Best of LinkedIn: Digital Products & Services CW 20/ 21

00:00:00: This episode is provided by Thomas Allgaier and Frennus, based on the most relevant LinkedIn posts about digital products in services.

00:00:07: In calendar weeks twenty-and-twenty one.

00:00:09: Frenness is a BDB market research company that supports enterprise product teams with building feature by feature competitive intelligence That shows exactly how their product stacks up against the competition.

00:00:21: you can find more info in the description.

00:00:23: So imagine building a million line software application where your human engineers are like strictly banned from typing A single line of code.

00:00:31: I mean it sounds completely absurd, right?

00:00:33: Yeah But that's exactly what we're looking into today for this deep dive.

00:00:37: We're exploring how AI is fundamentally rewiring product management Right

00:00:41: and we're pulling us From the top digital products & services trends across LinkedIn over The past two weeks curated specifically For you the professionals in the ict and tech industry

00:00:51: Exactly because way past treating AI like some novelty text generator.

00:00:56: Yeah,

00:00:56: oh for sure.

00:00:57: we're tracking a real fundamental paradigm shift here.

00:01:01: the conversation on LinkedIn lately makes it so clear.

00:01:04: We are redefining what our product team actually is when the cost of executing code basically drops to zero.

00:01:10: yeah let's start right there with that sheer speed of execution because Akash Gupta posted this incredible story about an open AI product manager.

00:01:19: Oh

00:01:19: I saw this one.

00:01:21: So, they wrote a document on a Monday and by Friday it shipped as working feature.

00:01:26: And no engineer wrote single line of code for

00:01:29: this.

00:01:29: Not one line?

00:01:30: Nope!

00:01:31: The team lead, Ryan LaPoppolo.

00:01:33: they built this million line app using a totally new workflow where humans were literally forbidden from touching the syntax.

00:01:39: Which

00:01:40: is wild because that forces you to completely rethink what an engineer even does like.

00:01:44: if and AI agent makes mistake in set up You can't

00:01:46: just go ahead fix it

00:01:47: Exactly!

00:01:48: Human can't jump into file and fix typo.

00:01:50: That's bandaid.

00:01:52: Instead had write test or update documentation.

00:01:55: so AI never make specific mistakes

00:01:57: again Right?

00:01:59: Akash noted that sometimes they had to recurse, like six or eight levels deep.

00:02:03: just give the agent the right credentials.

00:02:06: Wow!

00:02:06: Yeah Like if AI hits an API endpoint and fails because of an auth error The human doesn't hand over the token.

00:02:15: They have build scaffolding for it.

00:02:17: Exactly

00:02:17: Build a scaffolding define permissions And teach environment how pass those credentials securely.

00:02:24: After a month of this, the codebase itself literally became The Quality Gate.

00:02:29: I mean they

00:02:30: weren't building the feature at all.

00:02:31: They're building the environment that allows... ...the machine to build the features safely.

00:02:36: Right it's like we stopped throwing the bowling ball down the lane ourselves And instead were just meticulously building the bumpers You know?

00:02:44: So AI can't possibly throw a gutterball.

00:02:47: Oh, I like that analogy.

00:02:47: The bumpers are the compounding asset and that actually ties perfectly into what Knit and Ms.

00:02:52: Ra was talking about.

00:02:53: All

00:02:53: right!

00:02:53: The storyteller thing?

00:02:54: Yeah.

00:02:55: Knit observed in this AI era... ...the best product managers aren't just going to be the story-tellers anymore Like we're so used to.

00:03:03: PMs writing a perfect Jura ticket with all of these user empathy Right.

00:03:09: But now they need people who can turn human intent into what Nitin calls executable truth

00:03:15: Executable truth.

00:03:17: Man, that's a great phrase!

00:03:18: It really is.

00:03:19: The instructions have to be so mechanical and precise That the machine processes it with zero ambiguity Which Prank Jamr actually calls context engineering

00:03:30: Context engineer?

00:03:30: Wait explain that a bit more.

00:03:32: Well

00:03:32: Prank pointed out that traditional team scaling Is just dead

00:03:35: Like just hiring more developers.

00:03:37: Exactly...the old way was linear.

00:03:39: More features meant more headcount But now scale comes from providing the exact right boundaries Defining the guardrails.

00:03:46: So

00:03:46: what does a good boundary actually look like then?

00:03:49: It's defining ATI payload limits, data schemas error handling.

00:03:53: all of that because given AI to little context you get generic output give it too much Like dumping your whole company history into a prompt

00:04:01: and it just starts hallucinating.

00:04:02: exactly it invents features That shouldn't even exist.

00:04:04: so orchestrating that context is The new superpower.

00:04:08: okay.

00:04:08: But but let me push back on this a little bit sure if we've removed All the friction of coding right mm-hmm and we can just execute instantly with some boundaries.

00:04:18: Aren't we just opening the floodgates to churn out a massive amount of garbage features?

00:04:22: That is the big question!

00:04:24: Right, because manual coding was slow but that friction protected us from building bad ideas.

