went back to Deloitte. They had sponsored me to go to business school, and I got into robots and not the physical kind, the software kind. And it was really fascinating, so 10 years prior when I started at Deloitte, we were helping build the ERP system, and then 10 years later I joined that same client actually and we were building software robots to automate that software that we had built.
01:41 So within 10 years I had seen this really interesting automation cycle in the digital space from build system and software and train users to use it to build software robots software to software and people not having to operate software. And so with that sort of light bulb moment, I was like, “This is going to be pretty massive. This is going to be… a physical automation takes decades, this is probably going to be 10 years that it takes to really massively roll out these software robots.” So with that, I left Deloitte in 2019, I started Thoughtful Automation as a solo founder, and actually my co-founder now he had wrote the first angel check along with my dad into the company. We basically with an idea I was going to go out and try to find a path for how to get these software robots into as many hands as possible.
02:28 And we were fortunate. We found our first customer in South Florida actually. I had just moved there at the time to, I had met a lovely lady who’s now my fiance in Cabo actually and followed my heart and moved down there after three months after dating to see if the relationship would work out. It has. We’re now engaged. And I found a customer down there too, tried to sell him a bot, we call him digital workers, and we’ll get into that, the why behind that. It worked out. We actually found the business model there that is sort of our innovation automation as a service, and I’ll get into that more. And then from there we’ve been building a lot of digital workers or bots for customers. That’s been really exciting seeing the market really receive these well, and they’re actually hiring these digital workers as employees.
03:15 So the same way a CFO might look at a part of their org and say, “Hey, we’ve got 10 people running our revenue cycle department for this healthcare org.” CFOs are now hiring digital workers as employees, and we license them to other companies as employees with salaries and expectations and sort of a JD, and that’s sort of the business model at the simplest form. We’ve now got around a hundred digital workers, different types, different systems, different use cases we can offer as employees to companies. And our mission is to build the world’s largest digital workforce, getting a million digital workers live as fast as possible to help you alleviate the mundane and repetitive task that employees have to deal with data to day, basically just processing data between systems, kind of the data assembly line.
03:58 And quick stats on the company. We are seed funded 21 million from Drive Capital and Silicon Valley Bank. We’ve got 30 US employees and 30 full time Python developers that build on the Thoughtful platform helping build the digital workers. And yeah, we are growing and scaling. I think this is a time with economic uncertainty happening. We’re seeing a lot of demand increases for these digital workers. So I’ll pause there. A lot of information. Happy to double click wherever you want to go.
William Tincup: 04:28 Oh yeah. Oh, we’re going to be double clicking all over this. So digital workers, well, let’s go into a couple examples of some of the different digital workers that you have right now and also kind of where you see that expanding.
Alex: 04:44 Yes, absolutely. So one of our first digital workers was an invoicing digital worker, primarily in the healthcare space, behavioral healthcare, and the task that digital worker was to run through 50 clicks, four different systems, sort of electronic healthcare record system, your Microsoft Office suite, send some emails, upload some data. Two people were doing that full time before. They were basically just going through that sequence, sending invoices. Those two people were sending 30 invoices a day. The CFO of the company, Butterfly Effects is the name of the company, came to me and said, “Hey, this is a problem. We have a high turnover in this role where maybe six months max and then people want to leave because they hate that job.” And he is like, “I got to keep back filling it, training it. The cost of that obviously is very evident.” And then just quality. There was a backlog growing. They could only get 30 out a day, and they really needed to get out about 50 a day, and so that growing backlog was a pain for management.
05:42 So I came in, it was actually just me at this time, solo founder, solo company. We built the first digital worker. That one specifically was the invoice digital worker in a healthcare system. Built that in six weeks, so that’s basically how long it takes to build a digital worker and replicate this sort of employee’s task, and we license that as an automation, as a service model. So what that means is that digital worker costs $2,000 a month, and it’s a responsibility for Thoughtful Automation to keep that up and running and healthy, so we’re basically responsible for the performance, and the company gets the value. That’s why it is a very compelling business model for most companies because they basically just lock in cost savings or some other value metric that’s determined by the customer, so that’s one example.
