Best of LinkedIn: Digital Products & Services CW 12/ 13

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

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

00:00:20: you can find more info In the description.

00:00:23: so If you are an ICT or tech professional listening to this right now, You're probably feeling the whiplash.

00:00:31: completely tearing tech companies apart right now.

00:00:34: Oh,

00:00:34: absolutely!

00:00:35: It's everywhere

00:00:36: Right we've reached this point where AI has made building software cheaper and faster than at any other time in human history.

00:00:42: but paradoxically up to eighty-five percent of these new tech initiatives are still completely failing.

00:00:47: Yeah it is a staggering contradiction.

00:00:49: I mean essentially gave everyone the industry super power And somehow The failure rate didn't budge.

00:00:56: And the reason for that, well.

00:00:57: That's exactly what we're digging into today.

00:00:59: The bottleneck in creating digital products has fundamentally moved

00:01:03: Exactly.

00:01:03: So in this deep dive, we're going to figure out exactly where that bottleneck went.

00:01:07: We are looking at the top digital products and services trends seen across LinkedIn In calendar weeks twelve and thirteen And were going to explore how AI is you know actively eating The operational overhead of tech industry

00:01:20: Right as a road map for what were covering.

00:01:23: Were gonna look at How that shift Is forcing A massive redesign Of product operating models Like why the agile processes You rely on might actually be choking your company's bandwidth.

00:01:33: Yeah, which is terrifying it

00:01:35: is.

00:01:36: and then we'll get into how successful teams are using product ops to completely rewire their internal plumbing And finally will explore what happens when you give autonomous agents the keys to the castle?

00:01:46: I trust in human governance suddenly stopped being a boring compliance topic and become those critical features to build.

00:01:54: Okay, let's unpack this because to really understand that shift we kind of have is no longer just this cute site assistant you use to polish up an email, it's fundamentally becoming the new product operating

00:02:09: layer.

00:02:10: Oh totally!

00:02:10: It's the infrastructure now?

00:02:11: Exactly so.

00:02:12: there was this incredibly relatable story shared by David E on LinkedIn.

00:02:18: he admitted that he was spending something like seventy percent of his time on what he called PM taxes.

00:02:23: oh man...the taxes?

00:02:26: yeah if you are listening and work in products exactly what those are its'just a sheer administrative weight of the job, right?

00:02:32: Yeah.

00:02:33: It's soul crushing.

00:02:34: it

00:02:34: really is spending hours manually updating task crackers across four different tools cross referencing Jira tickets with GitHub commits pulling data into a spreadsheet and

00:02:43: then formatting at all to slide deck for some Tuesday afternoon alignment meeting

00:02:48: exactly.

00:02:50: And David thought was just inevitable.

00:02:51: reality

00:02:53: Right, but then he started using Claude code.

00:02:55: and the wild part is...he didn't use it to write consumer-facing software.

00:02:59: He used it to automate that entire seventy percent of his own workload!

00:03:03: Wow so you built a zone pipeline?

00:03:05: Yeah literally turned the AI into a personalized pipeline.

00:03:11: the development repositories and, you know surfaces exactly what needs his attention every morning.

00:03:19: That

00:03:20: is amazing!

00:03:20: He even taught it his specific slide formats & company branding.

00:03:24: So now presentations that used to take him half a day are generated in about twenty minutes.

00:03:29: He reclaimed all of those hours for actual strategic work and talking.

00:03:38: He literally had an AI read all of his Slack messages and meeting transcripts for an entire week to build a behavioral script of himself.

00:04:00: See, that sounds like science fiction to me.

00:04:02: it really does but the mechanism is totally accessible right now.

00:04:05: The AI basically looked for patterns in how Carl operated.

00:04:09: It noted exactly How he groomed product backlogs?

00:04:13: This specific phrasing used when asking engineers for status updates.

00:04:17: And get this Even how he scheduled his late afternoon Slack messages, so he appeared highly engaged.

00:04:23: Oh my gosh wait... So he essentially cloned his own work persona?

