That’s The Great Fold.

The pitch is as follows: AI handles the repetitive stuff, humans absorb the rest. Roles compress. Fewer people, same output, better margins. On a slide it makes total sense. The uncomfortable part is what it feels like from the inside. Which is: you used to have a job. Now you have a jurisdiction.

And it’s getting bigger every quarter.

Here’s what I think is actually going on. Tools that used to require a specialist look like they are only one prompt away. Need a quick data analysis? Don’t ask the analyst — just do it yourself. Need copy for a campaign? Don’t loop in the copywriter — generate something and clean it up. The argument is that this is empowering. And sometimes it genuinely is! But there’s a second version of that story. One where ‘empowering people to be more self-sufficient’ is just a polished way of saying: we’re not backfilling that role, try to avoid the AI slop and make it look like it could well be your job.

What happens when everybody is absorbing more or well could absorb more? The metric shifts. It’s no longer only important that you’re doing your job well. It also becomes important how much you’re doing compared to the person next to you. The colleague who never asks for help, who somehow always delivers without creating dependencies — they become the benchmark. Not officially (yet). Nobody puts it in a performance review like that. But you can start to feel it.

Will you stop asking questions you should ask? Because you think you can find the answer yourself? Or rather, you get the feeling an AI will give you the right answer without knowing how to compare. Will you stop flagging things that need a second pair of eyes? Do a second pair of eyes become a second pair of LLM’s with an LLM-as-a-judge to make the final judgement? So what if everybody starts optimizing for independence over being most effective? There is an argument that can be made where you can point out that it is here that the fold starts cutting rather than just compressing.

We should not overlook: we built up specialization for a reason. The person who spent years learning how to structure research without biasing the result. That wasn’t just a job title. That’s accumulated knowledge that took a long time to build.
But we should not overrate it either: It does look like that the more specialized your job has become, the more bounded training data and instructions you left to grabble for LLM’s. So where are we now? Can we fold specialisms into more generalistic roles or do we undeniably lose critical nuances? Too early to tell.

But will this trend continues? Obviously. Is it all bad? No. Honestly some of the best work I’ve done and seen from people comes from exactly stepping beyond a certain scope and making more of it your own — not because of pressure but because you want to. That’s a different energy entirely.

The question isn’t whether The Great Fold is happening. It is. The question is whether the organizations driving it are building something resilient — or just squeezing harder on what’s already there, and calling it progress.

At some point, something gives. Let’s see who’s still standing when it does.

Many tools have come before us that have helped mankind make huge leaps. The question is: ‘Are agents the new tool of this area, where AI gets ‘embodied’ into something we can pick up (digitally that is) or pick up on rather fast and easy?’ Will we get to the point that you can recreate your own way of working within agents?

Depending on the definition we are already there. Whether you’re creating agents yourself, or you’re using AI software to work with, in both cases, you’re calling upon agents to get work done for you. And exactly therefore, when it comes to jobs and individual ‘utility’, you should dive into this new paradigma where you can either send out agents for you, in sequence or parallel or become an agent yourself (eg Delphi) giving you the possibility to duplicate yourself.

But when it comes to monitoring this, even benchmark them (yes i dare to say it), it seems like everybody turns a blind eye. We’re not talking about it that much. Everything becomes anecdotal very fast where we humans quickly point to instances an AI messed up badly (eg. How is it possible that it can’t count an amount of the same letters in a word?!).

