Episode 32 — Artificial Intelligence with Adam Hofman

Episode #32
Original Air Date: 04.18.2017

30 Minutes

Listen Here
0:00
0:00

Our Guest This Episode: Adam Hofman

Adam Hofman joins the cast to discuss artificial intelligence and his role at CallRail -- a company that specializes in phone call tracking and analytics that helps businesses understand which marketing campaigns are delivering valuable phone conversions.

Adam is a product management at CallRail, an Atlanta-based startup focusing on call analytics.  His role is focused primarily on the company’s Conversation Intelligence products, which are centered on artificial intelligence and machine learning. 

Favorite Quotes from Our Talk:

"It’s always just a constant learning about how people are using it and a constant struggle of what do we include and what do we not.  And it would be great to do this, but the technology’s not quite there yet.  And how do we – we need to give customers an expectation of, like, this is the accuracy that you can expect.  And we’re constantly trying to improve on it."

- Adam Hofman

"I think the biggest challenge for me was being able to know that there is no way that I’m going to be able to learn this, but there is somebody that I can ask.  And being able to be willing enough to ask questions and to learn from the questions that I ask and to truly listen to the people who had the expertise in it was what made this project successful."

- Adam Hofman

"We’re all intimidated, and we’re all afraid everybody’s going to figure out how dumb we really are, And the truth is we’re all figuring this out."

- Adam Hofman

Share With Others

 

ANDY CROWE ● BILL YATES ● NICK WALKER ● ADAM HOFMAN

NICK WALKER:  Welcome to Manage This, the podcast by project managers for project managers.  This is our chance to meet with you and dig into what you think is important in the world of project management.  Maybe you’re new to the field, have questions about certifications, or perhaps you’ve been in the trenches for years with multiple projects under your belt.  Whatever the case, we’re here to give you some insight, some ideas, some principles that you can use and hopefully avoid some hazards along the way.

I’m your host, Nick Walker; and with me are the in-house experts at all this, Andy Crowe and Bill Yates.  And Andy, we have a guest with us today who can, among other things, probably answer the question, “Who you gonna call?”

ANDY CROWE:  You know, it’s exciting as we start looking at this and looking at artificial intelligence, this is something a lot of our listeners brush up against.  And there’s so much going on in this field that it’s going to touch all of our lives.  So we’re really excited to have Adam in here today.

NICK WALKER:  Adam Hofman is a product management at CallRail, an Atlanta-based startup focusing on call analytics.  His role is focused primarily on the company’s Conversation Intelligence products, which are centered on artificial intelligence and machine learning.  Adam, welcome to Manage This.

ADAM HOFMAN:  Thank you very much for having me.

NICK WALKER:  So Adam, I’m wondering if you could, first of all, maybe just paint a picture in our minds of how your products work.  What do they do?  Give us a scenario of how they’re used.

ADAM HOFMAN:  Yeah.  So primarily CallRail is a call analytics and call intelligence platform.  So we have a lot of marketing agencies that use us with clients who they have phone numbers on their website.  They put phone numbers on a postcard or on a billboard and really want to know where those calls are coming from and how many people are calling those, you know, calling those different campaigns that they have.  So we give those customers phone numbers.  And those phone numbers actually forward to – they forward to a business number of that company.

So when that call forwards, CallRail is able to capture the information from that caller, whether it’s basic information like the caller’s city and state that they’re calling from, or the caller’s name, or in certain scenarios whether they visited a website or they searched in a keyword or they visited a certain campaign before they placed the phone call.  So with that, those companies are able to really make better and smarter marketing decisions.  And they’re really able to hone in on what is being successful for them and grow.

NICK WALKER:  So if I call in to a certain business, and I get a recording that says, “This call may be monitored for evaluation”?

ADAM HOFMAN:  Right.

NICK WALKER:  That might be you.

ADAM HOFMAN:  Yeah.  So some of our customers choose to record their phone calls.  I would say about 80 percent of our customers choose to use that feature.  And that’s really based on local laws, and they kind of have to be aware of do we have to let these people know that these calls are being recorded?  In most cases it’s yes.  But they are able to use that, and we’re able to actually use that, to build more analytics products around that and be able to, like from a customer standpoint, they’re able to listen to a call after it happens and really analyze the quality of a call.  And going forward, CallRail, with these new products that we’re building in the call intelligence world, is able to do that without them having to listen to a call using artificial intelligence and machine learning.

NICK WALKER:  So how did you initially get into all this?