00:04:30: it's too expensive to build everything.

00:04:32: You're definitely not alone in thinking that Shippo Michele actually posted very blunt warning about this exact trap.

00:04:39: Oh, really?

00:04:39: Yeah.

00:04:40: He explicitly warned product teams not to use AI.

00:04:43: to quote build slop efficiently.

00:04:47: Build slop efficiently.

00:04:48: that's exactly it right

00:04:49: because execution is so fast now the bottleneck hasn't actually disappeared It just moved upstream its fully relocated into R&D and discovery

00:04:58: meaning validating The customer problem before you unleash.

00:05:02: Exactly because if you can build a terrible idea in a day instead of six months You're still polluting your product with the terrible idea.

00:05:10: Yeah, and SF mayors shared a really practical tactic to combat that urge to just ship immediately.

00:05:17: Oh what did he suggest?

00:05:19: He said that when you use AI To Build prototypes...you should actually prompt The tool..to make it look rough Like intentionally sketchy

00:05:28: Really like a whiteboard drawing

00:05:30: squiggly lines, mismatch fonts the whole deal.

00:05:33: Yeah because the psychology there is fascinating.

00:05:35: The moment you put a polished screen in front of a stakeholder Oh

00:05:39: they think it's final exactly.

00:05:41: They see high fidelity and they instantly start arguing about like pixel spacing at button colors?

00:05:45: They totally ignore the core workflow

00:05:48: that makes so much sense.

00:05:49: A rough sketch forces them to actually discuss the problem

00:05:52: right as half argues.

00:05:54: the smart move isn't shipping more features to market It's exploring more options internally.

00:05:59: generate three rough prototypes in an hour, test them and see what actually resonates before you commit.

00:06:05: Using AI to stress-test your thinking a

00:06:17: pre-mortem.

00:06:18: How does that work?

00:06:19: So he feeds his project files, market research all of it into the AI's context window and then prompts to imagine its six months post launch.

00:06:27: Oh wow!

00:06:28: Yeah He tells the AI look initially we had great adoption but then usage dropped churn spiked And product failed miserably.

00:06:36: Brutally analyze why this happened based on my documents.

00:06:40: That is so ruthless, I love it.

00:06:42: It instantly cross-references everything like.

00:06:44: you'll find where your acquisition strategy clashes with Your technical debt and point out exactly Where the logic breaks?

00:06:50: Basically tearing your assumptions apart before You write a single line of executable truth.

00:06:54: Exactly.

00:06:55: but um okay.

00:06:56: So If rough prototypes and pre-mortems help us filter out the noise upstream, we're still left with this massive pile of validated ideas.

00:07:05: How do you map those without causing a massive traffic jam on the roadmap?

00:07:09: Oh!

00:07:10: The Roadmap Traffic Jam.

00:07:12: Shahab Jain actually painted a picture of it that I think basically everyone in tech has lived through.

00:07:17: Staring at road maps with thirty items?

00:07:20: Yes,

00:07:21: every single one has a champion... a solid business case, and they're all mission critical.

00:07:27: And then you try to solve it with a scoring model?

00:07:29: Right everyone defaults too like rice reach impact confidence effort.

00:07:33: You run the math add the weights and then you still end up debating The same five items for three hours anyway

00:07:39: because those scoring models totally break down in an AI world Like think about the effort variable in Rice.

00:07:45: Right, if AI drops the effort to near zero.

00:07:47: Exactly!

00:07:48: If an AI can code it in an afternoon then math gets completely messed up.

00:07:52: Every feature suddenly looks like a high ROI quick win.

00:07:55: Yeah Shilab realized prioritization isn't actually a math problem anymore It's clarity problems.

00:08:01: That is great point And Raj Travedi expended on this saying that how has gotten so easy.

00:08:05: The complexity just moved up.

00:08:07: decision stack

00:08:08: Meaning defining WAY Is now hardest

00:08:11: part Exactly.

00:08:13: Deciding what actually deserves to exist, because if execution is instant but your strategy is fuzzy... Your product will just bloat in every direction It

00:08:22: drifts

00:08:23: Right!

00:08:23: Dennis Cupins posted that road maps shouldn't even be gaunt charts anymore.

00:08:28: They should act like universal traffic signs.

00:08:30: Traffic signs?

00:08:32: That instantly makes me think of what Strape did.

00:08:35: Panya Kakos posted about Stripes' new public-product roadmap.

00:08:39: Oh

00:08:39: I missed one.

00:08:40: What'd they do?

00:08:41: They didn't publish some huge laundry list of requested features, they organized everything around these massive crystal clear themes like programmability network activation and AI infrastructure.

00:08:53: Oh wow that kind of thematic clarity changes Everything for the engineers.

00:08:57: exactly if The theme is programmability And an engineer Is picking between two API structures?

00:09:03: They don't need a meeting.

00:09:03: just look at the traffic sign.

00:09:05: yes They pick the one that gives developers more flexibility, because that's their destination.

