Latest / Chief in Tech Deep Dive Shortcast / Chief in Tech S1E51 - Palak Kamdar, Sowmya Sundararagavan, Oqeania Pepellashi & Ajda Gruden : Co-Pilot or Cautionary Tale? Navigating AI in Product Strategy

Transcript

Chief in Tech S1E51 - Palak Kamdar, Sowmya Sundararagavan, Oqeania Pepellashi & Ajda Gruden : Co-Pilot or Cautionary Tale? Navigating AI in Product Strategy · Chief in Tech Deep Dive Shortcast

• 0:00
  So what if I told you that the person managing
• 0:02
  this incredibly complex software architecture
• 0:05
  at a major European tech company, well, they
• 0:08
  don't actually have a degree in computer science.
• 0:10
  They have a master's degree in wood science.
• 0:12
  I mean, I know it sounds like a punchline, right?
• 0:15
  But it's actually the reality of where modern
• 0:18
  business is heading right now. The traditional
• 0:20
  tech resume is just being completely obliterated.
• 0:23
  Welcome to the deep dive. And if you are, say,
• 0:27
  prepping for a massive strategy meeting right
• 0:29
  now, or maybe you're just trying to cut through
• 0:30
  the relentless daily hype around artificial intelligence
• 0:33
  without losing your mind, well, today's mission
• 0:36
  is for you. We are going straight to the source
• 0:38
  to figure out if AI is your ultimate career co
• 0:42
  -pilot or, you know, just a massive cautionary
• 0:45
  tale waiting to happen. Yeah, and we are pulling
• 0:47
  from a phenomenal strategy panel. It's titled,
• 0:50
  appropriately enough, Co -Pilot or Cautionary
• 0:52
  Tale, Navigating AI in Product Strategy. And
• 0:55
  it brings together insights from four major product
• 0:57
  leaders who have, gosh, decades of combined experience
• 1:00
  managing everything from highly regulated fintech
• 1:02
  to these massive B2B software ecosystems. And
• 1:06
  the core tension running through this entire
• 1:08
  discussion. I mean, what really stood out to
• 1:10
  me is that artificial intelligence just isn't
• 1:13
  a shiny new feature anymore. It is completely
• 1:16
  reshaping how products are built from the very
• 1:19
  first brainstorm on a whiteboard. all the way
• 1:21
  down to how you actually talk to a customer.
• 1:23
  Exactly. It's fundamentally changing the mechanics
• 1:25
  of the job itself. It's not just a tool you use.
• 1:29
  It's changing what the job is. Right. OK, let's
• 1:32
  untack this, because the panel really breaks
• 1:34
  this massive shift down into three distinct philosophies.
• 1:37
  First, you have the optimistic adoption camp.
• 1:40
  And these are the folks who see AI as this straightforward
• 1:43
  magical enabler. Oh, yeah, the magic wand perspective.
• 1:46
  Right, exactly. And then second, there's pragmatic
• 1:49
  realism, which is all about how our actual messy
• 1:52
  day -to -day jobs have to evolve to manage these
• 1:54
  tools. And third, the critical perspective, which
• 1:57
  frankly is the terrifying look at AI overreach
• 2:00
  and the ethical blind spots that companies are
• 2:02
  just actively ignoring right now. Yeah, it's
• 2:04
  a crucial trifecta. I mean, you can't just look
• 2:06
  at the shiny optimistic side without understanding
• 2:09
  the critical risks, but you also can't just be
• 2:11
  paralyzed by fear. You really need all three
• 2:14
  to survive the next decade of business. Right.
• 2:16
  But before we get into those doomsday scenarios,
• 2:18
  I feel like we really need to understand the
• 2:21
  mechanics of why the business world is so absolutely
• 2:24
  intoxicated by AI's promise. And intoxicated
• 2:27
  is definitely the right word here. Oh, for sure.
• 2:29
  The pressure from investors to adopt AI, it's
• 2:32
  just staggering right now. And it's creating
• 2:34
  this tense cross -functional complexity across
• 2:38
  literally every department. Well, the big promise
• 2:40
  here is AI as a force multiplier. So if you are
• 2:44
  a product manager or a developer or a designer,
• 2:47
  AI is being pitched as this ultimate tool that
• 2:49
  can predict market trends before they even happen.
