HRExaminer Radio Executive Conversations Badge Podcast Logo

HRx Radio – Executive Conversations: On Friday mornings, John Sumser interviews key executives from around the industry. The conversation covers what makes the executive tick and what makes their company great.

HRx Radio – Executive Conversations

Guest: Bobby Kolba, VP of Engineering, Textio
Episode: 320
Air Date: March 29, 2019




Important: Our transcripts at HRExaminer are AI-powered (and fairly accurate) but there are still instances where the robots get confused and make errors. Please expect some inaccuracies as you read through the text of this conversation. Thank you for your understanding.

Host: John Sumser, HRExaminer
Guest: Bobby Kolba, VP of Engineering, Textio

Full Transcript with timecode

00:00:14:12 – 00:00:30:12
Good morning and welcome to HRExaminer’s Executive Conversations. I’m your host John Sumpser and today we’re going to be talking with Bobby Kolba who is the vice President at Textio. So, introduce yourself please.

00:00:35:29 – 00:00:51:10
Yeah. So my name’s Bobby I’m the VP Engineering here at Textio I lead the engineering data science and technical program management teams. I’ve been with the company since the very beginning since the founding of it and they’ve helped us grow over the last four and a half years.

00:00:53:00 – 00:01:06:16
It’s been quite an adventure. So. So how did you get here? You didn’t wake up one day and say what I want is go be the VP of engineering at Textio you know so what’s the sequence that got you here?

00:01:06:17 – 00:02:52:03
Yeah, so it starts, you know, in a like, where all good tech stories start with theater production. That was my springboard into the tech industry. So I was a I would say a hobbyist and tinkerer with computers since a very young age I was very lucky that you know that my dad was into computers. I remember you know playing on a Tandy 1000 with my brother when I was very young and I so I always had sort of technology computers as part of my life. And then when I got into high school and college I got very very involved in theater production lighting design things like that. And so my major is political science which has nothing to do with either theater or my job but I spent a lot of time in college producing large shows building building productions and then I had a friend who in a very serendipitous conversation introduced working at Microsoft as a possibility for me. And I was able to sort of bridge all the work I had done in theater and what it takes to go from a script and a set of actors and a stage manager to opening night and realize that that’s basically the same thing we do with software you you have a spec and some engineers and you get to a ship date and so that led me to Microsoft where I worked originally an office on enterprise software and then moved in later to work on search. And so I got some exposure both to the enterprise side of the world as well to kind of the more machine learning side of the world. When I was working in search that’s also where I met Karen Hughes the founder of text you. So when she decided to go off on this adventure she knew me and asked me to join and I couldn’t say no yeah she’s a hard person to refuse isn’t she.

00:02:52:09 – 00:02:59:25
So so for the people in the audience who don’t text Joe would you talk about what the what the company does and what your father said.

00:03:01:03 – 00:04:34:02
Yeah. So tech steel is an augmented writing platform and what that means is you know with augmented writing you can see the future your words create and have a chance to change them before you post or publish or send anything. So as an example are our first application that technology is called textile higher which is two products textile for job posts and textile for recruiting mail. So as a very specific example I’ve hired a lot of engineers as we’ve grown our team from the five of us when we started to about 55 on the engineering team now. And of course that process starts with a job post. So when I write that post with text yo I can see the effectiveness of the language I’m using and any unconscious bias that may be seeping in. And then I can change it before I put it on our career site where it’s going to get picked up by crawlers and reposting across the Internet or I pay to put it on a job board. Or maybe I get kind of one shot at it and so the result is you know buy it by having this much stronger job post. I see a broader set of qualified candidates. And then once I sort of have that up there I can also start reaching out to people and so that’s where our product texture GOP recruiting mail comes in. Which is integrated into linked in an email where as I’m writing sort of outreach mails to potentially passive candidates I’m getting feedback about the language that’s most effective in getting people to respond and to respond positively so not just that you know please stop contacting me but you know to get them interested in the job that’s interesting.

00:04:34:03 – 00:04:35:20
That’s interesting so.

00:04:35:22 – 00:04:54:28
So you say that this helps with you really mean very specific kinds of unconscious bias not all unconscious bias in general but that is bias that leads to discrimination or regulated areas where bias is prohibited.