00:04:27: Pretty much!

00:04:28: He took these extracted behaviors, organized them into a workflow file and set the AI loose to run them autonomously And few weeks later His VP actually told him that was really hitting his stride…he even got nominated for spot bonus.

00:04:41: Wait..

00:04:42: Really?!

00:04:42: He got a bonus not doing the

00:04:44: work?!

00:04:44: Yup He got rewarded for designing the AI capability.

00:04:47: that did the work.

00:04:47: We are moving from prompting to designing AI

00:04:50: workflows.".

00:04:51: Okay, but let me play devil's advocate first.

00:04:53: second here because if we pause and think about the logic... If Carl can automate his own presence so effectively he gets a bonus machine doing this job.

00:05:02: doesn't it raise a massive red flag?

00:05:04: For whole industry!

00:05:06: How

00:05:06: do you mean?

00:05:06: Well

00:05:07: Paul O'Brien argued this exact point.

00:05:09: He pointed out is cost of creating software drops essentially zero.

00:05:13: the product itself just becomes a commodity.

00:05:16: We're seeing this trend of vibe coding where anyone can generate a functional app entirely through natural language prompts over a weekend without knowing any syntax.

00:05:25: Right, vibe coding is huge right now!

00:05:28: Yeah so if execution is basically free aren't we just engineering our own obsolescence?

00:05:33: Are we hitting like the Netflix streaming wars phase of software where there's so much cheap content that none.

00:05:42: This is an important question about how we actually define value.

00:05:46: When the manual labor of execution becomes free, The definition of quality fundamentally changes.

00:05:52: Akash Gupta had a brilliant insight on this specific transition.

00:05:55: Oh!

00:05:55: About the testing?

00:05:56: Yeah

00:05:56: exactly He pointed out that for last fifteen years... ...the gold standard for product teams has been the A-B test.

00:06:01: Right You build a blue button you build a green button You route half your live user traffic to each and sit around waiting two weeks To see which one gets clicked more

00:06:09: Exactly.

00:06:10: It requires lots of patience But best-in-class AI teams are realizing that a two week experiment is just a dinosaur in the current landscape.

00:06:18: They're replacing traditional AD tests with AI evils, you know.

00:06:22: evaluation scoring functions.

00:06:24: How does actually work and practice though?

00:06:26: Because how do you score something without a real user clicking on it?

00:06:29: Well let's say you are building an AI feature That summarizes long messy sales calls.

00:06:36: You don't A-B test the prompt on live users to see if they like it, instead you write a scoring function.

00:06:42: You tell this system to score every output at scale of zero to one based upon criteria Like did it capture pricing?

00:06:48: Is summary under two hundred words Did maintain professional tone?

00:06:52: Oh I see.

00:06:53: so set rules what good looks

00:06:55: right And once define that function you can run thousands automated tests locally in minutes.

00:07:00: Gupta showed a staggering statistic on this.

00:07:02: Teams running evils are doing about twelve point eight experiments per day.

00:07:06: Almost thirteen experiments a day, that's insane!

00:07:08: Traditional teams do what?

00:07:09: Maybe three-a month

00:07:11: Exactly Over a single quarter.

00:07:13: the team using evils explores over a thousand variations while the traditional team explores maybe nine.

00:07:20: But here is the critical part that answers your question about obsolescence.

00:07:24: The human value hasn't disappeared It's just moved up the stack.

00:07:28: Right, because AI doesn't know what business actually needs

00:07:31: Exactly!

00:07:32: The AI can generate a thousand variations but it has absolutely no idea what constitutes good outcome for your specific company.

00:07:40: Human value is now entirely in defining that scoring function.

00:07:44: deep customer context judgment, deciding what to evaluate.

00:07:48: that is the new premium skill.

00:07:49: So

00:07:49: The bottleneck isn't the engineering effort anymore it's knowing exactly What problem we are actually trying To solve spot on which perfectly transitions us into the second major theme.

00:07:58: We're seeing the complete redesign of product operating models.