• 2:51
  It can analyze mountains of real -time data to
• 2:54
  figure out exactly what a user wants, and it
• 2:56
  automates all those mind -numbing mundane tasks
• 2:59
  so you can just focus on big picture creativity.
• 3:01
  Yeah, but we have to look at how it's actually
• 3:03
  doing that. Because it's not magic, right? It's
• 3:06
  pattern recognition on a scale human brains simply
• 3:08
  cannot process. Right, like we just don't have
• 3:11
  the bandwidth. Exactly. You can look at millions
• 3:13
  of unconstructed data points, stuff like customer
• 3:16
  support tickets, social media mentions, click
• 3:19
  rates, and instantly map out a behavioral trend.
• 3:22
  See, I used to think of it like having an ultra
• 3:25
  competent superhuman intern, you know, one who
• 3:28
  doesn't sleep or steal your lunch from the break
• 3:31
  room fridge. But looking at these insights, it's
• 3:33
  not really an intern, is it? It's more like an
• 3:36
  industrial exoskeleton. Oh, an exoskeleton. That's
• 3:40
  a brilliant way to put it. Right. Like you strap
• 3:42
  into this AI exoskeleton and suddenly you have
• 3:44
  the power to lift a 10 ton truck. But you are
• 3:47
  still the one inside the suit deciding where
• 3:49
  to walk. And I think that leads directly into
• 3:51
  this concept of the full -stack product manager
• 3:54
  they talk about. Yes, exactly, the pragmatic
• 3:56
  realism perspective. Because when you strap on
• 3:59
  that exoskeleton, your job description changes
• 4:02
  overnight. You are suddenly expected to bridge
• 4:05
  these massive gaps across strategy, data analysis,
• 4:09
  user design, and software delivery. You become
• 4:12
  the conductor of a very fast, very complex orchestra.
• 4:16
  And looking at the real world leaders on this
• 4:18
  panel, well, it's the best way to understand
• 4:19
  how this works in practice. So take Palak Kumdar.
• 4:22
  She's a vice president of product management
• 4:24
  with, I think, 17 years in the game. Yeah, 17
• 4:27
  years. Right. And she started as a senior developer,
• 4:29
  moved up to technical product management. And
• 4:31
  now she focuses heavily on generative and agentic
• 4:34
  AI. And her career is just the perfect mirror
• 4:37
  for this. She has that really deep technical
• 4:40
  coding foundation, but now she's operating across
• 4:42
  domains like AI governance and MLOps. Okay, let's
• 4:46
  pause there for a second because MLOps is one
• 4:48
  of those buzzwords people love to throw around
• 4:49
  to sound smart in meetings, machine learning
• 4:51
  operations. What does that actually look like
• 4:53
  for her on a daily basis? Well, think of it like
• 4:56
  the plumbing of an AI system. Like building a
• 4:58
  cool AI prototype is fun, but MLOps is the gritty
• 5:02
  reality of keeping that system running reliably
• 5:05
  in a live business environment. Oh, so it's the
• 5:08
  maintenance. Right. It's maintaining the data
• 5:10
  pipelines, ensuring the servers don't just crash
• 5:13
  when thousands of people log on and constantly
• 5:15
  updating the algorithms. And Palak uses modern
• 5:18
  AI tools to manage that incredibly complex plumbing,
• 5:22
  driving global projects across all these different
• 5:25
  teams. OK, so the exoskeleton analogy makes total
• 5:28
  sense for Palak. I mean, she already knew how
• 5:30
  to build the suit. She has a coding background.
• 5:33
  But this brings me back to the wood science degree.
• 5:35
  How on earth does this apply to someone who doesn't
• 5:37
  have a computer science background? And that
• 5:39
  is where this shift gets incredibly inspiring,
• 5:41
  honestly. This brings us to Ajita Gruden. She's
• 5:44
  a product manager at Cardmarket, and her day
• 5:47
  -to -day involves managing monolithic ecosystems
• 5:49
  and scaling massive IT organizations. But as
• 5:52
  I mentioned at the start, her academic background
• 5:54
  is literally in wood science and technology.
