Tim Klever about the evolution of the role of developers in the AI era

0:01 All right. Um, so, uh, I wanted to have
0:05 you on this episode because, uh, we’ve
0:07 met multiple times, uh, in Phoenix. Uh,
0:10 we’ll talk about that in a few minutes.
0:13 Um, but first, uh, for, uh, for our
0:16 audience, please uh, introduce yourself,
0:19 tell us a bit more about your
0:21 background, your career path, uh, what
0:24 you’ve been focusing on.
0:27 Yeah, totally. Uh, you know, um, I like
0:30 to keep it simple. I like to just say
0:32 that, uh, I’m Tim Clever and I make
0:35 software. Um, I’ve been lucky enough in
0:37 my career to,
0:40 uh, make software at institutions both
0:43 large and small, ranging from three
0:45 people to 70,000 people, you know, giant
0:49 enterprises. Uh and along that way I’ve
0:53 been uh fortunate enough to sort of
0:55 mentor lots of software developers. Uh
0:58 get really into the weeds on software
1:00 development and the mechanics of
1:02 building software. Um and my focus is is
1:05 primarily in that area in in developer
1:07 experience and sort of like helping
1:08 organizations
1:10 uh make the best possible software they
1:13 can which again just sort of gets back
1:14 to that engineering root of I’m Tim
1:17 Clever. I like to make software.
1:22 Now because we have a past together, I
1:26 know from uh you directly that you
1:29 didn’t really start your career um in in
1:32 software, but you were a drummer in a
1:34 rock band. So tell us a bit more about
1:37 that and how does someone go from that
1:40 to leading developer experience in one
1:42 of the largest financial institutions in
1:45 the world.
1:46 Yeah. uh you know I am a technology as a
1:49 second career uh guy like so many people
1:52 are and and and uh much love to everyone
1:55 that comes to technology as a second
1:57 career. A lot of us come from other
1:59 places originally. Uh but yeah, I tried
2:02 to chase the rock and roll dream and it
2:05 actually has a ton of transferable
2:08 skills. Not that I knew that at the
2:09 time, but uh that it served me really
2:11 well. And honestly, that’s probably an
2:13 entire 30 minute podcast, you and me,
2:16 someday. Like I’ll have you on and we’ll
2:18 call it like the rock and roll engineer
2:20 or something like that. But uh both on
2:22 the business side and the product side
2:25 of sort of getting a small team or a
2:28 band down the road uh in an old van and
2:33 shleing gear and getting on stage and
2:36 getting off stage and selling t-shirts
2:37 and welling wearing multiple hats. Both
2:40 that side uh prepares you a lot for a
2:43 software engineering career as well as
2:46 uh the physical production of music.
2:48 When I think about the technical skills,
2:50 if you think about a large mixing board,
2:52 I don’t know if you’ve ever had to sit
2:53 in front of a 128 channel SSL, but you
2:57 know, what’s a mute button other than
2:59 sort of uh an if then? Uh you’re talking
3:02 about lots of channels. They need to be
3:04 coordinated. They need to be organized.
3:06 You’ve got a patchbay where you’re
3:07 moving. This is just a CI CD system. And
3:11 at the end, it has to come out the left
3:14 speaker and the right speaker and it has
3:15 to sound nice. So I mean in that way all
3:18 that logical gate I mean that prepared
3:20 me for a life in software and then you
3:21 know add in a tremendous amount of luck
3:25 uh and here we are.
3:28 So when people are now talking about
3:29 vibe coding you were in the vibe way
3:31 before it started.
3:33 Yeah. Yeah. Just it was it was I added
3:36 the coding to the vibes. Some people are
3:38 adding vibes to the coding.