06:30 So just to talk about that digital worker specifically, what happened as a result? So we deployed that digital worker, and the company is now able to, a quarter of an FTE can send 200 invoices a day versus two people sending 30, and so what that means on a, let’s say a per invoice sent transactional level, they were spending about 12, $13 per invoice sent, and now they’re spending less than a dollar on 87 cents per invoice sent. So again, that’s just huge savings obviously. But again, the efficiency, the digital workers help the team members find places where the process is broken and actually improve the process, and we actually build it with human in the loop. So this is a little bit how we’re different. We don’t try to get rid of people.
07:13 We know what people are best at, which is working with other people, creative ideas, calling people for collections. These are all things humans are really great at, not operating in software and just moving data. So we design with our front end system called Empower which is meant to interface with the digital worker and a human worker so that they can work together and become more even more productive. And that is one example of a digital worker.
07:38 And then we actually went from there and said, “Hey, what are other jobs that are really mundane and repetitive in the healthcare, specifically behavioral healthcare revenue cycle space?” So we saw claims processing, invoicing, the one I said in statements and secondary claims. So we created an entire suite of digital workers, the revenue cycle suite, and now we have CFOs of behavioral healthcare providers basically licensing the revenue cycle digital workers suite from us, and we’ve now been able to replicate that many times for several behavioral health providers, so that’s just one example of going from one to many digital workers at a company. And even further, we’ve seen companies take that and introduce us to other departments, specifically HR and talent revenue cycle. They had some touch points between the two departments, and so we got introduced there, and now we have some digital workers in the HR talent department.
William Tincup: 08:34 I know you’ve faced this question in probably a jillion different ways, but some people will listen to this and say, “Well, we’re replacing humans.” And of course that’s a knee jerk reaction. But it seems to me that what we’re really doing is that we’re making work more efficient. And again, a human being should have never been doing again, the invoicing. There was a better way. We found a better way those human beings can be redeployed into other things that are more human centric, and you probably have more of an eloquent way of saying that.
Alex: 09:11 William, that’s perfect. That’s exactly right. I use the Henry Ford assembly line example just to give a physical world representation of this. You look back in 1913 images of Henry Ford’s assembly line, you’re going to see people on the assembly line moving the car along. Fast forward to the eighties, you’re mostly going to see a robotic factory.
William Tincup: 09:30 Right.
Alex: 09:31 And so that same representation is happening in the digital world. Why is that important? Do people want to be in those factory jobs building widgets? These are dangerous jobs right back in the physical world. And then we got robots so we could get more throughput efficiency, right? So these jobs become not in demand anymore. We actually saw this with a great resignation. A lot of people said, “I don’t want to do these jobs. I don’t want to show up to work.” So employers like, “Well we still have this work to do.”
William Tincup: 09:57 Yes.
Alex: 09:58 And so what happens is employers have to end up spending more money to get people to pay to do these sort of mundane jobs, psychologically pretty terrible for a human. If all your job is to follow a script all day, and you get no creative freedom, no interpersonal relations, typically pretty bad for a human’s brain, so we just look at it like we’re actually able to free up that time. That’s how we look at it. We’re freeing up time for people to either one, look at the process and optimize the process which is value add, go work on creative solutions, go work on cross team collaboration. These are all things that are going to help the company grow and evolve and create more efficiency. Not just run the software, run the process. This is all basically commoditized work that should be automated.
William Tincup: 10:41 Right. And we’re just going to hunt as we should for other things that should be automated, so once this company that you’re just explaining, they’re already introducing you into HR and talent and sales and other things. It’s a constant, probably an audit, of where, what else can be automated? Where else can we gain efficiency? And again, not getting rid of humans, making humans more efficient at the things that they should be efficient at.
Alex: 11:17 Yeah, that’s right. Typically you’ll see creating 5x superhuman powers. So if you give a person a digital worker, and they were typically doing that process before, they’re going to be able to just offload all of that work to the digital worker and free up time and actually create a ton of value for the organization then. And I think all companies want to love that, is like, “Hey, let’s just automate all this back office work that we have to do and then let’s go sell more deals. Let’s go figure out how to create new products.” These are the fun things. These are the things that I think people really get excited by when they go to work. Not the I have to process my email. I mean, me for example, I spend a lot of my day, and I’m sure you do too, emails, slacks, updates. You realize when you count all the hours, you’re literally just clicking and moving and doing repetitive task all day. Wait, did I do anything today or did I respond like a machine would to my email?