• 5:57
  Yeah, it's wild. I just can't get over that.
• 5:59
  From analyzing tree rings to managing complex
• 6:02
  software architecture, How does she do it? Because
• 6:06
  the AI acts as the ultimate translator. Let's
• 6:08
  define a monolithic ecosystem really quickly,
• 6:10
  just to give it context. Imagine a software platform
• 6:14
  as a giant, incredibly complex Jenga tower. OK,
• 6:19
  I'm with you. Every single block. So the user
• 6:21
  logins, the payment processing, the product catalog,
• 6:23
  it's all glued together. If you want to change
• 6:26
  one thing, you risk bringing down the whole tower.
• 6:29
  So in the past, you needed to speak fluent, complex
• 6:32
  code to even touch that tower. Right, because
• 6:34
  you'd need an engineering degree just to know
• 6:36
  which block to pull without breaking everything.
• 6:38
  Exactly. But AI completely democratizes that
• 6:41
  technical execution. Ajita doesn't need to write
• 6:43
  the complex syntax to safely query the database.
• 6:46
  She can use AI tools to map the architecture
• 6:48
  in plain English. And what that does is it places
• 6:51
  a massive premium on her other skills. Skills
• 6:54
  like product discovery and behavioral design.
• 6:56
  Ah, so she understands how users think and behave,
• 6:59
  and the AI just translates her strategic intent
• 7:01
  into technical execution. Precisely. She uses
• 7:05
  structured discovery to figure out what the user
• 7:07
  actually needs, and she relies on the AI to help
• 7:10
  bridge those technical gaps with the engineering
• 7:12
  team. It essentially turns a non -traditional
• 7:15
  background into a massive superpower. OK, that
• 7:19
  is fascinating. But here's where it gets really
• 7:21
  interesting. Because if this exoskeleton gives
• 7:24
  anyone the power to lift a 10 -ton truck and
• 7:27
  it's processing all this data at warp speed,
• 7:30
  well, what happens when it learns the wrong lessons?
• 7:32
  I mean, what happens when the exoskeleton just
• 7:34
  walks you straight off a cliff? Yeah, and that
• 7:36
  brings us crashing right into the critical perspective.
• 7:39
  Because an AI is really only as good as its training
• 7:42
  data. If your AI is relying on faulty data or
• 7:46
  making decisions based on an incomplete picture
• 7:48
  of the world, it becomes a massive liability.
• 7:51
  And the panel outlines three very specific risks
• 7:54
  that create this cautionary tale. The first one
• 7:56
  they talk about is bias. And there's this chilling
• 7:58
  case study mentioned in the sources about an
• 8:00
  AI hiring tool. It was meant to make recruiting
• 8:02
  faster and more efficient, but it inadvertently
• 8:05
  started favoring candidates from a very specific
• 8:07
  historically dominant demographic. And we really
• 8:10
  need to explain why that happens because it's
• 8:12
  not malicious. I mean, the AI didn't wake up
• 8:14
  one morning and decide to be prejudiced. It's
• 8:16
  just an optimization machine. It looked at a
• 8:19
  company's historical hiring data from, say, the
• 8:22
  last 20 years. It analyzed the resumes of people
• 8:25
  who were hired, promoted, and ultimately succeeded.
• 8:28
  Right. Data created by flawed humans. Exactly.
• 8:32
  So the AI identified patterns. Maybe it noticed
• 8:34
  that historically, successful candidates happened
• 8:36
  to play lacrosse or belong to a certain club
• 8:39
  or use specific phrasing on their resumes. It
• 8:41
  just mathematically replicated those historical
• 8:44
  human biases and scaled them up, basically treating
• 8:47
  our past mistakes as the ultimate goal to optimize
• 8:50
  for. That is just terrifying. I mean, it just
• 8:52
  mathematically bakes inequality right into the
• 8:55
  algorithm. It really does. And then the second
• 8:57
  risk they highlight is over reliance on automation
• 8:59
  because it is so tempting for companies to just
• 9:01
  automate everything to save a buck. But AI does
• 9:04
  not have human intuition. It doesn't have empathy.
