Best of LinkedIn: Digital Products & Services CW 22/ 23

00:00:00: This episode is provided by Thomas Allgaier and Frenos, based on the most relevant LinkedIn posts about digital products and services in calendar weeks twenty-two and twenty three.

00:00:10: Frenose is a B to B market research company that supports enterprise product teams with building feature by future competitive intelligence that shows exactly how their product stacks up against the competition.

00:00:22: you can find more info.

00:00:23: I

00:00:26: want you to imagine sitting down, too.

00:00:28: write this massive twenty page product specification document.

00:00:33: Oh

00:00:33: man!

00:00:34: The dreaded PRD.

00:00:35: Right exactly.

00:00:37: So your an hour in and you're staring at the giant wall of text And then just realize something completely soul crushing That

00:00:42: no one is actually going read it.

00:00:44: Literally nobody Like not a single person on your engineering team Is gonna read that thing.

00:00:49: For decades i was kind of like the accepted reality Of building software.

00:00:52: Yeah You wrote the document.

00:00:53: Hope for the best Pretty

00:00:54: much.

00:00:55: But over the last two weeks across LinkedIn, we've seen this massive wave of evidence showing that this entire paradigm is just collapsing.

00:01:04: It really is and that's exactly what were getting into.

00:01:06: today We are doing a deep dive in to top trends digital products services from calendar week twenty-two and twenty three.

00:01:15: So if you're professional working at ICT and tech industry This Deep Dive specifically built for You?

00:01:21: Absolutely.

00:01:22: We're basically looking at how AI is fundamentally rewiring, uh...how tech teams strategize build and launch products

00:01:30: And there has been a lot of noise about this recently.

00:01:32: I mean if you follow the space You've probably seen all these like apocalyptic predictions online.

00:01:37: Oh!

00:01:37: The one saying that the product management role is dying.

00:01:40: Yeah..the panic is everywhere.

00:01:41: People are claiming the PM roll will be entirely wiped out by twenty thirty.

00:01:45: I see those posts literally every single day.

00:01:48: It's definitely causing some anxiety.

00:01:50: It is, but Sean Scott actually posted this highly viral and I think incredibly clarifying insight that really cuts through all that panic.

00:01:59: Oh right!

00:02:00: i saw that one.

00:02:00: Yeah he pointed out the product management isn't dying at all.

00:02:04: What IS dying?

00:02:05: And frankly what kind of deserves to die?

00:02:07: Is The role Of Backlog Manager?

00:02:09: Right...the person who essentially just acts as like a human router for JIRA tickets.

00:02:14: That

00:02:15: is the perfect way to describe it, A Human Router.

00:02:17: We're talking about the PM who spends you know eighty percent of their week producing these static artifacts

00:02:23: Just coordinating alignment meetings triage in the back lawn

00:02:26: Exactly and maybe spending zero percent other time actually driving customer impact.

00:02:30: Yeah that version Of The Discipline Is Absolutely Being Automated But Rather Than Killing The Profession.

00:02:37: this automation is actually clearing the path for what product management was always supposed to be.

00:02:43: Which

00:02:43: is focusing on strategic judgment and market outcomes?

00:02:46: Exactly, instead of just drowning in administrative overhead

00:02:49: And we are really seeing that play out in real time Akash Gupta shared an amazing story about Avi Muchal.

00:02:56: Oh!

00:02:57: The PM over at OpenAI.

00:02:58: Yeah exactly This story completely changes how we think about documentation.

00:03:04: Obby essentially proved that the traditional product requirements document, the PRD is just dead.

00:03:10: Basically because he just refused to write one?

00:03:12: Right

00:03:13: yeah He was asked to write a prd for this new platform investment and like twenty minutes into drafting This massive document you just stops.

00:03:21: it hits The wall

00:03:23: right.

00:03:23: he realizes the format itself Is fundamentally flawed For what they're trying To do.