ADAM HOFMAN:  Good question.  If you had asked me, like, two and a half years ago if I thought that I was going to be a product manager for a software company, I’d have been like, well, maybe, but I don’t really know.  Really when I moved to Atlanta about four and a half years ago, got just fascinated by startups.  And I was working out of Atlanta Tech Village for a while, just kind of doing copywriting for different startups and different companies, and really just got fascinated by how startups work, really what it looks like to be successful as a startup, what it looks like to do it wrong.

And throughout time, all of a sudden CallRail kind of came up in my Twitter feed, and they got some funding from some capital investors.  And I took that as my opportunity to reach out to the CEO directly and tweeted at him and said, “Hey, I think you can use me.”  And then from there I had multiple interviews.  And I ended up actually on the customer support team for about eight months, and then moved towards the product side of things.  That’s when I started to become aware that I might be pretty good at this.  And they recognized that, which was awesome.

NICK WALKER:  Now, I know Andy and Bill have some questions about the nuts and bolts of how this works.  Bill?

BILL YATES:  Yeah.  I love the way that you came into a startup.  And you pursued them, and then you saw the opportunity and took advantage of it.

ADAM HOFMAN:  Yeah.

BILL YATES:  That’s fantastic.

ADAM HOFMAN:  Yeah.

BILL YATES:  And it’s really cool to me to see that you kind of went in as a problem-solver.  You were looking and seeing, okay, these are ways that customers are struggling right now.  How can we make a better product?

ADAM HOFMAN:  Right.

BILL YATES:  So here’s where I want to start to jump into the project management side.  As you are influencing the product and building more features and capabilities, we’re talking artificial intelligence.  So when you’re looking at requirements and figuring out what can we add to this product or not, what can we build out, how could you determine, you know, this is something that we can do; this is something that we can’t do.  This exists; this does not, this capability.  And, you know, figuring out – so you were vetting customer needs with viable requirements.  How did that work out?

ADAM HOFMAN:  Really, like from the start, from a process standpoint, for CallRail we experimented a lot with what specific processes worked the best.  So at the beginning, when I started that role, it was kind of disorganized, which was okay.  And we felt okay with that.  But throughout time it just got more and more – it got more and more organized from the more projects that we did.  It used to be that we would have an idea; and we would go, yeah, let’s develop that, and let’s build that; and then we would build it.  And it would take a long time.

But now that we have, like, processes in place, and we’re using, you know, we’re using like an Agile-ish sort of process, that has taken our project time down dramatically.  And from actually, like, starting to get the idea of we want to build a conversation intelligence platform, and we want to use machine learning to accomplish that, that just came as an idea and was presented to me as, like, do you want to work on this?  And I said, “Yeah, I don’t know really anything about that.  But I can sure learn.”  Yeah.

So I took a lot of time.  I took the first three months of the project basically like learning as much as I could about artificial intelligence, about machine learning, about natural language processing, about what that looks like.  And we actually worked with a third-party provider to help us kind of start with that process.  And after that we kind of decided, like, hey, we can possibly build this ourselves.  And really it took a full year for us to get the initial product out, the initial models out, and for customers to start using them and us to kind of ideate from that.

But it’s always just a constant learning about how people are using it and a constant struggle of what do we include and what do we not.  And it would be great to do this, but the technology’s not quite there yet.  And how do we – we need to give customers an expectation of, like, this is the accuracy that you can expect.  And we’re constantly trying to improve on it.  And allowing them to use it, but also helping them understand that it’s not going to be perfect, and that it’s amazing that a computer can do this.  But, yeah, from a project side, like, it’s been a whirlwind of different things every day, which I love.  Yeah.

ANDY CROWE:  Adam, I’ve got a question for you.  And I want to come back to the sort of Agile-ish approach you guys use in just a minute.  But what can you tell our listeners about the technology or the platform this is built on?  How do you build something like this?  And I assume you guys probably didn’t try and solve the natural language problem on your own.  That problem’s already been solved by somebody you ought to be able to incorporate.  But in general, how do you – what kind of technology do you use for something as impressive as this?

ADAM HOFMAN:  We built the model that we’re currently using, using Python language.  And we researched a ton about what algorithms are used in natural language processing, and which ones are proven to work, and which ones can we experiment with and combine with other proven ones to get the model that we really wanted for our customers.