3:43 that that’s definitely not a classic uh
3:45 career path, but uh I’m glad you you
3:48 joined the team and you know and and
3:50 you’re now part of this world. Um so you
3:52 you’ve been um in the past years working
3:55 for American Express. Um can you tell us
3:58 a bit more what does it mean to be
4:00 responsible for you know tens of
4:02 thousands of engineers building software
4:05 daytoday and their experience you know
4:08 what is in you know a few sentences what
4:10 is the responsibility there
4:12 yeah uh that role and and and I think
4:15 any DX role in an IND in in a large
4:18 organization or even a smaller
4:19 organization you you play this role is
4:22 like the great negotiator and there’s
4:24 three sort of like vectors pulling in
4:26 different directions that I like to
4:28 think about as you sit in the middle of
4:29 that. You have you have the business and
4:32 you have productivity and you have
4:34 delivering for your customers. So
4:35 important, right? Like work has to get
4:38 done, money needs to get made, things
4:40 need to get delivered. You have another
4:42 vector pulling um in a different
4:44 direction. You’re talking about risk,
4:47 security, safety, trust, uh long-term
4:51 viability, maintainability, all of this
4:53 stuff. And then you have the third
4:55 vector, the the quite frankly the most
4:58 uh enjoyable vector. Uh you’ve got
5:01 developer satisfaction and sort of just
5:03 employee satisfaction, right? Like just
5:05 everybody enjoying the work. You can get
5:08 a ton more productivity by beating
5:10 everybody with a shovel, but that’s
5:13 going to last maybe, you know, three
5:16 weeks before everybody quits and you’re
5:17 left with nothing, right? So it’s not a
5:18 very durable to let any one of these
5:21 vectors pull too hard. the the most
5:24 secure computer is one that’s often
5:26 unplugged. So if you let that vector
5:28 pull too hard, nothing gets done. If you
5:30 let the developer satisfaction vector
5:32 pull too hard, we’re all vibe coding
5:35 funtime personal projects and like we’re
5:37 running in a thousand D. So you sit at
5:39 the middle of th those kinds of
5:41 intersections of desires and you
5:44 negotiate to make sure frankly in any
5:47 negotiation it’s sort of the like if
5:48 everyone’s kind of mad at you, you’ve
5:51 done a great job. uh if everyone gave up
5:53 one piece sol you know when I think
5:55 about developer experiences to sit right
5:56 there at that and look at those three
5:59 vectors and just like trade-offs in
6:02 software isn’t life just trade-offs just
6:04 like trade-offs in software um you know
6:06 negotiate sort of the best outcome for
6:09 all of those interested parties
6:13 understood so um it’s just we are going
6:17 to talk about many transformations that
6:19 you’ve been you know orchestrating
6:22 Um but when you and you had a very large
6:25 overview of the SDLC right in in the
6:28 organization
6:31 when you looked at the organization and
6:33 at the SDLC what was clearly broken and
6:37 that’s kind of like a background we want
6:39 to set you know to discuss later on what
6:42 are going to be the the main uh
6:44 transformations you you would be
6:46 eventually running there. Yeah, we you
6:49 you know when we talk about broken um
6:52 I’d hesitate to ever say anything’s
6:54 broken. I know I’m not and I’m not
6:55 dodging the question there, but if you
6:57 look at any business that produces
7:00 software and is still in business,
7:04 they’re somehow getting the job done.
7:06 Like the business works. Um and this
7:10 this applies to small startups a and
7:14 large huge multinational organizations
7:16 but like at some level it works. Uh the
7:20 big thing I think we talk about when we
7:22 look at SDLC’s is waste and sort of like
7:26 how much effort, how much money, how
7:28 much resource and how much just sort of
7:31 like human heroics have to go into
7:35 making it work. And so are you running
7:38 at, you know, truthfully, and these
7:40 these things are hard to measure and
7:41 everybody measures them sort of
7:42 differently, but are you measure are you
7:44 running at 20% efficiency, right? And
7:48 are you leaving 80% on the table? And
7:50 you you’ve seen it, I’m sure, as you
7:52 talk to CEO or CEOs or CTOs, you know,
7:56 where they’re say they’re sitting there
7:57 saying, you know, my my my engineering
7:59 budget gets bigger every year and I feel
8:03 like I get less out the door every year
8:05 and I keep putting money in the system.
8:08 And sure, you know, we still have
8:10 customers and features come out and
8:13 like, you know, the thing is sort of
8:14 moving forward. Our business is viable.
8:18 uh like the SDLC technically works but
8:22 are you losing um 80% of your investment
8:26 85% of your investment you know how much
8:29 are you leaving on the table and that
8:32 that problem is not different between
8:35 large organizations and small
8:36 organizations I think you just see the
8:39 scale of it sort of exponentially grow
8:41 when you’re at a 330
8:45 300 person startup
8:48 um they have these problems too. Now,
8:50 when it’s just it’s just me and you
8:52 sitting in a room, we can sort of get
8:54 over some of these efficiencies by
8:55 turning around and like tapping each
8:56 other on the back or saying, “Hey, let’s
8:58 go to lunch and sort this out. We got a
8:59 problem here.” Right? You can’t do that
9:02 in a 20,000 person engineering organ.