00:04:55:00 – 00:05:21:06
Zahra Yeah I think that that’s a way of seeing what I mean. Bias is such a broad term that you know you can have a bias. You know I’d say from my background of working at Microsoft there were probably a lot of things I had to unlearn that might help is bias to corporate jargon and things like that which is something that the platform looks at. But generally when we’re talking about bias and we are talking about sort of unintentional discrimination got it got it.

00:05:21:09 – 00:05:36:04
OK so so you use A.I. in this process what thing resembling a I always get befuddled by what people were calling so. So tell me a little bit about the technology that you use to get this done.

00:05:37:14 – 00:07:11:16
Yeah. So like you said I had a pretty overloaded term and I think you know a way that it’s talked about Indian shit is that sort of the A.I. revolution is like electricity it’s sort of this you know empowering commodity that it’s coming and rapidly emerging. And so I really is the foundation of what we built in Mexico the subfield of A.I. that were most closely associated with this natural language processing. So how do you take the text and the meaning behind text and turn that into something that computers can reason over and we can make judgments based off. And so the way that we implement that is really not in a a one size fits all solution. So if I’m trying to predict a classic example is this piece of email spam or not spam that’s a that’s a single a model it’s making that decision. And what we do a text of that I think is really special on the technology is used lots and lots of different A.I. approaches and then blend them together seamlessly into a single writing experience. So when we’re looking at you know what kind of document. That’s one kind of model or what type of role is this hiring for. That’s another kind of model. What was the sentiment of an email response separate pieces. What language to be most effective in this context specifically. And then we pull that all together in a way that is really transparent to sorry users so that they don’t feel like they’re using a you know a product. It’s just the A.I. is there empowering they’re all they’re sort of human activity of writing.

00:07:11:22 – 00:07:50:18
That’s that’s fascinating so. So I think what I heard you just say is that that that in the system one of the one of the modules you have in the system is sort of a model of how a conversation reaches its conclusion on a specific topics. So so we’ve got some emails some care of over and over the course of the exchanges it’s building towards an interview or deciding the. So there’s some process map and better do this.

00:07:50:18 – 00:08:39:08
Is that a reasonable way to think about the sort of matching and the identification with that you do for customizing the advice you give about overwrite Yeah I think the phrase use module is a really good way to think about it of sort of did that kind of fan out to ask a bunch of different technologies different questions and then use kind of the ensemble response from that to to pull together an entire experience and so I’m just like you use different parts of your brain in different scenarios that tackle a different problem that the same thing that we’re doing with our technology. So it’s really about blending all that together. It is really I think the magic part of it.

00:08:39:27 – 00:09:21:12
So I’m getting this picture of a of a of a conference room with a whiteboard in the room on the whiteboard is kind of a it’s not a country it’s more like a flowchart with 10 mins likelihood gates in the flowchart that is starts from I don’t know you at all. 2 you’re hired and at each sort of step in that process the advice that one gets about the language that you use to take the process forward shifts to to match the stage of the process.

00:09:21:13 – 00:09:29:21
Are you there or is that just be whoever your fancy this word I think that it’s a good way to think about the two products that we have. So

00:09:29:21 – 00:10:15:07
So if you think about textual for job post that’s sort of applying augmented writing to kind of a top of the funnel problem and then when you think about textual for recruiting mail that’s kind of that next block on that flowchart that I think you’re talking about which maybe is mapping a hiring process and then you can imagine we’re not there there yet that you could apply that for interview feedback for performance reviews for a communications internally within a company. You know I think the vision that we have for text Deo is really to bring augmented writing to communication across the business. And so I think we’re really just getting started with what we have out there today and definitely plan to continue building towards that vision over the next years that’s everything so.

00:10:15:09 – 00:10:15:26

00:10:15:29 – 00:10:32:11
I guess it does something about the working in office with you know is there is there a group or a person in your R and D function who is thinking about what are all the type relocation that happened inside of the organization Yeah.

00:10:32:12 – 00:11:05:08
Certainly at the top of mind thing for for our product team and you know like you said we have a mix of folks here in the company I think you know one to one thing that’s really important process to build a team that represents a huge diversity of skill so that we can tackle these problems because you know we we have a set of problems we know we’re solving today and there’s going to be a difference at six months from now and so we want to have the most flexible team possible so product and ah ah R leadership team is it’s always thinking about where can we apply this technology what’s what’s best for our customers.