• 9:07
  It cannot read a room or understand complex shifting
• 9:10
  market sentiments the way a seasoned human leader
• 9:12
  can. Right. And the third major risk is data
• 9:15
  privacy and security. because in this absolute
• 9:18
  gold rush to innovate and appease impatient investors,
• 9:22
  organizations are rushing products to market
• 9:24
  and overlooking the ethical responsibility of
• 9:27
  handling user data. If you aren't transparent
• 9:30
  about how this AI is using customer information,
• 9:33
  you destroy user trust and your whole product
• 9:35
  ecosystem can just collapse overnight. I want
• 9:39
  to zero in on a very specific phrase from this
• 9:41
  critical perspective, because I know it's going
• 9:43
  to resonate deeply with anyone listening who
• 9:45
  actually uses these tools at work. The panel
• 9:47
  warns against the organizational tendency to
• 9:49
  expect one person to do five jobs with AI. Ah,
• 9:53
  yes, the five jobs trap. OK, wait. I need to
• 9:55
  stop you there and push back on this for a second,
• 9:57
  because isn't the entire point of business efficiency
• 9:59
  to use software so one person can do the work
• 10:01
  of five? Like, if I'm a business leader and I
• 10:04
  can pay one salary but get the output of five
• 10:07
  people because of AI, I mean, that sounds like
• 10:09
  the holy grail of return on investment. Why is
• 10:12
  that being framed as a negative? Well, what's
• 10:14
  fascinating here is the absolute distinction
• 10:16
  between capability and sustainability. So yes,
• 10:20
  AI gives one person the raw capability to generate
• 10:22
  the output of a coder, a copywriter, a data analyst,
• 10:26
  a designer, and a strategist all before lunch.
• 10:29
  But the reality of managing that output raises
• 10:32
  severe concerns about human well -being and governance.
• 10:35
  How so? Walk me through what that actually looks
• 10:37
  like in an office. Imagine it viscerally. It's
• 10:40
  9 a .m. and you've used AI to generate a thousand
• 10:43
  lines of backend code. At 10 a .m., you use it
• 10:46
  to write 20 pages of marketing copy. At 11 a
• 10:49
  .m., you generate complex financial projections.
• 10:51
  You are producing a massive amount of material.
• 10:54
  But you aren't actually creating anymore. You
• 10:56
  have just become a nervous proofreader of an
• 10:58
  infinite fire hose of content. Oh, wow. Yeah,
• 11:00
  I can see that. The cognitive load required to
• 11:02
  critically review, manage, and take legal or
• 11:05
  financial accountability for the output of five
• 11:07
  distinct professional disciplines. It's impossible
• 11:10
  for one human brain. You simply cannot evaluate
• 11:12
  all of it with the required depth. So it stops
• 11:15
  being a helpful co -pilot and it basically becomes
• 11:17
  a tyrant demanding constant review. Exactly.
• 11:20
  It turns AI from a tool of collaboration into
• 11:24
  a relentless engine of burnout. And when that
• 11:27
  burnout hits, governance drops. That's when the
• 11:30
  bias hiring algorithm slipped through. That's
• 11:32
  when the security flaws get published to the
• 11:34
  live server. That's when you make a massive market
• 11:37
  misjudgment because you're just too tired to
• 11:39
  realize the AI hallucinated a financial trend.
• 11:42
  I see. So it's not just about doing the work.
• 11:43
  It's about being responsible for the quality
• 11:45
  and safety of the work. If you are doing five
• 11:48
  jobs, you are definitely dropping the ball on
• 11:50
  governance. Which brings us to how we actually
• 11:53
  fix this. If we can't put the AI gene back in
• 11:56
  the bottle, how do we build the guardrails? And
• 12:00
  this is where the philosophy of another panelist,
• 12:02
  Soumya Sundar Govind, comes into play. She's
• 12:04
  a global product leader focusing on B2B software
• 12:07
  ecosystems, and her central philosophy is the
• 12:10
  absolute necessity of keeping the human in the
• 12:13
  loop. The human in the loop. I mean, it sounds
• 12:14
  like a great bumper sticker, but mechanically,
• 12:17
  what does that actually require a company to
• 12:19
  do? It means designing systems where AI informs
• 12:22
  but does not replace human judgment. It is the
• 12:26
  exact opposite of a set it and forget it mentality.