00:09:09: The strategy does all

00:09:15: of

00:09:20: this

00:09:22: heavy

00:09:23: lifting.".

00:09:24: People tend to think AI is this great equalizer, right?

00:09:27: Yeah.

00:09:28: It levels the playing field for everyone

00:09:29: but The data shows it's not!

00:09:31: AI Is an amplifier.

00:09:33: Smaller mature executive-led product teams are seeing huge benefits But large orgs like over five hundred people AIs actually making things worse For them.

00:09:43: Wait

00:09:44: Making things worse.

00:09:45: So what's the actual ceiling here?

00:09:46: Is the AI tech just not capable enough yet for complex enterprise stuff?

00:09:51: No, The Tech is fine!

00:09:52: The Ceiling isn't capability...the Ceiling is

00:09:54: trust Trust.

00:09:56: Yeah, Malton Schultz shared this AI maturity gap report.

00:09:59: Forty percent of product leaders ranked trust as their number one blocker.

00:10:03: Wow

00:10:04: Higher than security.

00:10:05: Higher then security, higher than data quality.

00:10:07: because think about the friction if an AI writes ten thousand lines of code in a day but it has to sit in a queue for monthly security review board

00:10:15: Your system speed is still zero

00:10:16: Exactly!

00:10:17: The AI just hits those legacy bottlenecks hundred times faster and creates massive gridlock.

00:10:22: And Marisia Cardin brought up another angle on this.

00:10:25: In enterprise software users will totally tolerate a missing feature.

00:10:29: They'll build or work around.

00:10:30: Oh yeah

00:10:30: they use a spreadsheet

00:10:31: Right But they will instantly lose trust if the product's behavior becomes inconsistent.

00:10:37: Because, If AI ships code fast but like breaks a validation rule in a finance module

00:10:43: Or permission structure changes subtly The accountants stop trusting this system.

00:10:48: They export everything.

00:10:49: It does not matter how fast you ship.

00:10:51: if users don't trust output.

00:10:53: That makes total sense And it really ties into what Diego Granados was saying about future of product design.

00:10:58: Oh!

00:10:58: About next user?

00:10:59: Yeah

00:11:00: He pointed out that your product team's next user is actually an AI agent.

00:11:04: Like, Diego uses Claude Code to interact with SaaS products for him.

00:11:08: He literally lets the agent test software before he buys it?

00:11:11: Yes!

00:11:12: The agent checks APIs and documentation to see if they can execute tasks.

00:11:17: If the agent can't navigate the product... Diego just doesn't buy it.

00:11:20: That is incredible.

00:11:21: So we aren't designing human psychology anymore.

00:11:24: We have design machine execution.

00:11:27: Right Diego argues there are three distinct users now.

00:11:31: You have the human who wants intuitive UI, then you have the agent that doesn't care about the UI at all.

00:11:38: it just want predictable APIs high rate limits and perfect documentation.

00:11:42: And who's third?

00:11:43: The admin.

00:11:44: Yeah!

00:11:44: And the Admin is terrified of the Agent.

00:11:46: Oh I bet

00:11:47: They need scope permissions bulletproof audit logs and instant kill switches to stop a runaway agent.

00:11:54: Wow.

00:11:54: So if your operating model is messy and you're code base lacks the consistency Marisa was warning about, an external agent just won't be able to use your product?

00:12:02: It'll hit an undocumented inconsistency in just

00:12:05: break And then your product becomes practically invisible to machine users that are starting to dominate enterprise.

00:12:10: Exactly

00:12:11: so taking a step back looking at everything we've synthesized today it's very clear narrative.

00:12:16: We have officially transitioned out of era defined by ruthless feature prioritization based on limited developer capacity.

00:12:24: Right, the bottleneck is no longer the coding itself.

00:12:27: we're entering an era where your ultimate competitive advantages are extreme strategic focus enterprise grade trust and context engineering

00:12:36: Which means knowing exactly what deserves to be built And building that heavily guarded environment for AI to execute safely

00:12:42: Precisely.

00:12:44: Well We always like leave you with a final provocative thought to mull over.

00:12:48: Based On The Deep Dive

00:12:49: Let's hear it.

00:12:50: So We just talked about how your next user is likely an AI agent, and how you're code base is basically evolving into a constraint layer for coding agents.

00:12:58: Right?

00:12:59: So what happens when an AI-agent user acting on behalf of the client starts negotiating API access in feature requests directly with your AI generated product roadmap?

00:13:10: Oh wow!

00:13:12: In world where machines define demand and machines write codes to supply it What exactly becomes the definition of a product manager?

00:13:19: It

00:13:19: basically becomes a pure exercise in defining intent.

00:13:22: Exactly!

00:13:23: If you enjoyed this episode, new episodes drop every two weeks.

00:13:26: Also check out our other editions on ICT and tech Artificial Intelligence Clouds Sustainability and Green ICT Defense Tech And HealthTech.

00:13:33: Thank You so much for joining us On This Deep Dive.

00:13:35: Don't forget to subscribe.

00:13:37: We will catch you next time.