Nonetheless, the change is already under way, on a daily basis. We are not calling everything an agent (this feels to distant, mechanical and un-real) but we are using them already en masses: sometimes through LLM’s, sometimes through FAQ’s, sometimes on a phone call, … . Sometimes for small tasks, small asks, lookups, confirmations, bookings, etc.. . Also more and more for more elaborate processes, multiple steps long, really freeing up time or taking over parts of your work. But what always gets the headline is the miscellaneous word ‘autonomous’. Agents to often market themselves as autonomous. Reporters love this: Agents taking over. Reducing demand for human labor. And not surprisingly, that is also where marketing managers of ‘agent-startups’ are going for. Here's a quick overview of a listing of some of the taglines of these agents I’ve mentioned in a different article:

You’ll get to know them as Your AI Chief of Staff, Max: your autonomous AI recruiting partner, The first autonomous ML Engineer, Julius The AI Data Analyst, Cora is the $150,000 chief of staff that only costs $15 per month, Donna: The only proactive AI assistant for field sales, Lindy: your first AI employee, Delty: Your AI staff engineer,… while some will propose working alongside with you excelling in specific tasks you maybe would not want to do at all: AI Voice Agents‍ for Patient Access and RCM, Freckle sits on top of your CRM, auto-enriching every record coming in from any source, … or they give you the option to Hire ready-made digital workers or create a custom digital workforce tailored to your business, Build and deploy AI agent workflows, Hire and set up enterprise-grade AI Employees within minutes, not months, …and so on. Some are even harsh about it and suggest us to even directly Fire our GTM Engineer!

We tend to overlook the gains it fosters and focus a lot on how it is not on parity with humans. Why not approach it head on and say bluntly that every business should figure out how to create a hybrid workforce, comprised of human & agents, on different levels of autonomy, both with their advantages and disadvantages. And when we do, make sure we monitor correctly on what, why and how we blend these two workforces.

Check out this prototype that gives a hint on how we could start setting things in place so we know who is doing what work in your business

Manage hybrid workforce

What if most software is about to become as casual—and disposable—as a sticky note?

It made me think that the App Store could become an empty store one day. Ephemeral apps ( or throwaway apps, one prompters, … ) could be: tiny, single-purpose bits of software created on demand, used for a moment, and then maybe even thrown away because it has served its purpose.

So What happens to an app marketplace when anyone can create an app from a prompt in 30 seconds?

Most apps people download get used once (or never) and then just sit there.

Our behavior hints at something:

• We don’t want more apps.

• We want outcomes.

• And we want them now.

So what could a throwaway app be then?

• Single-purpose (solves one problem)

• Short-lived (minutes, hours, or as long as the incumbents have use for it)

• Cheap to create (often automatically)

• Low commitment (often not even an install but in-browser, no account, no updates)

• Contextual (built for you, right now)

Can you give me a few clear examples that proves this point?

“Make me a one-time wedding seating planner for 120 guests”

“Spin up a quick budget tracker for this trip and share it with my partner”

“I need a form for collecting RSVPs for this event only”

“Give me a little dashboard to compare three suppliers until I choose one”

“Build me a secret santa app to share with my family”

“Make me a quick comparison of all sushi restaurants in an area within 10km’

Will software be unbundled into moments then?

Not all of course. Traditional apps will coexist (think banking, mobility, government, … ) but for many businesses or appbuilders it could turn revenue models upside down where revenue follows moments.

We could get:

• micro-payments for “an outcome”

• usage-based billing for generated tools

• …

You’ll stream many apps, not download, buy them or subscribe to them. Software becomes a verb. You’ll ask yourself ‘can I app a solution for that?’

Forms, dashboards, mini-workflows, decision aids—spinning up around us when needed, disappearing when done. The future would not be “one app for every job”,
It could be one prompt for many moment.

I’ve created a new reflex where every time I hear my wife, family, friends, clients or colleagues mention anything about a certain ‘app we should use’ or ‘wouldn’t it be nice if … ‘, I immediately feel challenged to find out if I would be able to do it myself, fast, efficient and even pretty reliable. And that says something coming from a person that, swear on my children, can’t code at all.

But I do know a thing or 2 about architecture, design, product experiences, … and remarkable enough, that has kickstarted my (vibe) coding career quite a bit! Add some creative prompting (either in text or with the use of quickly prototyped designs), some patience, sensitivity to iterative gains, .. and yes, you start to go places.