And from the technical side, the brains behind all that is the main developer that I work with.  She’s probably one of the smartest people at CallRail.  And I don’t know how any of this would have been possible without her.  But she really took on the brunt of that challenge and said, you know, I’m going to research.  I’m going to learn about this.  She was at one point taking three different AI and machine learning classes, and also studying for her master’s, and also working at CallRail.  Like it was just like when do you – it was like a when do you asleep? kind of thing.  But, yeah.

So primarily we built it on Python.  We actually started experimenting with multiple models throughout the different verticals that our customers were using, whether it was, you know, whether they cared about appointments, or whether they cared about requesting like a free trial of a product or quoting for a specific price.  We experimented with that and then started to realize that, like, we could possibly just build a single model.  And now we’re really experimenting with that some more around that.

ANDY CROWE:  So most of our listeners are involved in project management in some way.  Most of them are project managers.  And when you look at something like this – so it’s interesting.  I’ve got two friends that own companies that they’re not exactly what you do by any stretch, but they both – one is OrcaTech and the other is DocAuto.  And they both analyze large bodies of data.  Most of these are written documents, you know, and they analyze them and try and apply machine learning.

And machine learning, just for anyone who maybe isn’t familiar with it, the idea is just that the software gets smarter, the more it’s exposed to – it can learn from mistakes.  It can incorporate new information.  So that’s more or less what I’m talking about when I talk about artificial intelligence.  And so this idea, as they’re approaching these big projects, it’s fascinating.  Both of those companies have done Agile development and have chosen an Agile approach.  And it sort of fits with that.  What has been you guys’ experience with incorporating Agile?  And if you care to incorporate on the “ish” part.  What makes it Agile-ish?

ADAM HOFMAN:  I mean, we use – we have multiple teams throughout the whole product.  We have multiple teams that all focus on a specific aspect of the project, or of the product.  And my part is the Conversation Intelligence suite of products that we’re building out.  So how that looks from a project management standpoint is really starting the project, being able to work it to a point where it’s in a – from an idea phase to a design phase.  And then from there we have five different stages of the project.

So it’s in design.  It’s in, you know, once it gets handed off to a developer, it’s in development.  And then a review stage, a QA stage.  And then, finally, just a part where we’re 90 percent done, we’re almost ready to ship it, and then code review and all that stuff.  And then finally shipping the product.  I think what makes it Agile-ish is that, like I said before, we’re constantly kind of adjusting how we do projects and how we can kind of ideate our own process to make it – to improve it.  And we’ve gone through, you know, we’ve gone through times where it’s taken six months to do a simple project, to now where it takes two minutes.

ANDY CROWE:  So, question for you.  In a true kind of Agile environment, the customer is embedded with the team.

ADAM HOFMAN:  Right.

ANDY CROWE:  Is that the case in your environment?  Or do you represent the customer from that standpoint?

ADAM HOFMAN:  Yeah, I mean, it’s a little bit – it’s a little bit of both.  So we try our best to talk with customers as much as we can to get an idea of how they would use it.  We show them designs of possible ways that we’re going to build this out and get their opinion on things and really just get their input on how they would use it; and primarily, like, if that would be successful for them.  And then on our side we do make some assumptions and go, okay, we think that the customer is going to use it this way.

So that’s kind of how we begin to build it.  And now we’re getting more into kind of doing more and more customer research from that standpoint, which I’m really excited about because our assumptions are only as good as what we think.  And those customers that we talk to, they might just throw something at us that we’re going, “Oh, we never thought about that before, but that would be awesome.”  And being able to put that in the product so that more and more customers are excited about using CallRail has been awesome.

BILL YATES:  I’ve got a follow-up on that.  How do you, I mean, I think we all struggle when we’re dealing with projects that involve a product, something that we’re going to build out that has features and capabilities.  We all struggle with knowing when to cut it off, you know, what features are going to be the ones most used, and at what point are we just adding technical debt, featuritis, and losing control of our product?  How are you able to measure those features that are delivering most of the value to the customer?  Is that an ongoing conversation, or are there metrics that you can use?

ADAM HOFMAN:  Yeah, so there are metrics.  And we talk to customers every day, whether it’s from a support side or sales or, you know, our customer success team talks to those customers every day who are using those things.  And if they have feedback on something, they’re going to give it to us, which is what I love about our customers is that they’re not afraid to say, like, this doesn’t work for me.  This is how it could work for me.

But from a project standpoint, we have various flags in most of those features, the big features.  We have flags that we just, from a development standpoint, can put on a product and go, okay, somebody’s actually using this.  And we can look over time about how many people are using it?  What’s the growth?  And with this, with CallScore, which is our machine learning and call analysis platform, with that we decided at the last minute to make it free for everybody, which was pretty crazy from a project standpoint to just all of a sudden change that.