00:03:28: so instead of writing he Just opens up an AI tool and spends that exact same time building a working interactive prototype with the feature instead.

00:03:38: That is

00:03:38: such a better use of time!

00:03:39: It really is, so he takes that functional prototype into his alignment meeting Instantly.

00:03:48: well.

00:03:48: Yeah, think about the mechanics of why that works so much better.

00:03:51: I mean a traditional text document.

00:03:53: asks everyone in a room whether that's engineering design legal marketing to you know simulate A complex software product and their own heads

00:04:02: which is impossible to do consistently

00:04:03: exactly.

00:04:04: naturally they all picture something slightly different And that creates this this illusion of alignment.

00:04:09: right until The product has actually built never one goes.

00:04:12: wait That's not what i thought

00:04:13: yeah the classic?

00:04:17: But a working prototype removes that entire cognitive gap.

00:04:20: Everyone is clicking the exact same buttons and reacting to the exact Same object.

00:04:24: it totally bridges The translation gap.

00:04:26: yeah, and you know To satisfy the absolute necessary documentation because obviously You still need to define success metrics in safety constraints And things like that sure abby just paired the prototype with A really short ten-question FAQ.

00:04:41: That's brilliant.

00:04:42: It is, and Michael Bankhole weighed in on this exact trend too.

00:04:46: He argued that we are officially moving out of that Agile waterfall hybrid era And into an era of what he calls continuous prototyping

00:04:55: continues prototyping.

00:04:56: I like yeah.

00:04:57: In this AI native world discovery actually his delivery you just don't hand off static wireframes anymore.

00:05:02: You handle these messy ai generated version ones?

00:05:05: The velocity of the shift Is when i think people Are really underestimating right now for sure.

00:05:09: Justin LaHullier actually put out this incredibly stark warning, specifically targeting CIOs and tech leaders who are trying to navigate those.

00:05:17: What did he say?

00:05:18: He emphasized that transitioning to AI-assisted software development is not like a standard cloud migration or just adopting new DevOps tool.

00:05:28: it compounds exponentially

00:05:31: Right because the models can write code faster than humans can even read

00:05:35: it.

00:05:35: Exactly, I mean just look at the data coming out of Anthropic.

00:05:38: their engineers are currently merging eight times as much code per day as they did in twenty-twenty four.

00:05:44: Eight

00:05:44: times?

00:05:45: That is insane!

00:05:46: Eight

00:05:46: times Claude Is now authoring more then eighty percent Of The Code that gets merged into Their entire codebase.

00:05:53: Oh

00:05:53: So LaHolier's core argument here is that the biggest risk right Now isn't experimenting too aggressively with AI.

00:06:00: The actual existential risk is waiting so long to adopt these workflows that your organization just mathematically loses the ability to ever catch up.

00:06:09: Okay, I do have to push back on the raw value of that speed though.

00:06:12: Oh how so?

00:06:13: Well

00:06:13: if a team has shipping code eight times faster but you know their strategy's flawed aren't they going to build the wrong things eight time faster?

00:06:21: That

00:06:21: is very fair point.

00:06:23: It feels like putting a Ferrari engine into car which doesn't have steering wheel.

00:06:28: If you don't know exactly where your driving, hitting the gas pedal just means you are going to crash much harder.

00:06:34: And that is the exact tension tech teams are wrestling with right now.

00:06:38: it leads directly into this huge bottleneck around discovery and strategy.

00:06:43: Your Ferrari analogy is actually spot on because the execution of code has become so incredibly fast and cheap.

00:06:50: Deciding what to actually build, it's becoming an ultimate scarce resource.

00:06:54: Because

00:06:55: building is practically free now?

00:06:56: Exactly!

00:06:57: Nadjiv Ornach and Mari Kagan pointed out that this dynamic is creating a golden era for product discovery.

00:07:04: That makes total sense because engineering speed no longer a competitive edge right It's just the baseline requirement even exist in market.