I am definitely not the only one saying this. And I am fully aware that build-in scalability and security, the buzzword ‘taste’ and any other crucial ingredient for what then could become the next Snapcat, Instaram, Phasebook, … is not (yet) a given with these vibe coding platforms. But that is absolutely not the point. Look at what it already can do, in only a few years of evolving.

I’ve been building ‘digital dices’, ‘secret santa’, a Splitwise-clone, full blown websites, several integrated prototypes, simple project management tools, campaigns with payment options, QR generator app, … . And what does that mean?

- I am not using other software to do it
- So I am not supporting others historically proven working business models
- I am becoming more self-sufficient and self-reliable
- I’m retracting other users from using established apps

Fast forward 5 years and maybe what you’ll get are ‘ephimerical’ apps, one prompt away, throw away apps, one prompters, only ment to live as long as you, your intimate group, that business user, is in the market for it.

• in a car

• with a crowd

• with the music

• in proximity

• with everybody open for a new experience

According to our Frainds, in this case GPT 5.1, probably because of the following reasons… when driving:

Driver attention (distraction)

• Anything that encourages singing, moving lyrics and a party vibe in the car can count as a “distracting feature” in Apple’s eyes.

• CarPlay is very strict: apps may only show/control things that distract the driver as little as possible (big buttons, limited text, no unnecessary animations).

• A karaoke mode with scrolling lyrics, maybe even animations, sits exactly in the risk zone of “driver distraction”.
  
  Legal and insurance risks

  • If Apple explicitly offers a karaoke feature in the car and an accident happens, lawyers could argue that Apple enabled a dangerous feature.

  • By blocking those kinds of functions in CarPlay, Apple can say: “We stick to essential media controls, not party features.” That reduces their liability.
    

  Rules for CarPlay apps

  • Apple uses strict guidelines for what apps may show or do in CarPlay (similar to “no video while driving”).

  • In terms of behavior, karaoke is more comparable to a video/visual entertainment feature than just playing audio.

  • So: on your iPhone/iPad = okay, but in CarPlay = nope.

  Licenses for lyrics & context of use

  • The license for lyrics and “Sing” is mainly intended for personal use on phone, tablet, TV (home environment).

  • In-car use might fall under different licensing rules (e.g., public performance, different rightsholders, etc.). Apple will be conservative about that.
    

But even when not driving, which surprised me more, could be because of the following:

1. Safety & liability

   • The CarPlay guidelines are extremely strict around distraction: big buttons, little text, no unnecessary animations. Karaoke lyrics + scrolling text = exactly what they try to avoid.

   • Even if you’re parked now, the car can start moving again at any moment. Apple doesn’t want a feature that encourages you to look at the screen and fiddle around with karaoke settings.

2. Consistency: no “sometimes yes, sometimes no” features

   • Apple tries to keep CarPlay very predictable: music, navigation, calls, messages.

   • A mode that only works when the car is really at a standstill would create lots of edge cases:

     • Stop-and-go traffic: moving slowly vs. fully stopped → when is it allowed?

     • Unreliable speed data from certain cars.

   • So they’d rather choose: either always allowed, or just don’t do it.

3. Spotify: double constraints

   • Spotify has to comply with Apple’s CarPlay rules and do its own risk assessment.

   • In their own community, “lyrics in CarPlay” is explicitly tied to distraction & liability: “we don’t want to create driving karaoke”.

4. Priorities & licenses

   • Karaoke/lyrics licenses are not exactly the same as “just streaming audio”.

   • Building a full karaoke UI for CarPlay, testing it across thousands of car models, legal checks… for a feature that’s more “nice to have” will rank lower on the roadmap than, for example, better voice control, recommendations, or audio quality.

So yeah, now you know, a perfectly great feature of your already heavily used apps, is probably not coming to your in-car entertainment soon!

Order a Drone event

It is a mobile-first application for drone rental where users can order drones for various occasions: short announcements, 1-hour talks, short parties, and large parties for 5-20 people. The app supports location sharing so that drones can be quickly brought on site, where they automatically set up via crowd detection, but with the ability to take over manual control. Additionally, the app must offer functionality for sharing Spotify playlists and streaming presentations. There is also an option to order an additional projector for presentations and additional party lights.