00:08:02: because I mean if you can run a thousand experiments A quarter but your company only approves budgets once a year You hit a massive wall or

00:08:09: the biggest wall.

00:08:10: Yeah, same close you see pointed this out on LinkedIn.

00:08:12: He said that historically our delivery models all that endless backlog grooming the pixel perfect design handoffs The months of discovery they were designed for one purpose to protect scarce engineering time.

00:08:26: Right because letting an engineer write a line of code was most expensive thing You could do?

00:08:29: You had be absolutely certain

00:08:31: exactly.

00:08:32: But if building is now the cheapest and fastest part up the process those legacy models are actually actively harmful.

00:08:39: Yeah, and Jesper Haye Jensen argued this perfectly.

00:08:41: He said that Agile isn't dead It's stuck.

00:08:45: it stalled at the team boundary.

00:08:47: Oh here's where it gets really interesting?

00:08:48: Its like we upgraded The company from a horse-drawn carriage to A Ferrari But management is still trying To steer with leather rains.

00:08:56: That Is the perfect analogy.

00:08:57: The development teams can sprint At light speed but the wider organization With its twelve month budgeting cycles And committee approvals are just choking the bandwidth Completely.

00:09:10: Mark D. Orlick shared data showing the traditional project focused IT models fail up to eighty-five percent of the time.

00:09:18: Wait, eighty five percent?

00:09:19: Yeah...up

00:09:19: to eighty five per cent whereas product oriented companies see sixty percent higher shareholder returns.

00:09:25: Okay but Practically speaking, for the listener sitting in a tech firm right now it's very easy For CEO to stand on stage and declare that they are a product-led organization.

00:09:36: What is the ultimate litmus test?

00:09:38: That a company has actually made this shift from a project mindset To a product mindset?

00:09:43: what really comes down to where the ideas originate?

00:09:46: Josh Forrestine offered The Ultimate Diagnostic.

00:09:49: Yeah.

00:09:50: He asked, are you consistently prioritizing solution ideas that came from engineers?

00:09:54: Oh I love that!

00:09:55: Because if an organization is truly empowered, engineers aren't just sitting in the basement waiting for a spec sheet.

00:10:01: they're actively sitting-in on customer discovery calls

00:10:04: right there and just handed a JIRA ticket and told to code it.

00:10:07: yeah

00:10:07: exactly when engineer watches user struggle with interface first hand their motivation changes.

00:10:13: instantly we go from executing orders to collaboratively solving human problem.

00:10:18: If your engineers are just taking orders, you don't have a product operating model.

00:10:22: You just have a very fast future

00:10:23: factory.".

00:10:24: Yeah but getting engineers into those conversations and letting them rapidly prototype solutions that requires a massive amount of internal coordination.

00:10:34: like the organization itself needs an nervous system to function which explains product operations.

00:10:42: Yes, someone has to build the tracks while the train is moving

00:10:45: exactly but Melissa Perry came out with a very stark warning about this.

00:10:50: she said that if you staff your first product ops higher like an admin do road map tracking and meeting coordination You will completely kill the function

00:10:58: because That completely misses the strategic value Product Ops needs to solve The single highest friction problem in Your product development process

00:11:06: right?

00:11:07: And Antonia Lendi framed this beautifully.

00:11:09: She said, product ops is simply service design applied to the company's own operating model.

00:11:14: The user is the organization and the services how decisions get made.

00:11:18: That a great way of looking at it.

00:11:19: You're basically fixing internal data pipelines As Marielle Vlander noted.

00:11:24: when you fix those pipelines Product Ops naturally evolves into portfolio management.

00:11:30: They act as guardian strategy.

00:11:32: Okay, but I want to push back on this evolution a bit because Dinka Morali raised a really good point.

00:11:36: She said that with AI coming in, ops teams need evolve into systems.

00:11:41: architects managing fleets of AI agents

00:11:44: Right the agent's handling all connective tissue Yeah

00:11:46: gathering data synthesizing notes All invisibly in background.

00:11:50: So wait is product ops just becoming IT department for AI agents?