00:07:13: When AI gives unlimited execution speed to you, it obviously gives a tall your competitors too.

00:07:19: But what AI cannot automatically generate is a proprietary understanding of your specific customer.

00:07:25: so knowing the user better than anyone else and then turning that knowledge into the right strategic problems to solve That Is The New Battleground.

00:07:33: but You Know Even That Deep Customer Research Is Being Fundamentally Altered By These Tools Right Now.

00:07:38: Oh Comprehiling Like Paul Wehran Ran An Experiment Recently Did is genuinely hard to wrap your head around.

00:07:45: He essentially gave an AI architecture this massive product discovery job, okay?

00:07:50: But rather than just pasting a transcript into a standard large language model you know in LLM and asking for a summary which is what most people do right.

00:07:58: he used the javascript orchestrator A hundred and thirteen

00:08:06: agents operating simultaneously on a single data set.

00:08:09: Yes,

00:08:09: And in twelve-and-a-half minutes this entire fleet read through a hundred full length customer interviews.

00:08:15: That's

00:08:15: unbelievable!

00:08:16: They

00:08:16: extracted the raw opportunities clustered all those insights into eleven validated user needs... ...And then took these needs and built three clickable prototypes

00:08:25: In twelve minutes?

00:08:26: Twelve minutes.

00:08:27: But The truly staggering part is the underlying mechanism.

00:08:31: here Heron noted that the orchestrator code spent zero model tokens.

00:08:36: Oh wow, okay let's break down why.

00:08:37: that matters especially for anyone listening who hasn't built these models yet.

00:08:40: Yeah

00:08:41: please do

00:08:41: Think of tokens as the currency you pay to an AI provider.

00:08:46: Every single word the AI reads or generates costs a fraction of ascent, right?

00:08:51: So if you have one hundred and thirteen AI bots reading a hundred long interview transcripts those API cost could rack up incredibly fast.

00:08:59: But he bypassed that massive cost entirely.

00:09:02: The JavaScript orchestrator Which is basically the central brain deciding which agent does what, routing data and filtering duplicates.

00:09:08: It didn't use an LLM to make its routing decisions.

00:09:11: Oh

00:09:11: I see.

00:09:11: Yeah we just used standard deterministic coding logic.

00:09:15: Simple if then statements.

00:09:16: That's so smart.

00:09:17: So the coordination in management of this massive Agent fleet was entirely free totally stable and instant.

00:09:25: it means a PM can run deep discovery cycle over one hundred interviews for pennies Basically at that time.

00:09:32: go grab coffee.

00:09:33: That is an unbelievable amount of leverage, but and I think this is important.

00:09:37: We do have to kind of ground this enthusiasm a bit because that kind of automation introduces some pretty severe risks.

00:09:44: Dr.

00:09:45: Els van der Berg in Roman Pichler actually offered unnecessary reality check on this exact practice.

00:09:51: What

00:09:51: do they say?

00:09:52: Well, Dr.

00:09:52: Vandenberg uses AI extensively to extract opportunities from transcripts but she treats the AIs strictly as a second set of eyes

00:10:00: Because an AI model will aggressively hallucinate patterns if you let it run wild.

00:10:04: Exactly!

00:10:04: It'll find connections that just don't exist in real world and completely lacks any human context.

00:10:10: Right...it's reading text!

00:10:11: It doesn't know whether customers sound sarcastic or if there were sighing during interviews.

00:10:16: It only reads words.

00:10:17: That context is everything

00:10:18: It really is.

00:10:19: And Roman Pichler argues that this explosion of AI capabilities actually increases the need for rigorous product strategy, because without a firm strategy acting as a guardrail teams will inevitably use these massive cheap agent fleets to just build, you know tech for tech sake.

00:10:37: Strategy is what ensures all this.

00:10:39: rapid output actually serves a viable business

00:10:41: model.".

00:10:42: Emily White had a brilliant way of framing this human element.