William Tincup: 12:11 What are the edges that you found? Because somebody asked you if you could do something, you probably evaluate it and say, “Okay, yes, this can be done.” What are the edges that you’ve seen that you’re not quite there yet? Or the technology’s not quite there yet?
Alex: 12:26 Yeah, great question. So we do see, we use OCR a lot. So every digital worker, think of it as we’re assembling a robot, and we’re adding all the sensors, and so we use a lot of our technology to solve problems, right, and to replicate what a human’s doing when they’re processing that digital work. So one example is you receive a check, typically there’s a handwritten thing or an unstructured part of that, and then you need to use OCR to kind of read that unstructured handwriting and then put that in a system so you can process that check. And OCR capability’s very high. We can process most checks. Can we process a hundred percent of all handwriting if someone has chicken scratch? No, it’s not there yet. We’re not at perfection with a lot of this being able to see handwriting. And I think even some humans can’t notice their handwriting point either.
William Tincup: 13:15 I’m thinking about my son when you’re saying this, but yes.
Alex: 13:19 Yeah, I mean my dad’s a veterinarian, and I always, his handwriting was horrible, and I don’t think any digital worker would be able to identify his handwriting.
William Tincup: 13:27 No. It’s hieroglyphics at this point now. No, the irony is he understands.
Alex: 13:31 Yes, that’s what he always told me. He’s like, “I just need to know it. You don’t need to know it.” Yeah, no, I think the capabilities there where it’s just not structured enough or the machine learning models haven’t figured out that slight variation, humans are still going to be best there. In addition, where digital workers have a problem is if a process has non-codified logic, right? So typically part of our design process we’re like, well, we interview people, and we’re like, a lot of people keep the process in their heads, so we have to get that out. And because Elon must doesn’t have neurolink really live yet, we can’t just extract that from the brain, so we have to interview people to say, “Hey, well why did you click that screen?” “Oh, because of this decision in that system.” So that’s our codification of logic process.
14:18 But if we ask a question and go, “Hey, why did you do that?” And it’s just an intuition, can’t codify the intuition yet unless they can give us a rule. So we will run into instances where we’re building a digital worker that replicates a certain job function. We do the design process, and we realize, “Hey, there’s actually no rule for that.” So what we do is we will put basically a human in the loop gate, and this is where the digital worker stops, the human logs into our system Empower. They go, “Oh, I need to go do this part of the work because that’s just creative or in my brain or I don’t know how to codify it.” And then they can also give it back to the digital worker to keep it completing it. So we have solved for that human in the loop. But yeah, that’s the stuff that is the intuitive stuff that we can’t solve for. That is really the human creative part.
William Tincup: 15:07 Over time, maybe again, we don’t know what we don’t know, but maybe over time we start to learn more about that. So you mentioned a phrase, automation as a service. I want to go back to that for just a second. So as you’re explaining this to prospects and customers, et cetera, it’s probably new to them as it says, it was new to me as you said it. So how do you frame that up for them?
Alex: 15:35 Yeah, so it’s the typical buy verse kind of rent, the buy verse rent model. And when I was at Deloitte, we would do the typical builds. We would go onto our client site, we would build an in-house capability. I was consulting with large enterprises, Fortune 50 clients. We would build an in-house capability, almost like if you kind of think back to the early two thousands, late nineties, companies building servers on Prem, that was in very vogue then. Same things happening with automation. Companies want to build their in-house capability, they want to scale the team that automates the work, and that costs a lot of money to be honest with you. So to build a bot in an enterprise use case, you’re looking around $250,000, and that’s just like to buy it, right? Or to buy it from a consulting provider like a Deloitte, IBM, Accenture. And then once you get that bot, you have to maintain it.