9:05 So, that efficiency really really just
9:08 grows. We always talked about it a lot
9:10 like I don’t know if you’ve watched uh
9:12 television shows like like uh Kitchen
9:14 Nightmares or or Bar Rescue or or these
9:17 kinds of things where they come in and
9:19 they say you’re you’re pouring all the
9:22 liquor down the drain. You’re throwing
9:23 all the food in the trash or you’re b
9:25 you know all these things. It’s like you
9:26 are serving food
9:29 but you’re not make you’re not able to
9:31 like capitalize or make money on your
9:33 investment because you’re just throwing
9:34 in the trash. So like that’s that’s the
9:36 big thing I think uh when you talk about
9:38 an SDLC at at enterprise scale is is
9:41 fighting the waste finding it fighting
9:44 it.
9:45 Okay so uh productivity is definitely
9:48 you know one way to look at what you
9:50 just said how do you improve it and and
9:52 I think that was and it’s still uh of
9:56 course one of the the main promises of
9:58 AI right? Yeah.
10:00 And you’ve you’ve you’ve run different
10:03 processes before LLMs came to our live.
10:06 I know you you’ve been involved with
10:09 bringing GitHub to the organization and
10:12 and it’s a lot of processing when you’re
10:14 working at the scale of such
10:16 organization. But let let’s talk about
10:19 AI
10:20 and you can’t have a podcast and not
10:22 talk about AI.
10:23 Yeah. Exactly.
10:25 and but but mostly I think that what
10:28 interesting is to see the and to learn
10:30 from your experience about the friction
10:33 and you know challenges around that and
10:35 how do you bring AI at such a scale so I
10:39 know you started with copilot and you
10:41 know um I was curious to like learn how
10:45 you know how did developers respond to
10:47 that what was the first uh the first
10:50 impression impression that you you got
10:52 when you started introducing these
10:55 tools.
10:56 Yeah. You know, I think as you look
10:58 across um the industry and and at large
11:02 organizations where I’ve participated in
11:03 these kinds of things, I think I think
11:07 you really got to think about these
11:09 things like a product. You know, I was
11:11 talking with a a friend recently. Uh, I
11:14 won’t name the employer nor the friend,
11:16 but you know, like they got a mandate
11:18 from the top um that was basically like
11:21 here’s a bag of AI tools, do AI, we’ll
11:26 catch you later, right? And like it was
11:28 like it was that it was like do AI and
11:32 you’re like okay and and and and take
11:34 out the the magic and the fun of AI and
11:37 all the exciting things that are happen
11:39 and just like go back to, you know, old
11:41 school tools. uh go into a room of of 10
11:45 people, give them all power tools,
11:47 right? Give them all like saws alls and
11:50 just say do saws. You’re doing saws now.
11:54 And come back in an hour and just see
11:56 what kind of chaos you have unleashed.
11:58 They’re probably sawing in different
12:00 directions. Someone’s probably injured
12:02 themselves terribly. like the like all
12:06 of these things are probably outcomes
12:07 because you say well you just you just
12:09 lobbed power tools into a room and
12:11 walked away and expected magic and we
12:13 know that doesn’t work. Uh and so like
12:15 as you look at these things I think it’s
12:18 about
12:19 thinking about them like a product and
12:21 thinking about them like and packaging
12:23 them up and sort of saying like here’s
12:24 what we’re trying to achieve. Here’s how
12:26 you can be successful. Here’s how this
12:27 is going to improve your life. The same
12:29 kinds of product discussions you have
12:31 when you do anything. Um and then I
12:34 think you you you like all great things
12:36 you have to manage um you know community
12:39 right in inside an organization or even
12:41 in um outside of an organization change
12:44 is hard uh and you talk about people
12:48 being afraid especially with AI is it
12:50 coming for my job is this am I training
12:52 my replacement all of this stuff which
12:54 again I I don’t think to be true and I’m
12:56 I’m super optimistic I’m grandly
12:58 optimistic about the AI future but um
13:01 you know like If you don’t manage those
13:03 kinds of things in addition to sort of
13:05 giving people proven paths, ways to
13:08 succeed. We see this across the industry
13:10 right now. Um you see a lot of companies
13:14 uh turning off their sort of like token
13:16 maxing dashboards and things like that.