BILL YATES:  Yeah, do you still get paid, Adam?  Are you a paid employee?

ADAM HOFMAN:  Yeah, luckily, yeah.

BILL YATES:  Okay.

ADAM HOFMAN:  But we decided to make it free, which was awesome.  We’re the only company in the industry who does that.  And that was a challenge from a project standpoint because we had it all built, and we had all the pricing.  Like we had that all embedded into the code.  And for at the last second to be like, hey, by the way, we’re making it free, you know, you just – okay, all right.  And you’ve got to deal with that.  But it’s been awesome.  Ever since we decided to do that and decided to market it very heavily, we’ve seen so much more uptake on how many people are using it.  We went from 1 percent to, like, 7 percent of customers who are using it.

BILL YATES:  Oh, wow.

ANDY CROWE:  Just a note along what you were asking, Bill, with that whole idea of featuritis, Adam said something earlier about, you know, when we’re 90 percent sure.  And it just reminded me of an old quote that 90 percent of something is better than 100 percent of nothing.  And, you know, there are times when it’s just time to get it out in the customers’ hands.

ADAM HOFMAN:  Yeah.

BILL YATES:  Right.

ADAM HOFMAN:  Right.

ANDY CROWE:  Start letting them interact with it and see how they use it and what works and what doesn’t.

ADAM HOFMAN:  Yeah.

BILL YATES:  Mm-hmm.

ADAM HOFMAN:  Yeah.

BILL YATES:  And I would imagine, with machine language and artificial intelligence, you’re able to get smarter and smarter with your product; right?

ADAM HOFMAN:  Yeah.

BILL YATES:  Because of the metrics you’re able to capture and the conversations you can have with your customers, you’re able to feed that back.  There’s like a feedback loop.

ADAM HOFMAN:  Yeah, yeah.

BILL YATES:  How does that communication work?  Because you’ve got really stupid customers like me on one end, and then you have brilliant people that are kind of on – they’re in a special room.

ADAM HOFMAN:  Yeah.

BILL YATES:  And you’re kind of stuck in between.  So how do you manage that conversation?

ADAM HOFMAN:  Our feedback loop currently is the more, I mean, the more and more calls that come in, the more we can learn from those, the more that the model can learn from that.  But also our hope is that, like, if people, if a customer rescores a call, so say we score it as no, it’s not a lead, but they score it as yeah, that was – they think it was a lead.  We use that rescore to basically feed it back into the model and really just improve it from there.

ANDY CROWE:  I think I’m a little bit in denial that every phone call we make, every click we do on a website, is going to be recorded in a database, every place we visit.  And it’s all going to be vectored in and triangulated for some purpose.

BILL YATES:  That’s right.

ANDY CROWE:  Adam, I’ve got a question for you, back to your project management and your methodology and your approach here.  What do you see as maybe your biggest challenges from a project standpoint?  Where have been the things that you’ve run into and struggled with?  Where have you learned and grown the most since moving into this role?

ADAM HOFMAN:  I think from a mere technology standpoint, I mean, there are a lot of things going in that I was not – that I didn’t have knowledge of.  So there were a lot of things like artificial intelligence as an idea and as an industry that I wasn’t comfortable with and that I didn’t, you know, and my discomfort came from me just not knowing about it.  So being able to actually, like, really dive down and learn about those things as much as I could, as much as my brain could handle, you know, I’m not a guy who knows a lot of math, who knows like the algorithm portion of it.  But I know that somebody does.

And being able to really – I think the biggest challenge for me was being able to know that there is no way that I’m going to be able to learn this, but there is somebody that I can ask.  And being able to be willing enough to ask questions and to learn from the questions that I ask and to truly listen to the people who had the expertise in it was what made this project successful.

ANDY CROWE:  You know, I’m mentoring somebody right now.  And one of the things that I’m trying to teach this individual is ask questions to the point where you’re comfortable.  And don’t be afraid of sounding ignorant or stupid because that keeps a lot of us from asking basic questions.  We sit in a meeting.  We keep our mouths shut because we don’t want to look silly.  And I’ve kind of built a brand as being the guy that just annoyingly asks questions, you know, sort of the Socratic method of project management.  Just keep pestering until I understand it.