16:28 And what I saw, a real big issue was a lot of companies, they’re all doing it differently, there’s no standard protocol, the bots or digital orders will break and then it takes two weeks to get fixed, and there’s a lot of chaos. And then people stop using them when they break. And so what I found was, hey, why don’t we just follow the same playbook that cloud went is why don’t people just rent the bots, just like you kind of give an employee a salary. They need to come up, show every day to work, and they need to provide value. That’s sort of the same corollary that I think automation is trying to replicate, but in the current model, trying to build an in-house and from scratch is just, it’s so expensive. And also at $250,000, most of the work really that’s getting automated is providing around 75 to a hundred thousand dollars of value, so it could take, sometimes with the maintenance, three years of payback.
17:18 In today’s world, three years. I mean that’s forever in technology terms. So when we were playing with the automation of service idea, it’s like, well let’s find the price point which makes sense and provides value and lock that in for the buyer, which is typically CFOs or COOs. And so when we scope a process, we actually scope it, we have a mechanism, it’s a proprietary model where we scope our digital workers to one to two human FTE depending on the process complexity. There’s a lot of factors that go in there. And we lock in typically between 50 and 80% pure cost savings. The customers give us those inputs. We’re just saying we agree to that, and then that’s our job is to provide that value back to the customer. And that is the essence of automation as service. We build digital workers faster than anyone, I believe in the world, just coming from my experience doing a Deloitte, we can do them in as fast as six weeks if we’ve done it before, and if it’s a net new replication of a person, we can do that in three months.
18:14 So we basically can replicate any person in three months and license that back as a digital worker. And so what that means is if you’ve got people, and they’re doing work, and they’re expensive, and they’re doing work that should not be being done by a hundred thousand dollar a year person, that’s typically we’re going in and we’re going to go look at that work and go, “Okay, our average digital worker is 25 to $30,000 a year, going to lock in $70,000 of savings.” And typically we’ve seen as those people go into work on more interesting projects, like the process improvement initiatives within the department, scaling, they actually, they’re like digital worker managers now, so they manage the digital workers, they optimize them, they gave you more output, just kind of changing how we think about humans interacting with automation.
18:58 And so that is the core of automation service. What we actually do, and this is the kind of the cool tech, you aren’t going to see this on our website, we’re building the gigafactory. So if the digital workers are the Tesla cars, we’re building the machine that builds the machine. So a lot of our digital workers are already self building, so they generate code. So 20 to 30% of it can just be spun up, so you save that in a time to build it.
19:22 And then we use a mechanism called work blocks. So because we own the asset, and we’re renting it, if we’ve used a work bot before, let’s say it’s a log in to a certain system, we can reuse that to make it go faster, so that drives down our time to value our implementation of digital worker even further. Our personal goal is in two years to get this down to two weeks, that you can basically put out a JD out there for an entry level data role or claims processing and we can replicate that in two weeks, and then you can basically spin up the digital worker and then it’s trained, it’s ramped, and it’s adding value. And that’s sort of, I think, the mission here as we think about it. Again, building the machine that builds the machines.
William Tincup: 20:00 Oh, I love this. So questions that prospects should ask if they’ve never been down this road before, which I’m assuming a lot haven’t. They’re intrigued, obviously, but what should they be asking you?
Alex: 20:13 Yeah, absolutely. A lot of people ask, “Should I automate? Should I outsource to BPO facilities?” That’s typically like a question I get, and I say, “Well, I’m going to save you a lot of time. When you go to think about the outsource question, digital workers are going to be less, and they’re going to be perfect. You’re still going to see issues with the BPO facility, and digital workers are going to cost you less than a BPO facility.” So that’s typically a question I get typically.
20:40 Next up is, “Well, how much time do I have to put into this, right? Is this a full time job for me to build a digital worker?” Another beauty of the automation of service is we do everything from design the digital worker, train it, deploy it, manage it, and optimize it, which is part of the learning. So it actually gets smarter over time, so that’s all included as automation of service, kind of full end to end. And why that is important is because I can tell you when we did this in the enterprise, it would take us six months to build and that was a full-time team working on that. So people really just need to provide the inputs, provide the questions, sort of the design parameters. That’s typically about anywhere between two maybe max four hours per week for a process champion or expert, and that’s it. That’s all they need to do is they have that free time, then we can go build it and then sort of manage it and optimize it. The next question is-
William Tincup: 21:33 Oh yeah. Go ahead. Finish your thought.