• 12:29
  In practice, this means building continuous feedback
• 12:31
  loops. So if the AI flags a customer account
• 12:34
  for fraud, the AI doesn't automatically close
• 12:37
  the account. It pauses, surfaces the exact data
• 12:40
  points that lead to the conclusion, and forces
• 12:43
  a human to review it and click approve or deny.
• 12:46
  So you are constantly learning from the AI's
• 12:49
  performance and actively intervening. The human
• 12:51
  remains the ultimate editor and like the moral
• 12:54
  compass. Exactly. And this connects directly
• 12:56
  to a crucial point the panel makes about ethical
• 12:58
  AI design and the specific necessity of having
• 13:01
  women and diverse voices in tech shaping these
• 13:04
  tools. Right. Because if we want to fix those
• 13:05
  blind spots we talked about with the hiring algorithm,
• 13:08
  you have to look at who is actually building
• 13:10
  exoskeleton in the first place. This raises an
• 13:12
  important question about the engineering process
• 13:14
  itself. Because if the teams building and training
• 13:17
  these AI models all share the exact same background,
• 13:21
  the same historical advantages, and the same
• 13:23
  cultural blind spots, the AI will naturally bake
• 13:26
  those blind spots into its pattern recognition,
• 13:29
  actively promoting fairness, inclusivity, and
• 13:31
  diverse leadership in the development phase.
• 13:33
  It isn't just some corporate PR exercise. It's
• 13:36
  a technical necessity. Exactly. It's an engineering
• 13:39
  fix for a structural flaw. Diverse teams patch
• 13:42
  the holes in the training data, leading to fundamentally
• 13:45
  safer, less biased, and frankly, more profitable
• 13:48
  product outcomes. Okay, that makes perfect sense.
• 13:50
  Ethical design, human well -being, avoiding the
• 13:53
  five jobs trap, I mean, it is all incredibly
• 13:55
  compelling in theory. But let's be pragmatic
• 13:58
  here. How do you actually justify slowing down
• 14:00
  to build these feedback loops to a board of investors
• 14:03
  who only care about the bottom line and beating
• 14:05
  the competition to market? And that is the ultimate
• 14:08
  hurdle, right? And the panel answers this by
• 14:11
  introducing what they call a business -first
• 14:13
  framework. You have to demonstrate to those investors
• 14:16
  that ethical AI is directly tied to measurable
• 14:20
  ROI. So you have to prove that ethics equal profits.
• 14:24
  When applied correctly, yes. Responsible AI practices
• 14:28
  don't slow you down. They actually accelerate
• 14:30
  your long -term time to market because you avoid
• 14:33
  catastrophic regulatory setbacks. And this is
• 14:35
  where the fourth panelist, Okwainia Papalashi,
• 14:38
  comes in, right? Because she has over 10 years
• 14:40
  of experience specifically in fintech developing
• 14:42
  innovative financial solutions. I mean, she has
• 14:44
  diplomas from Oxford, Google, and Bocconi. Yeah,
• 14:47
  and fintech is the perfect arena to prove this
• 14:49
  point. When you are dealing with people money,
• 14:51
  mortgages, and credit scores, the stakes are
• 14:54
  existential. You cannot beta test an algorithm
• 14:56
  with someone's life savings. O 'Quinney understands
• 14:59
  that in highly regulated industries, sacrificing
• 15:02
  user trust for a slightly faster product launch
• 15:04
  is basically corporate suicide. The balance of
• 15:07
  rapid innovation with strict corporate responsibility
• 15:10
  is what actually builds a moat against your competitors.
• 15:13
  The panel also connects this responsible framework
• 15:15
  directly to the concept of corporate sustainability.
• 15:18
  Usually, when I hear sustainability, I think
• 15:20
  of the environment or carbon footprints. But
• 15:23
  here, embedding sustainability into AI means
• 15:25
  something highly specific to operations. Right.
• 15:28
  It's about reducing operational waste. Every
• 15:31
  time an AI runs a complex query, it costs server
• 15:34
  power and it costs money. If your AI is hallucinating
• 15:37
  or providing biased, useless data, you are wasting
• 15:41
  massive amounts of computing power, human effort
• 15:43
  on failed prototypes, and capital on bad decisions.