00:11:05:08 – 00:11:36:03
Oh that’s that’s fascinating. So do you imagine that that evolves from adjacency to recruiting or or are you going to go from get rid of driving into the recruiting process and jump over to getting or without writing into all eating agenda development processes. Do you have a sense of where the map takes you yeah.

00:11:36:03 – 00:12:29:24
I think if you if you look forward far enough it does take us into all business communication. The specific ordering and staging of that comes down to priorities and and the team that we have at any given time. We’re investing sort of our energy and our resources on the engineering team into supporting our existing customers what do what are they saying. What do they need. Are they running into issues. Are there new things we could be building for them at the same time we’re building to advance the division of augmented writing and you know in the early days of a startup you’re you’re small enough that you sort of have to pick one or the other as the team’s grown it’s enabled us to turn that on multiple time horizons at the same time. So some of the work we’re doing is for this week or this month and some of it is for this year or next year and we can work on that simultaneously.

00:12:29:24 – 00:12:36:01
How big is Texas you are. I don’t have a was associated with the employees. How many customers.

00:12:36:09 – 00:13:03:26
Yes it’s Tokyo started in 2014 late 2014 with five of us. And over the last four and a half years we’ve grown from that original group of five to one hundred and thirty five and the company and then among that fifty five is on the engineering team split between front end engineering back and engineering data science and technical program management.

00:13:05:03 – 00:13:51:07
So this this is probably a couple hundred serious customers at this point. I don’t have a good sense of the scale it’s order 1 hundreds of customers and thousands of companies that have tried the product at this point through a pretrial That’s interesting so scaling being able to deliver the right was to the right person without it appearing to be a latency problem must must take a lot of your time as as you as your breath increases in your depth increases. Being able to deliver that real time feel is an increasing challenge. How’s that going.

00:13:52:20 – 00:13:58:28
Yeah I guess so I think about this the way that I used to think about lighting design actually going back to my deep path.

00:13:59:00 – 00:14:41:23
So lighting design is one of those things that when it’s done well you don’t notice it. It just sort of seamlessly enhances whatever whatever skiing you’re looking at. And as a lighting designer that’s what I was always striving for I was never trying to show off with my lighting design. It was always more subtle than that. And that’s the same with reliability and infrastructure engineering that nobody notices that unless it goes wrong. If we do our job well and so building in that resilience building in that performance and having you know that the systems and the property to maintain that so that as our company is growing and we’re bringing in new customers I’m staying ahead of that so that I don’t have to worry that you know.

00:14:41:29 – 00:15:03:09
Oh this week we closed a large deal and a bunch of new people are going to be onboarding to the system. I have confidence that the system of scale that that’s been something that has been front of mind for us since the earliest days of the company and continues to be one of the surprises of you know your tackling an area.

00:15:03:21 – 00:15:58:18
So I would say before you started tackling the area and nobody existed and and and the the core of the product as I understand it is the capacity to have your desktop tell you how to improve the quality of your posting or your liver in doing so that you get something that approximates the results you’re after. In particular in the job posting stuff it begins the right mix of responses that overcome internal biases towards one glory and this gets you a better mix of gender and a better mix of ethnicity responses. What are the pages what have you learned.

00:16:00:19 – 00:16:08:24
That is a very very broad question. You know I think there’s both building the products like you’re talking about and building a company.

00:16:08:24 – 00:16:40:09
I mean this is the first time you know my past was always at large companies with established infrastructure. And so for me personally there’s been a lot of learning about how a company operates one of the early projects I worked on when we were quite small was taking our initial private beta and bringing it to commercial availability and to do that we had to have a way to collect money. And I knew very little about how to build a billing system how to sort of go into these large companies and sell to them.

00:16:40:24 – 00:17:28:00
And what that motion looks like and so and how I could then encode that in the software and how we supported it. And so each I would say every month I’m learning something new as as the company has grown and I think you know there’s one of the I guess the delights has been to see how successful it’s been to take all of this deep deep data and knowledge and intelligence we have about language and by wrapping it in you know a writing experience. It’s so beautiful and intuitive that you don’t have to be a data scientist to use it. It’s just the response from our customers from our users. You know I see people talking about it at conferences or on Twitter it’s been really gratifying that that’s interesting so.

00:17:28:03 – 00:17:34:01
So the interface design is one of the the most extraordinary parts of Texas.