13:18 You’ve seen those kinds of things in the
13:19 news because again go back and say use
13:22 AI and so the bill went up and we saw
13:26 all this token usage. Um but then are we
13:29 getting better business outcomes? Are
13:31 our customers happier? Is there more
13:32 money in the bank? All those kinds of
13:34 things you got to think about if you’re
13:35 a CEO. Um, so like I think when you look
13:38 at those things, I think where you could
13:40 we can all do a better job is is
13:43 thinking about this from a product sense
13:45 even internally internally in your
13:47 organization and say I am delivering a
13:49 new experience. you know, think about it
13:51 with that product mindset of improving
13:54 people’s lives, outcomes that they will
13:57 actually benefit from, and then, you
14:00 know, hopefully adoption comes along,
14:02 right? And then again, you’re still
14:03 always going to have your your your
14:05 adoption curves like those are real
14:07 inside organizations as much as they are
14:09 outside of organizations.
14:11 Just double clicking on on this part
14:14 because there is a difference between
14:16 adoption and impact, right? And
14:18 right
14:19 and what you talked about is um how do
14:22 we we give the tools right the the
14:25 product to the people so they start
14:28 using it but in a way that is measurable
14:31 and creates impact for the company
14:34 right
14:35 can can you give us like one example of
14:37 something that you did or experienced
14:40 during this time that uh moved the
14:43 needle from just playing with and
14:46 hopefully not cutting any fingers with
14:47 but actually delivering value and how
14:51 did you like identify uh what to do and
14:55 how you you actually executed that?
14:58 I think you know I think that’s the
15:00 great challenge we have with um AI right
15:05 now as an industry. I think uh I’m going
15:08 to steal Kent Beck’s term here is like
15:11 we got this genie and it’s exciting and
15:14 it’s new and you know how engineers like
15:16 me and you new tools, new shiny things
15:19 are always so wonderful. We always dive
15:21 in, right? And it seems like in cycles
15:23 we dive in and we forget everything
15:25 we’ve learned since Bell Labs in the
15:27 60s. Then we have to kind of like
15:28 refigure it out. And when I think about
15:32 that, I think that’s the struggle we’re
15:33 actually on right now. Like we all have
15:35 this wonderful new thing and we lost the
15:40 collaboration, we lost the coordination,
15:43 we lost the orchestration and we all
15:45 became sort of like very siloed single
15:48 units and we know that siloed single
15:52 units firing in every direction does not
15:54 a cohesive product or you know
15:57 organization make right and so I think
16:00 that’s the really interesting space to
16:02 play in now where you’d say, you know,
16:05 we all got our feet wet with this. We
16:08 all experimented. We all saw how fast it
16:10 could go. We all saw some interesting
16:12 tips and tricks, right? Like go up and
16:15 down your LinkedIn feed and it’s pretty
16:16 much nothing but here’s a neat thing I
16:18 got AI to do yesterday, right? Which
16:20 again, great play and experiment. I
16:22 mean, so much innovation comes out of of
16:26 play. Um but moving into turning it into
16:30 business processes, turning it into
16:33 durable workflows that provide leverage
16:35 and say we want you to put you utilize
16:37 it in this way or we have orchestrated
16:40 it to do this here is the part you play
16:43 in this system here’s what you do right
16:46 finding out where humans add the most
16:49 value putting them in that position to
16:51 win and then letting agents provide
16:53 value where they provide value right so
16:55 I think as we arc across the industry
16:59 right now moving away from
17:02 um sort of like individual AI
17:05 babysitters
17:07 to
17:09 um you know orchestrators of automated
17:13 and agentic business processes is the
17:16 next cool leap and there’s a lot of
17:18 people working on that that’s not unique
17:21 to what I’m doing but I think that
17:23 really changes our focus maybe about
17:26 like what we want these things to do,
17:28 right? Like where are the real
17:31 bottlenecks and um again smarter people
17:34 than me have talked about how it’s not
17:38 making code. You know, you think about a
17:39 large organization
17:41 um you know with sort of I don’t want to
17:43 say limitless because everybody’s got a
17:45 budget but like you know these really
17:46 really large organizations have never
17:48 had a trouble making code. They they can
17:52 offshore things. We can hire contract
17:53 labor. You can have the you can hire up
17:55 as many people as you want, right? It’s
17:56 it’s the ability to make code is sort of
17:58 magical for the individual because now I
18:00 can run, you (0:00 – 34:12)