Alex: 21:35 Oh, last I would just say is, “Do you have a champion? Do you have an advocate? Do you have someone who really wants to do one?” But we see the benefits when you really get to five, and they’re all working together, you’re going to see a multiplier effect on the value, so it’s really great if you do one. It’s even better if you do two, but if you get to 5, 10, that’s when you’re going to start to see whole departments. I mean you’re going to really see the value. So having a champion kind of run that is really important.
William Tincup: 22:03 And then this will drive you crazy on some level, but your favorite customer story, not naming names or brands or any of that type of stuff, but just somewhere where you’re like, “Oh, this was groundbreaking. This was actually really cool.”
Alex: 22:24 Yeah, there’s a lot of great use cases and stories.
William Tincup: 22:31 Go with your most recent.
Alex: 22:37 I want to go back to one of our earlier customers. Actually, it was in our first year. I had told you about the first bot, the invoicing bot, and we were really excited by that one because that was the proof of concept. We got it out there. And then we were like, “Hey, let’s go see if we can resell this one.” And so our second customer, another behavioral health provider out of Denver, they were really excited because they had actually specked this work with a Deloitte-like entity, and someone quoted them $300,000 to fill all their digital workers.
23:14 And for most companies in sort of the mid-market SME space, that’s just like a no, too much, and then they have to spin up a team to do that. And so this was our second customer, I remember going in for the sale, and I’m just still like, “Hey, this is such a new concept.” And people were like, “What is this automation on service?” They’re like, “This is too good to be true. But we just got quoted $300,000.” And we came in, and I think we charged, I know exactly how much we charged. I mean we charged $8,000 a month for everything, and this was a suite of 5. This was the entire revenue cycle department, and that got the deal done. And what’s amazing is we were able to spin up an entire department.
23:55 And this was early days by the way. We were still four employees, and now we’ve got 60 total people, so we’re able to get in faster, but we did it in four months.
William Tincup: 24:03 Oh wow.
Alex: 24:04 That would’ve taken us at Deloitte probably million dollar plus project. So you’re talking, it costs $32,000, $8,000 a month over four months. I really knew we had something there. That was 10% total cost of ownership to get those deployed, and that’s when I was really excited because I knew we had something. I knew we had something that wasn’t just single digital worker. We had a department, a demand for a department. They got implemented, users were using them, and they were working at a much higher efficacy.
24:33 So another thing is when I’m doing this in the enterprise, a lot of these would fail about two out of five times, about 60% hit rate success, which is not great. If you’re thinking about an employee only achieving 60% success. And with these, we were able to get them up, once they were fully trained and optimized, to above 95, 96, 98%. That’s incredible. And that doesn’t mean they fully fail. It means sometimes they don’t have a record and so they won’t process a record.
William Tincup: 25:01 Right.
Alex: 25:01 But fully successful runs, we’re getting up to those rates, just incredible progress moving the needle, and so I knew we had something when we got the TCO down and the quality and the value was being provided was high. I knew whatever you do in life, if you can provide a high quality product at a 10th of the cost, something’s going to happen.
William Tincup: 25:20 Yeah.
Alex: 25:22 Something magical will happen.
William Tincup: 25:22 Not sure what, but good things will happen.
Alex: 25:24 Good things will follow. Correct.
William Tincup: 25:28 Brother, this has been absolutely wonderful. Alex, thanks for carving out time for audience.
Alex: 25:33 Awesome. Thanks, William. I really appreciate it.
William Tincup: 25:34 Absolutely, and thanks for everyone listening to the Use Case podcast. Until next time.
Announcer: 25:40 You’ve been listening to Recruiting Daily’s Use Case podcast. Be sure to subscribe on your favorite platform and hit us up at recruitingdaily.com.
William is the President & Editor-at-Large of RecruitingDaily. At the intersection of HR and technology, he’s a writer, speaker, advisor, consultant, investor, storyteller & teacher. He's been writing about HR and Recruiting related issues for longer than he cares to disclose. William serves on the Board of Advisors / Board of Directors for 20+ HR technology startups. William is a graduate of the University of Alabama at Birmingham with a BA in Art History. He also earned an MA in American Indian Studies from the University of Arizona and an MBA from Case Western Reserve University.