00:11:54: Because if they are automating all connected tissues don't we risk losing messy human conversations that actually drive alignment?

00:12:01: Well, if we connect this to the bigger picture.

00:12:05: The goal isn't to eliminate the conversation.

00:12:07: when execution becomes a cheap commodity genuine alignment actually becomes the most expensive and rare resource in your company.

00:12:14: Oh that makes sense.

00:12:15: yeah.

00:12:15: so product ops Isn't replacing human Conversation.

00:12:19: it is removing the administrative baggage So humans can focus purely on the checkpoints.

00:12:24: as strategic judgment you isolate the Human element where matters Most.

00:12:28: So you are arguing about whether a JIRA ticket is updated.

00:12:31: You're arguing about if the feature actually solves customers' problems, which brings us to the final and honestly most critical part of this equation speed without governance is massive liability.

00:12:43: when we let these autonomous AI systems loose We hit wall reality Which bring up sudden urgency around trust Human loop design And ethics.

00:12:53: Nowhere's more evident right now than in healthcare tech.

00:12:56: Yes.

00:12:57: Jennifer C. Wrist highlighted a critical shift happening there, she pointed out that PMs must focus on agentic AI meaning AI that autonomously submits prior authorizations or even flags deteriorating patients.

00:13:11: I mean it is amazing how fast this is moving into production.

00:13:14: It is but the risk factor is off-the charts which is why Wrist established golden rule for designing these systems.

00:13:21: You must build the explainability UI and the audit log before you ship the Agent.

00:13:26: You can't just bolt it on later.

00:13:27: No,

00:13:28: absolutely not.

00:13:29: making the AIs reasoning visible and human override totally frictionless is The only way to earn adoption if a chief medical officer Can't instantly see the clinical logic of why an AI flagged patient.

00:13:39: They will never adopt that tool

00:13:41: right.

00:13:42: but honestly Let's pivot to the darker more ethical side of trust here.

00:13:45: Teresa Torres And Petra Will did a podcast episode warning about ai clones.

00:13:49: you know bots built from podcast transcripts That give advice in an expert's name.

00:13:53: oh yes the ai clones.

00:13:55: But honestly, isn't imitation the sincerest form of flattery?

00:13:59: I mean if an AI can act like Teresa Torres to help scale her knowledge and

00:14:10: a system thinking like them.

00:14:18: Expertise isn't just the database of past transcripts, it is context nuance and judgment.

00:14:24: It operates on deep comprehension whereas AI currently operates on pattern matching.

00:14:29: So the AI only knows what you said yesterday not how would react to brand new variable today

00:14:34: Exactly.

00:14:35: And when an AI confidently gives mediocre or even harmful advice with real humans name attached to it destroys trust entirely.

00:14:43: Trust Design is no longer a late compliance layer.

00:14:46: It's the core of modern product architecture.

00:14:49: Yeah, that makes total sense.

00:14:50: If users can't trust their reasoning The speed development just doesn't matter Right.

00:14:54: And it actually leads to one final forward-looking thought I want leave you with.

00:14:59: We've spent this entire deep dive talking about how humans build products for other human using AI.

00:15:04: But what happens when consumer of your product Is not more than a human but an AI agent acting on there behalf?

00:15:12: bots buying from bots?

00:15:13: Exactly.

00:15:14: If agents start doing our shopping, booking our software and evaluating our B-to-B tools product managers might soon have to design entirely new interfaces.

00:15:22: we're talking B-To-A business two agent interfaces.

00:15:26: how will your products stack up when a machine algorithm is the one experiencing it?

00:15:29: Wow

00:15:30: that is wild thought to leave on.

00:15:32: The customers buying the products may not even be human anymore.

00:15:35: We definitely have to explore in future.

00:15:37: If you enjoyed this episode, new episodes drop every two weeks.

00:15:40: Also check out our other editions on ICT and tech artificial intelligence cloud sustainability in green ICT defense tech And health Tech.

00:15:49: Thank You so much for joining us For This deep dive.

00:15:51: make sure to subscribe.