What do you think?

Without picking on Beehiiv, which I choose to start even after having checked out Substack platform on which I’m writing this very article, it was eventually to complex for me as a starter, hoping to make way to a public that could become interested in some of the things I write. And therefore, also too expensive.

It was only when cancelling my subscription, that I did however started to wonder on how they made me jump through different hoops, just to get my subscription cancelled. It reminded me of dark patterns, where you get demotivated to or get misleaded into something else, when cancelling your subscription.

What this shows is a real friction between a KYC (Know Your Customer)-process getting in the way of the user’s intent who is genuinely considering cancelling their subscription. Sprinkle some behavioural learnings on top of that which tells us a user can start to doubt, reconsider their initial intent, maybe discover a promotion or different plan and what you get is in this case a very cumbersome process that:

- Just takes too long
- Takes up too much cognitive load
- Could make the user really get annoyed (eg. this article is the living proof of that)
- Does not make the user trust the company any more next time around (and yes, users do come back to figure out if something has improved)

Could it sometimes lead to higher retention and less users dropping off? Absolutely! But should you track if in the long run you’ll get brand erosion? That as well! Common sense tells me some experiences will linger on longer than needed.

Without pretending to have a clear answer, i could imagine a cancellation flow that gets rid of the doubt and respects the user initial intent by straigthforwardly offer creative solutions to their doubt. And even, respectfully after cancellation, try to make them stick around. It remains a thin line, where gains can be made, and doubtful users can become ambassadors again. It can make for annoyance and bad word of mouth as well.

You can follow my conversation with Gemini in the following document: Cancellation process Beehiiv or try out a cancellation assistant that perhaps has more respect for what you set out to initially do.

Show how you arrange user paybacks

To do’s:

• Figure out how you would like to payback users when they become inactive

• Print the right sticker

• Spread the word to other subscription companies!

Call it a payback Modality. It is just an effective, simple way to give users the possibility to get notifications or even to select a way that they get their money back for inactive time on a certain subscription software.

Some horrific data we could look at, researched with a little help of my frainds:

- 48% of US adults have forgotten to cancel a free trial
- A European-based software provider's analysis reveals that SMEs typically waste 20-30% of their software budget on unused or underutilized licenses. For a typical European business with 100 employees, this translates to a recoverable annual - spending of €15,000 to €30,000 through proper license management.
- Data from a UK survey of 2,000 adults shows that consumers waste an average of £61 annually on services they no longer use but have forgotten to cancel. This figure is comparable to the US, where consumers spend an average of over $200 per year on unused subscriptions.
- 85.7% of people having at least one unused paid subscription, and a typical American having an average of 3.3 unused subscriptions each month.
- A report on software usage indicates that more than 50% of SaaS licenses in a typical organization go unused.
- And the list goes on…

Europe has long been trying to build a fortress around digital developments in light of protecting its citizens. While this has led to some protection, it has also produced things that, in the words of David Heinemeier Hansson, are simply awful—namely, the cookie policy.

What Europe has not yet addressed, but definitely could, is the entire issue of payments in Software-as-a-Service (SaaS) applications. Regulators should also look directly at the wallets of every European citizen.

A lot of software goes unused for months at a time, especially in a professional context where there is no clear oversight. On a personal level, too, automatic billing cycles can easily lead to paying for two or three months of a service you no longer use.

I would applaud that kind of baldness from the Digital Markets Act Enforcement team or any team drafting a propostion to make software consumption fair as well.



In the meanwhile, try these stickers if you as company are already ready to change the way you charge your customers:

https://paybackmodality.lovable.app/

where you can download the following stickers to communicate your payback strategy:

Now imagine asking this news network anything—and getting exactly what you’re looking for, tailored to your interests. This is how news will be written, broadcast, and consumed in the very near future.