• 15:46
  But if you have explainable AI, meaning your
• 15:49
  human -in -the -loop can clearly understand why
• 15:51
  the AI made a recommendation, well, the quality
• 15:53
  of your business decisions skyrockets. That is
• 15:55
  true long -term sustainability. And to achieve
• 15:58
  that, organizations have to fundamentally redefine
• 16:01
  talent. Whether you're in sales, design, or the
• 16:04
  C -suite, the role of a human is shifting toward
• 16:06
  collaboration with AI, not a defensive posture
• 16:09
  of trying to compete with it. So, what does this
• 16:12
  all mean? If you synthesize all these incredible
• 16:15
  perspectives, you know, the MLOops evolution
• 16:17
  of Palak, the behavioral design focus of Ajit's
• 16:20
  wood science background, the human -in -the -loop
• 16:23
  governance from Somia, and the strict, high -stakes
• 16:26
  fintech discipline of Okunaya. The ultimate takeaway
• 16:30
  for you listening right now is profoundly empowering.
• 16:33
  It really is a paradigm shift. The cultural narrative
• 16:36
  that AI is going to make human skills obsolete
• 16:39
  is completely backwards. According to these industry
• 16:42
  leaders, critical human skills like emotional
• 16:45
  intelligence, leadership, empathy, and strategic
• 16:48
  thinking are more valuable today than they have
• 16:50
  ever been in human history. Because the AI can
• 16:53
  write the code, but it doesn't know why the code
• 16:55
  matters. Whether you are a developer, a designer,
• 16:58
  or a manager, your objective should never be
• 17:00
  to compete with the sheer speed of a machine.
• 17:02
  You will lose that race every single time. Your
• 17:05
  goal is to master governance. Learn how to explain
• 17:07
  the AI's logic, mitigate its biases, and maintain
• 17:10
  that crucial human intuition. AI is your co -pilot,
• 17:14
  but you have to fly the plane. Now, as we wrap
• 17:18
  up this deep dive, I want to leave you with one
• 17:20
  final thought to mull over. It builds on a tiny
• 17:23
  but massively important detail we brushed past
• 17:26
  earlier regarding Pallup Komdar's specific focus.
• 17:29
  Ah yes, agentic AI. Exactly. We've spent this
• 17:33
  entire time talking about AI as a co -pilot or
• 17:36
  an exoskeleton. Both of those metaphors imply
• 17:39
  that the human is in control. A copilot sits
• 17:42
  next to you, reads the instruments, and gives
• 17:44
  you advice that you can choose to take or ignore.
• 17:46
  But Paloc is focused on agentic AI. And these
• 17:49
  are systems explicitly designed to take independent
• 17:52
  action. Right. They don't just recommend a course
• 17:54
  of action. They execute it on your behalf. They
• 17:55
  don't just draft the email. They send it. They
• 17:57
  don't just suggest a code fix. They push it to
• 17:59
  the live server. The exoskeleton literally steps
• 18:01
  off your body and goes for a walk by itself.
• 18:04
  Exactly, and if organizations are already burning
• 18:07
  out on the five jobs trap, and if we are already
• 18:11
  fighting tooth and nail just to keep a human
• 18:13
  in the loop with a co -pilot that only gives
• 18:15
  advice, well how on earth will organizations
• 18:18
  maintain governance and oversight when the AI
• 18:21
  becomes an autonomous agent pulling its own levers?
• 18:24
  How do you keep a human in a loop that moves
• 18:27
  at the speed of light and executes without ever
• 18:29
  asking for your permission? That is a whole new
• 18:32
  frontier of the cautionary tale. It's the ultimate
• 18:34
  governance test. It really is. It's definitely
• 18:37
  something to think about the next time you ask
• 18:38
  an AI to just go ahead and automate this task
• 18:41
  for me. Thank you so much for joining us for
• 18:44
  this deep dive. The next time you step into the
• 18:45
  cockpit of your career, make sure you know exactly
• 18:48
  who has their hands on the controls. We'll see
• 18:50
  you next time.