00:10:45: she compared product discovery to songwriting.

00:10:48: Songwriting?

00:10:48: Yeah!

00:10:49: She drew on insights from Wilco's frontman Jeff Tweedy and she pointed out that great software products much like great songs do not just arrive via divine lightning strikes And they certainly don't pop out of an automated prompt.

00:11:03: They require internalizing messy insights, developing a distinct taste and consistently showing up to do the hard work Crazy duct tape workarounds they use every single day just to get their job done.

00:11:30: An AI can easily prompt a generic song structure, and an AI can obviously execute standard code but identifying that deep unspoken human tension.

00:11:41: That requires empathy in judgment

00:11:44: Which brings up the immediate next problem for product teams which is The digital experience.

00:11:50: Because if AI is commoditizing the execution layer, meaning literally any team can build a functional bug-free app overnight.

00:11:58: How do you actually differentiate your product?

00:12:00: That's the million dollar question.

00:12:02: Tim

00:12:02: Herbig warned this week that digital products are at severe risk of drowning in what he calls a sea of sameness.

00:12:08: Oh, because underneath the user interface almost every new product is calling The exact same foundational AI models from OpenAI or Anthropic Exactly.

00:12:18: And if the underlying brain is identical the outputs Are naturally going to converge

00:12:21: Right.

00:12:22: so Herbig argues that having taste and a highly distinct point of view like, a POV are basically the only durable competitive moats left.

00:12:31: Interesting He points to tools like IA Writer.

00:12:34: It's this minimalist writing app that has maintained very strict distraction-free POV for fifteen years.

00:12:41: or you look at the highly curated opinionated feel of Airbnb.

00:12:46: So taste isn't just about premium graphic design or making things look pretty, it is about making deliberate highly opinionated choices about what your product will not do.

00:12:56: Yes and how the user should actually feel when they're using it.

00:13:00: Diana Mattei posed a really profound question to help product teams navigate this sea of sameness.

00:13:05: Oh was that?

00:13:06: She asked Which moments belong to agents and which moment's belonged humans?

00:13:10: Oh, wow.

00:13:11: That is a great lens to look through.

00:13:12: it

00:13:13: really reframes everything I mean.

00:13:14: we have spent the last twenty years essentially training ourselves To think an act like machines totally.

00:13:19: We type in these clunky keyword queries into search bars.

00:13:22: we navigate Through endless nested drop-down menus just to accomplish basic tasks right.

00:13:27: but now The technology has finally evolved.

00:13:29: understand our messy human requests.

00:13:33: AI agents are faster they're cheaper And They never sleep.

00:13:37: So if your product's only value is executing a single task, the AI agent wins every time.

00:13:43: But a great product is rarely just a task?

00:13:45: Exactly!

00:13:46: As Matei points out, products are portfolios of moments.

00:13:50: Agents can handle mundane data entry but humans are ones who actually experience that moment….

00:13:56: The emotional beats.

00:13:57: Yes It's feeling relief when complex work flow finally finalized or it's frustration on page load.

00:14:04: slowly The agent executes the task flawlessly, but the product's design must protect and elevate that human moment.

00:14:11: That concept of protecting the moment applies so incredibly well to the blurring line between physical and digital spaces too!

00:14:18: Oh

00:14:18: how

00:14:18: so?

00:14:19: So Sally Iterley shared a breakdown of her work rebuilding the Digital Presence for Noma.

00:14:23: Noma...the world-renowned Michelin star restaurant.

00:14:26: Yes

00:14:27: exactly she pointed out that Noma obviously has this incredible premium highly orchestrated physical brand when you go.

00:14:35: But their legacy online experience felt entirely disconnected from that reality.

00:14:40: Oh, I can see how it creates a really jarring cognitive dissonance for the user?

00:14:44: Yeah!

00:14:44: It's like paying for a world-class Michelin star meal but the dining table is sticky and the waiter just keeps dropping your silverware... ...it

00:14:53: completely ruins the perception of the food....It

00:14:55: does so.