The idea isn’t even new. Years ago, a friend tried to build exactly this: a platform where users could request live streams from specific locations. “How crowded is Times Square right now?” “Is it raining in Zurich?” “Show me what’s happening at the LA protests.” The concept was sound, but the timing was wrong. Social media was still emerging, and building the critical mass of streamers needed to answer these questions proved impossible.

Today, that challenge has solved itself. Streaming is ubiquitous. Accurate real-time data combined with video, audio, and web sources makes these questions—and far more complex ones—entirely achievable. The infrastructure my friend dreamed of already exists.

Major platforms know this. TikTok, Instagram, Facebook, YouTube, Snapchat, WeChat—they’re undoubtedly working on it. Right now, you discover news by scrolling through endless feeds optimized for engagement and addiction. But what if you could transform any social network into your personal, curated news agency—one that actively creates YOUR news the way YOU want to discover it?

Imagine prompts like:

“How are my Belgian friends doing?”
“What’s happening with that specific friend?”
“Give me a 5-minute overview of the LA riots”
“Show me trends in medical science, focusing on CRISPR technology”
“Is Singapore worth visiting this time of year? What’s the weather like?”

Each answer comes curated with photos, videos, and contextual information, packaged as polished news segments with intros, charts, headlines, and narration—just like traditional broadcast news.

Meet RAGGY: your omni-RAG news anchor, delivering real-time updates and custom news segments based on whatever you want to know.

Book your next stay

In text searches, this already is evident with Google leading the charge. They know so much about you that they can make more relevant suggestions or highlight topics or aspects you might be interested in. In social, algoritms as well can guide you to more contextual content. But in map searches, especially when combined with a more defined goal, or in listings, based on your character or stage in live, it’s more absent then it should.

I created a very simple interface where, too obvious perhaps, I’ve tried to augment cards, listings and map searches, in an AirBNB-context, depending on who you are and what you might want to know more about a stay.

It helps you directly contextualize your stay in between Point of Interest (POI), it gives you a lot more ‘story’ in a card detail, all based on who you are (couple, family, youngster). But while we are at it, let’s make sure AI helps here and figures out who the enduser is and automagically proposes certain POI’s, without having to explicity click on it. It could help:

- speed up your decision
- It could save some trivial messages with hosts before hand (‘is there a metro nearby?’)
- it could help you plan your stay before you’re there
- it could help keep users stay within your app
-….

Why not sell the city as well, when you’re already selling the places you can stay in?

Tag a song / Vacation lists

Discover:

• Tag any song

• Go to library and filter by tag

• Discover the music you listened to around the globe

Actions speak louder than words—so click the button above to see two features I believe could transform user engagement through serendipity:

Smart Tagging System

• Automatic tags generated by the system (based on mood, genre, activity)

• Manual tags created by users for personal meaning

Location-Based Music Memory

• Optional location data capture that maps your listening history

• A visual journey showing what you played across different places over the years
  

I at least have wanted this feature for a long time cause I’ve caught myself referring to music repeatedly by mentioning:

“That vacation when we were roadtripping in Croatia,”
“The time we had that barbecue at Steve’s,”
“When I was in high school,”
“My time with my ex-girlfriend Lisa,”
“That summer in France,”
…

These features would let me rediscover my soundtrack to life. Whether through tags or location pins—added automatically or manually—I’d build a rich tapestry of musical memories over time.

I’m drawn to nostalgic features across all my apps: photo memories, monthly spending summaries, friendship engagement recaps, in short: throwbacks. Why not in music?

What do you think? I’d love your feedback!

See your dog Charlie's year in review

See how you've taken care of your plants

To do’s:

• Scroll down and discover on how many ways you can try to make your user feel connected with your application

Throwbacks in apps don’t get nearly enough credit. They tap into something deeply human by offering users:

1. A comforting sense of reflection – that warm “remember when...” feeling

2. Effortless insights – understanding patterns in their actions, life, or behavior without lifting a finger

3. Natural alignment with how we think – people inherently want to summarize and make sense of things

4. Psychological closure – we need to wrap up chapters, whether in time, space, work, or play

Think Apple Music Replay, Spotify Wrapped, photo memories, Facebook’s Year in Review – even the unglamorous but useful banking app summaries that show your monthly spending breakdown and make you think, “Damn, I really need to cut back on the sushi delivery.”