And that has served me well.  I’ve been in so many situations, so, so many where I’ve asked a question that, you know, kind of gotten past that point of discomfort.  And then other people in the room go, yeah, I was wondering that.  I was wondering that, too.  And that always validates.  But I’m glad to hear you say that, that, you know what, I may not understand it, but I’m not afraid to ask questions.

ADAM HOFMAN:  Right, yeah.

ANDY CROWE:  Keeps a lot of people in the dark because they are.  We’re all intimidated, and we’re all afraid everybody’s going to figure out how dumb we really are, And the truth is we’re all figuring this out.

ADAM HOFMAN:  Yeah.  And being okay with something just changing, like randomly; of you just being, like, all of a sudden it’s something different than what you thought.  Or being able to kind of overcome adversity in the way of, like, things that are unplanned are going to happen.  And being able to take that in stride and still maintain your head from a project standpoint is huge.

NICK WALKER:  I’m curious.  Could we go in a little deeper on that?  You know, from the whole start to finish on development, I know what they say about hindsight.  But any surprises along the way?  Anything that you just say, “Boy, if I’d known then what I know now?”

ADAM HOFMAN:  Yeah.  I mean, if I know, yeah, if I knew then what I knew now, it would have gotten built, like, a lot faster.  And but that’s, like, that’s the whole – I think that’s the whole, like, artificial intelligence and machine learning world is that a lot of these things are not – a lot of these things aren’t proven.  A lot of these things are brand new.  So, you know, for somebody to do it well, or somebody to do it a little bit better, is a win.

And I think, you know, I definitely think that, if I knew what I know now, that this project would have been even more successful.  You know, I’m definitely happy with where it is, with where it is right now and where it’s going.  But to be able to know how a customer wants to use it before the fact, and to really be able to dive down and get that information, if I would have known that right from the gate, it would have been so much more effective.  Yeah.

ANDY CROWE:  Isn’t there a country song about I didn’t know then what I don’t know now?  And if there’s not…

ADAM HOFMAN:  There needs to be.

ANDY CROWE:  We’ve got another, yeah, we’ve got another thing we can do.

NICK WALKER:  Well, I was thinking of a Martin Mull song that says, “I just button my lip, and they all think I’m hip.”  But that doesn’t really work in this scenario.

ADAM HOFMAN:  Yeah, yeah.

NICK WALKER:  Well, Adam Hoffman, we thank you so much for spending time with us today on Manage This.  It’s likely, though, that some of our listeners would be interested maybe in trying out CallRail’s Conversation Intelligence products.  How can they get more information?

ADAM HOFMAN:  Yeah.  So CallRail.com, just how it sounds.  Yeah, CallRail.com is probably your best source of information.  Of course we have, like, Twitter; and we post a lot of stuff on Instagram about fun things that we do.  And we have an entire page for Conversation Intelligence on there.  We do have, like, a free 14-day trial that a lot of our customers take advantage of.  And just, you know, our sales team is awesome.  And if you just call in to them, they’ll be able to answer any of your questions about any of this.  And if they’re unable to answer a question, they’re able to find it really, really quickly.

NICK WALKER:  Well, Adam, once again, we thank you for sharing your expertise with us.  We have a gift for you.

ADAM HOFMAN:  Oh, awesome.

NICK WALKER:  And like you, we thought about charging for this, and we had all the pricing information in place.  But we decided to make it free, and that is this coffee mug.

ADAM HOFMAN:  Nice.  Awesome.

NICK WALKER:  You can take that home with you, and it’s yours.

ADAM HOFMAN:  Yeah, that’s great.  Thank you.

NICK WALKER:  Well, thanks again, Adam.  Andy and Bill, as always, great to benefit from your perspective.  And listeners, is anybody looking for PDUs?  Who’re you going to call?  Us, of course.  Your time with us means professional development units toward your recertifications.  To claim your free PDUs for this podcast, just go to Velociteach.com and select Manage This Podcast from the top of the page.  Click the button that says Claim PDUs, and just click through the steps.

That’s it for us here on Manage This.  We hope you’ll tune back in on May 2nd for our next podcast.  In the meantime, you can visit us at Velociteach.com/managethis to subscribe to this podcast, to see a transcript of the show, or to contact us.  And tweet us at @manage_this if you have any questions about our podcasts or about project management certifications.  We always like hearing from you.  That’s all for this episode.  Thanks for joining us.  Until next time, keep calm and Manage This.

Leave a Reply

Your email address will not be published.

{"cart_token":"","hash":"","cart_data":""}