00:14:56: Noma had to entirely rebuild their digital flow because a clunky checkout cart or an outdated visual design fundamentally breaks that human moment Mattea was talking about.

00:15:06: That makes perfect sense!

00:15:08: If there is friction online, the customers trust waivers regardless of how good the physical food is.

00:15:14: taste just had to be consistent across every single touch point

00:15:17: right.

00:15:18: so if we look at all these philosophies We've been talking about you know The death-of-the-back log the zero token discovery agents the absolute necessity of taste How are builders actually translating this into real?

00:15:29: products right now.

00:15:29: Yeah,

00:15:30: let's look at the actual launches

00:15:31: because we saw several fascinating operating models emerge in weeks twenty-two and twenty three.

00:15:36: Let's

00:15:36: start with Samantha Park.

00:15:38: She is building her first AI native product which is a cooking assistant called Sammy.

00:15:43: Okay

00:15:44: And what I love about her approach?

00:15:46: Is that she didn't start was some massive you know world changing platform vision.

00:15:51: She started by focusing purely on what she calls the mustard finger problem.

00:15:55: Mustard finger problem?

00:15:57: What is

00:15:57: that?

00:15:57: think about the last time you cook a really complex meal Using your recipe blog and your phone,

00:16:03: okay Yeah

00:16:04: Your hands are covered in food.

00:16:06: A pan is sizzling on the stove yeah And you have to keep scrolling up and down The page just to check if it was like a teaspoon or a tablespoon of specific spice.

00:16:14: Oh

00:16:14: my gosh.

00:16:15: yes, and meanwhile pop-up ads are shifting the text around exactly.

00:16:19: It is an incredibly frustrating user moment.

00:16:21: It's

00:16:21: pure friction at the worst possible time!

00:16:24: So her AI agent doesn't just summarize the recipe, it actively alters the interface... ...it embeds exact measurements and required kitchen tools directly into step-by-step instruction text as you read them.

00:16:37: Oh wow… It removes need to scroll entirely – a perfect example of focusing on specific human moments rather than throwing a chatbot onto a web page.

00:16:48: That kind of hyper focus is so critical right now, and Eric Manrique shared a very similar realization while building his product Ambedo.

00:16:56: What's the story there?

00:16:57: Well he originally set out to build this massive company operating system.

00:17:02: you know He wanted to capture organizational memory project tracking cross-functional visibility basically A massive heavy enterprise

00:17:10: platform which usually requires employees do like hours of manual data entry just to keep it updated.

00:17:15: And that is exactly why those tools always fail to get adopted, right?

00:17:19: So through his discovery process Manrique realized he was trying to solve way too many abstract problems at once.

00:17:25: The fundamental issue is that knowledge workers just have too much scattered information.

00:17:30: They're overwhelmed

00:17:31: completely.

00:17:32: so he pivoted the entire architecture.

00:17:34: He shifted from building a heavy company OS two building a lightweight personal AI chief of staff.

00:17:40: I like He

00:17:41: boiled the entire value proposition down to answering one single question.

00:17:46: If I only have five minutes before my next meeting, What do I absolutely need to know?

00:17:51: Wow, that cuts through so much corporate noise.

00:17:54: It really does!

00:17:55: It isn't about giving the user more dashboards to look at.

00:17:58: it is about giving them clarity when they are under pressure

00:18:01: Exactly...it takes a heavy cognitive load of synthesizing emails and documents off the human And gives it to an agent which preserves humans' actual mental energy for their meeting itself.

00:18:14: I have one more example from the last two weeks, and it might just be... ...the ultimate case study in aligning a tech team.

00:18:20: Let's hear it!

00:18:21: Axel Soria & Richard Sorter spoke at La Prodiconf Paris about how the Atlassian Williams Formula One Team operates.