00:17:34:01 – 00:18:21:21
You know we haven’t really talked about that that much is the interface design the result of a background in theater and lightning we I have to give full credit to that to Jensen who’s our co-founder and CTO and I think you can you can really look at our our founders for the bones of the company. So from here in with her background and computational linguistics you really get the brain of the textual platform and then Jensen who has an incredibly deep deep expertise and is a world class an enterprise user interface design. You get the packaging that that makes that so successful and so I think the you know the blending of of their two skill sets is really what set us up for success.

00:18:22:23 – 00:19:10:27
That’s that’s that’s interesting. So. So how does that how does that work in practice. I mean as as you get smarter the pressure of interface to be simpler has got to be pretty tawdry and the tendency for the interface to get complicated is is one that the good the growth can be channeled right. That’s one of the things that you often see in a company like yours that scale this space. Is that the way that people talk about what the company does gets complicated the way that you interact with the product. It’s complicated. That’s sort of the opposite of what you’ve stood for so long. So how do you keep that in check.

00:19:10:27 – 00:20:07:28
I mean I think it starts with core values. So having sort of a belief in and craftsmanship having you know like you’re saying understanding that simplicity inherently has value and knowing that that design is is art more than science. You know I think there are there are approaches that are very sort of rigorous science based where you end up with with overly confusing UI. Each little button tested well versus stepping back and understanding what is what is the customer experience. How do we take the UI and get it out of the way of people especially when you’re doing something so fundamentally human it’s writing. So you know a couple years ago we rebuilt our entire editing experience from the ground up so that we could have it be as simple as possible so that the writers could focus on the words and not all of the UI around it and every feature we add we’re looking at you know is this adding to the experience or is it just adding confusion and adding confusion.

00:20:07:28 – 00:20:36:22
We won’t ship it you’ve just said something that might be the favorite thing that I’ve heard this year and that is that to just to sort of synthesize what you said we we we are rooted in a set of values and then ship is amongst them. And I I’ve been asking people in the technology business questions for 25 years I’ve never heard anybody say that it’s genius.

00:20:36:27 – 00:21:33:13
So talk to me about craftsmanship I think it starts with caring about the thing that you’re building and I mean I think we all have that experience in the real or you you use whether it’s an object or a service or something that you can feel the care and the intentionality that went into the design of it. And it takes work and it’s hard and it’s it’s worth that. I think what you get. On the other side of it it’s so much better than than if you didn’t do that. And so that’s just something that you know very practically if we’re shipping a piece of software and there’s a small bug I think a lot of companies you know maybe we’re something has shifted in the UI a lot of companies might say well we need to hit this date so we’re going to not fix that. But that’s the kind of thing that we would look at and say you know we we aren’t going to be proud unless we fix that bug and ship it. And so it’s time that intentionality and sort of pride in what you’re building and having that infused through your development processes and how you think about your products.

00:21:33:19 – 00:21:54:18
I love this. You should you should plan reviews of the web. This is this is this is a really really wonderful thing to talk about. Now you’re in the you’re in the presentational bias in mediation of bias. This is sort of the ethical issues that matter to you guys.

00:21:54:19 – 00:24:20:26
Yeah I mean I think the first one which you know I think you’re going out is any A.I. system you have to be cognizant of bias and and really like many things the sort of output of one of these systems is only good as the data that you put into it. So it’s input data it’s fundamentally biased your output is not going to be is it going to be biased and not useful. So you know as an example of that if we get a data set that has aggregate statistics about gender identity you know how how many folks who identified as men apply to a job versus identified as women or other. If that data is based on the individual’s self identifying that’s really high quality if that data is based on you know maybe some algorithm that looked at the name and tried to guess that’s fundamentally biased data and we’ll reject that we just won’t use those fields because we don’t want that bias to seep into our system. So I think being mindful of that and it’s not something that you can sort of solve and say you’re done with It’s a constant vigilance that has to be built into our processes. I think also the responsible use of data you know I think we’ve seen with some of that Facebook stuff recently and actually some of the some of the work I did in being on privacy it reminded me of that that any company that’s doing data work has to be transparent about what they they are collecting. Why are you courting and how are you using it. And don’t use it for things that you say you’re not. So you know I think that we don’t we sell our data we don’t use it to power targeted advertising and things like that. I think there’s a lot of ways that people get in trouble with data. And if you can just have a human conversation about yes you’re giving us the data and this is the value you’re getting back that the much better footing to be on ethically. And then I think finally the there’s this sort of fear of A.I. replacing humans. And so the way we think about what we’re building is really to hold a mirror up to what you’re writing that when you write with textual you feel like you have a superpower you’re not it’s not letting the machine write for you and just you know type in a couple feet is a meditative click a button and get the best job description has to be. Fundamentally you and so lending that power of all that data that we have but leaving the human in control I think is important for us.