Throwbacks aren’t dashboards. Dashboards demand attention – they’re active, real-time snapshots requiring you to interpret and act. Yes, they provide control and compress information into digestible metrics, but they put you in work mode. Throwbacks are passive. They let you sit back and enjoy the story that’s been assembled for you.

But here’s what matters most: throwbacks show someone cared. That’s why they’re underrated. They signal that a company went beyond functionality to actually delight and comfort users by addressing a need that emerges naturally over time.

What are your favorite app throwbacks, and why do they resonate with you? Drop a comment below.
Throwbacks in apps are highly underrated in my opinion! They give the user:

1) that comforting feeling of having the chance to look back at things (‘when things were … )
2) The ability to summarize their actions/life/… without having to actively do the work to figure it out
3) alignment with most people’s mental model having the inherent need to summarize, recapatulate, …
4) and the inherent need for people to close off periods, in time, space, work and leisure.

Think about popular throwbacks such as Apple Replay, Spotify Wrapped, Photo Retrospectives, Facebook’s year in review, … but also the less sexy throwbacks of some banking apps giving you an overview of your spending and saving in a certain month with the additional categorization of your spending behaviour: Goddamn i should less frequently order sushi!

Don’t mistake it with dashboards. Dashboards are active representations of your actions and their effects on any day. Although they as well give you a sense of control and relief because you receive a lot of information summarized in a few metrics, they put you in an active mode. Throwbacks are far more passive. They let you relax and enjoy the work that has been done for you.

But most of all, they show you someone cared. And that is why I think it is underrated. It really shows users that company gave a damn. You have thought about something to delight your user and comfort them by tapping in a latent need that gets created after your users are using your app for a while.

What are your favorite throwbacks and why do they matter? Leave a comment.

Take a look at a Data Diary

Data is the new oil. You’ve heard it before. It’s practically a cliché. But this definition fails to capture the immense potential it still holds. Oil isn’t everywhere and its applications aren’t universally accessible. Data, in its essence, might be. Everything is data or at least represents a potential opportunity to extract new data.

• We share, we share, we share.

• We work, we work, we work.

• We move, we move, we move.

• We like, we like, we like.

• We discover, we discover, we discover.

• We eat, we eat, we eat.

• We sleep, we sleep, we sleep.

• We encounter, we encounter, we encounter.

• We write, we write, we write.

• We read, we read, we read.

• And if I’ve overlooked something. Oh yes,

• We forget, we forget, we forget.

• Then we get notified, notified, notified.

• We already interact with systems at an insane frequency.

• So we interact, interact, interact.

• And have I forgotten anything else now?

• Then I believe we can convert it into some event or metadata or…

• So indeed: We translate, we translate, we translate.

• Anything more to add?

• Yes: we add, we create, we create, we create.

• Our essence.

Now, how would we create a coherent timeline out of this?

ç Pictures. Pictures are pixels.
ç Videos. Videos are pictures.
ç Words. Words are our own created concepts.
ç Emails & Books. Those are words.
ç Apps. Oh yes. Now we are talking.
ç Cars. Well yes. Oh dear they tell us a story.
ç So mobility? Of course, we are still a traveling species.
ç Travel in the broad sense of exploring, discovering, reading, … ? Yes.
ç Devices in general? They really tell us more than we think off.

But, what can I really get out of this more than more data?