00:18:28: Formula One is arguably the most demanding execution environment in the world?

00:18:33: It really is—I mean they have twelve hundred people working on this engineering team….

00:18:36: …and they align every single one of them using just ONE North Star metric.

00:18:40: What is it...?

00:18:41: Laptime just lap time.

00:18:42: Just laptime.

00:18:43: they don't use abstract at all story points.

00:18:46: They don't used generic okay ours.

00:18:48: Every single Jira ticket every proposed software feature is measured strictly in milliseconds.

00:18:54: that completely changes the psychology of a planning meeting.

00:18:57: it fundamentally removes The ego.

00:19:00: its stops being a political debate Of you know, the marketing team's idea versus the engineering teams idea and just becomes a pure data equation.

00:19:08: which idea shaves off more milliseconds.

00:19:10: That's

00:19:10: incredible!

00:19:11: And the way they leverage AI here is brilliant, They use internal AI agents to constantly scan their backlog.

00:19:17: Wait,

00:19:18: how does an AI estimate a car's lap time from a text-based JIRA ticket?

00:19:23: It is wild!

00:19:23: The bot parses the technical specs of a proposed feature and automatically runs those variables against teams' historical telemetry data & physics simulations.

00:19:32: Oh my goodness...

00:19:32: Yeah it bridges the gap between text and physics estimating the laptime benefit before human manager even opens that ticket.

00:19:39: That provides immediate objective feedback to engineer who actually wrote.

00:19:43: And the behavioral result was totally unexpected.

00:19:47: Engineers started arguing with AI about its estimates.

00:19:51: No way!

00:19:53: To prove it wrong, engineers had to continuously refine and improve their own ideas without engineering management ever having to intervene.

00:20:02: It drove an incredible cultural shift just by relentlessly focusing technology on a single undeniable outcome.

00:20:10: Which brings us to a really crucial piece of advice from David Pereira, specifically for you the listener as you navigate this transition in the tech industry

00:20:19: What's it take away?

00:20:20: If you want to survive and thrive In an AI native world You have to fundamentally change your vocabulary

00:20:26: because The language that team uses dictates their reality

00:20:29: exactly.

00:20:30: Pereira argues that tech professionals need to permanently stop using project language.

00:20:35: Like what?

00:20:35: Stop asking about scope, arbitrary timelines and static deliverables.

00:20:40: That language protects the initial plan But it does absolutely nothing to protect the value of what you are actually building.

00:20:46: Instead we must strictly use product language.

00:20:49: You should be asking about the expected impact The required investment And key decision points

00:20:54: Because projects have a fixed end date Regardless whether they succeed or fail.

00:20:58: Right Product initiatives are strategic bets that you should abandon the moment.

00:21:02: The data proves you wrong.

00:21:04: if your team is debating outcomes in milliseconds rather than output and features, You were on the right

00:21:10: track.

00:21:11: I love that.

00:21:12: before we go We want to leave you with one final thought to mull over yeah?

00:21:16: We talked earlier about Michael Bankle's concept of continuous prototyping.

00:21:20: Yeah And the fact that Anthropic is merging code eight times faster then they did last year.

00:21:24: Right be insane speed.

00:21:26: So, if AI is drastically shrinking the time it takes to go from a raw idea... ...to a functional prototype.

00:21:33: What does that actually mean for traditional two-week agile sprint?

00:21:37: Oh!

00:21:37: That's great question.

00:21:38: Are we about see complete extinction of the Sprint Cycle in favor of continuous realtime deployment?

00:21:45: Think About It.

00:21:46: Thanks so much for joining us on this deep dive.

00:21:48: Don't forget to subscribe so you never miss an update

00:21:50: If You Enjoyed This Episode.

00:21:52: New Episodes Drop Every Two Weeks.

00:21:54: Also, check out our other editions on ICT and Tech – Artificial Intelligence, Clouds, Sustainability in Green ICT, DefenseTech & HealthTech.