00:24:20:26 – 00:25:23:27
One of the things that I like most about the textual approach is is a lot of a lot of the emphasis in bias mitigation in hiring processes boils down to finger wagging people for being biased and rather than wag your finger at me biased you simply focus on the outcome I am trying to achieve with regard it’s about achieving the outcome without it ever being a question of am I a screw up because I’ve got bias. So I think that I think that’s an interesting approach to solving bias that you don’t but you don’t inherently attempt as the primary thing to change people. So what you look to do is change the behavior that results in a consequence that’s different from the one that want is that. Is that something. Is that something that’s just there their Yeah.

00:25:23:28 – 00:25:49:26
I mean I think for me it it’s pretty simply this belief that people are fundamentally good. And if you just hold up a mirror to them and show them these unconscious mistakes they might be making they’re going to want to want to act on it. And so you know we’re not like you said trying to sort of finger wag. We’re not trying to judge people. We’re just trying to show them what’s in their language and then trust the good human they’re to decide to act on it.

00:25:50:00 – 00:26:05:18
So last question. You’re in Seattle. There is of enormous pressure on the labor market in Seattle to compete for the very people that you need. How do you how do you success for talent work.

00:26:07:18 – 00:27:49:26
Yeah. So first we use our own product as you might imagine. That’s that’s step one just a little plug there. And then you know I think there’s a couple of things we’ve done to to be successful. So one is really just building a great place to work and having a company that really is values focused and empowers folks just to solve heart problems. Like if there’s a you know you can have the best recruiting program in the world but if people get your company and it doesn’t live up to those promises that’s not a that’s not a successful hiring plan. So it’s really important to build that sense of belonging for everyone here and then I think a specific thing that we’ve done is you know we we for engineering is looking beyond a small set of companies or schools. So we’ve had a lot of success with what we call career changers who may have entered tech and engineering through a non-traditional path. So if I look at my team you know I have folks who used to be aviation technicians physical therapists clinical psychologists physicists all who did that and then decided to go to a boot camp to be self-taught. And then we’ve been able to find them and I think some of that just comes from putting the work into the process. I still remember before we ever had a recruiting team looking at the path of inbound we got off of indeed. And I would look at every resume and I would go to the portfolio sites and I would look at GitHub and it takes a lot of work but I didn’t trust a résumé scanning tool because it was going to filter folks out and it would be biased. And so I think some of it is just putting the work in and opening your horizons of who you would look for and trying to avoid some of the classical traps.

00:27:50:00 – 00:27:54:27
That’s great. So last thing In a nutshell what makes the company different?

00:27:56:22 – 00:28:35:22
I think you know we want to build a great product and build a great business. Those are really important but we also want to build a different kind of company that doesn’t look or act like a stereotypical tech startup. And you know I think that starts with not being afraid to challenge our assumptions and figure out what does it look like to do with the textiol your way whether that’s how we host our events how we think about career growth build our software design or our physical office space. And I think that you know going back to the conversation we’re having about craftsmanship that intentionality around the kind of company we want to be really infuses the way that we operate across the entire team.

00:28:36:10 – 00:28:43:16
What a fantastic conversation. Thanks so much for taking the time to do this. Would you reintroduce yourself and tell people how they might get ahold of you?

00:28:46:03 – 00:28:59:02
So my name is Bobby and you can find me on LinkedIn. You can follow textio on Twitter. We have a pretty active Twitter there. And then you can catch me at Bobby at textio dot com.

00:28:59:04 – 00:29:20:23
All right thanks again for doing this Bobby. We’ve been talking with Bobby Kolba who is the vice president of engineering at Textio an augmented writing firm in Seattle. You’ve been listening to HRExaminer’s Executive Conversations and we will see you back here next week. Thanks for tuning in. It’s been a blast as usual. Bye bye now.


Read previous post:
When Should Recruiters Trust Their Gut?

"We know from experience that after a rigorous interview process, almost half of the people who ultimately get the job...