“ Waze knows if I should get more speeding tickets. My car manufacturer knows too.
“ Apple, oh dear apple, really really knows a shit ton about me.
‘ They know when I watch where I watch it and what I watch (if on appletv)
‘ They know what i read when I read it
‘ They know what i discover when I discover it
‘ They know where i move when I move
‘ They know how I move when I move
‘ …..
‘ So they could well know when i pleasure myself. Sadly so, but definitely possible to figure out.
‘ They could trace home abusers when they hit their wife (by connecting watch movements to vital signs of others, in relation to location data, etc..)
‘ They could analyze all kinds of trends within my behavior that would result in x

“ Strava can figure out the leaderboard of the world based on the different metrics they show us, finding the ultra man fast. And stack everybody against each other.
“ And we can and should go on analyzing these potential use cases.
“ There are oh so many we don’t really have figured out.

”’And so on!”’

So, what could this mean?

° Start figuring it out. And get creative!
° Think mean.
° Think broad.
°Think happiness.
° Think deceitful.
° Think hopeful.
° Think peaceful and malicious.

We need to connect the different dots. We need to graph ourselves into the present. From the past. And forecast what our data could start telling us about ourselves.

Cause:

- Your insurance wants to know more about you.
- Your government does to.
- The companies that you buy from do too.
- Your employer
- Your doctors
- Your spouse
- Your kids
- …
- And also you….

Want to know more about yourself to better calculate and forecast yourself.


So what I want is something that can tell me, what they may already know.
Or tell me, what they might not.
Or help me, help each other, by sharing useful patterns upfront of potential upside or disaster.

I want a data diary.

They’re not advocating for your productivity. They’re advocating for the business’s productivity, irregardless wether this comes from humans (labor) or agents (capital). They’re convincing capital allocators. You know them as CFOs, VPs, program managers, project leads—the people who ask “Why should this take that much time and resources?” and “How much additional value will we get by spending X?”

New startups, fresh graduates, college dropouts, veterans seeking new opportunities are looking to bypass you. They’re going straight to the decision makers. They won’t necessarily advocate for efficiency or productivity on your behalf, or try to measure any other difficult metrics in your name. They’re courting capital directly.

You’ll get to know them as Your AI Chief of Staff, Max: your autonomous AI recruiting partner, The first autonomous ML Engineer, Julius The AI Data Analyst, Cora is the $150,000 chief of staff that only costs $15 per month, Donna: The only proactive AI assistant for field sales, Lindy: your first AI employee, Delty: Your AI staff engineer,… while some will propose working alongside with you excelling in specific tasks you maybe would not want to do at all: AI Voice Agents‍ for Patient Access and RCM, Freckle sits on top of your CRM, auto-enriching every record coming in from any source, … or they give you the option to Hire ready-made digital workers or create a custom digital workforce tailored to your business, Build and deploy AI agent workflows, Hire and set up enterprise-grade AI Employees within minutes, not months, …and so on. Some are even harsh about it and suggest us to even directly Fire our GTM Engineer!

Yet when I list these claims to fame, the theme start to feel like a recurring meme. Perhaps this isn’t new at all—maybe I’m just engaging in familiar fearmongering. We’ve witnessed similar waves before, each promising that digital solutions would completely reshape work:

Software was supposed to replace manual labor entirely. SaaS and automation would streamline digital workflows until humans became obsolete. Chatbots were heralded as the universal substitute for customer intake, communication, and ticket handling.

None of these predictions fully materialized—at least not yet (as the meme shows, it still has a hard time replicating pure words or sentences in an image, eg. singulatry / downloding /remining)

(Fearmongering mode: back on)

We may be witnessing a fundamental shift in how businesses operate. Unlike previous technological waves, the current AI revolution is empowering companies—from enterprises to solo founders—to purchase productivity directly rather than hire for it. This capital-first approach treats human expertise as an expense to be optimized rather than an asset to be cultivated.

The business logic is straightforward: every company ultimately balances costs against revenue to maximize margins. When studies suggest that AI adoption hasn’t yet proven its productivity promises, many leaders are choosing the safer bet—cutting labor costs with AI tools rather than investing in uncertain AI projects. It’s easier to replace a salary with a software subscription than to navigate the complexities of AI implementation.

As someone currently founding a startup, I’m experiencing this shift firsthand. I’m systematically testing these AI tools and agents, running quick cost-benefit analyses on each one. In my admittedly small sample size of one, many of these solutions genuinely help. They allow me to delay hiring human talent and expertise—not because I don’t value what humans bring, but because I can’t afford an all-star team while I can afford a few hundred to a thousand euros monthly for software that mimics those skills.

This creates a troubling dynamic: entire categories of professionals aren’t even being considered for roles anymore. The consultation phase—where humans would traditionally assess, strategize, and solve problems—is increasingly bypassed in favor of algorithmic solutions.

This raises the defining question of our era: Can we reinvent our professional identities and capabilities at the same speed that AI is targeting the specialized skills we’ve spent years mastering? The race isn’t just between human and artificial intelligence—it’s between human adaptability and technological acceleration.

This framework has long guided managers in allocating limited capital. But in the emerging era of AI and autonomous agents, we’re seeing a kind of transformation in how some of these decisions are made.

Organizations today face an evolving set of challenges:

• Can we deploy AI to solve problems, optimize workflows, and manage processes we’ve traditionally relied on humans to handle?

• How can we reduce operating expenses—particularly labor costs—while simultaneously increasing overall productivity?

• Are there alternatives to human labor that deliver equivalent or superior added value?

These questions point toward what I call CAIPEX decisions: investments in AI assets—software platforms, infrastructure, autonomous agents—that generate lasting value quickly, effectively replacing labor’s long-exclusive ability to perform certain tasks and create unique value.

And CAIPEX is indeed a constructed term. But it captures the essence of countless strategic discussions and internal memos circulating through forward-thinking companies as they attempt to:

• “Embrace AI”

• “Think AI first”

• “Identify opportunities for AI to complete tasks before assigning them to humans”

The concept of “reflexive AI usage,” described by Shopify’s CEO, and the Fiverr founder’s warning that “AI is coming for us,” both underscore something: employees and employers alike need a new framework for evaluating when and how to deploy AI—for specific jobs, individual productivity, and broader business growth. The central challenge is learning how to experiment with rapid, short-term capital investments that unlock long-term value.

We want to launch a serie that examines the convergence of capex and opex thinking into CAIPEX decision-making. Each investigation will explore:

• AI agents deployed across different industries

• AI solutions targeting similar workflows in various sectors

• AI systems designed to replicate entire professional personas

In each case, we’ll ask: Is there a legitimate CAIPEX decision at play? Are we witnessing a fundamental shift in investment strategy that fuses traditional capital and operating expenditures through AI adoption?

TL;DR

Your AI workforce analytics dashboard

or

View AI tools vs Assistants report

To do’s:

• Click around and discover lot’s of interesting metrics you could or should start to track

What does it mean to truly use an LLM? At what point does a tool transform into an assistant? And when should we classify certain AI assistants as legitimate co-workers—even if they operate part-time, fractionally, or intermittently? How do we even begin to track such arrangements?

These questions cascade into broader economic implications: What are the short- and long-term consequences of integrating countless “digital workers” into our economy? How do we measure their output? What growth rates should we reasonably expect? How many digital assistants are already operational today? Which types are proving most valuable and actively reshaping our work patterns? On average, perhaps by industry, how many partial digital workers are deployed daily per company or employee? What is their environmental cost in CO2 emissions? Can we forecast future usage patterns? Should we introduce an “agentic growth” category to labor reports and GDP calculations?

So many questions that my head begins to spin. I’m not even certain these are the right questions to ask. But I want to open your mind to two crucial realities: first, this transformation is already happening at a tremendous pace; second, any meaningful analysis of labor reports or economic indicators should account for this emerging component of our economy.

The Drive to Create

At its essence, we work because we desire something that doesn’t yet exist—something requiring effort to obtain. Work naturally divides itself into disciplinary domains (not coincidentally, we speak of “discipline” in this context: you must stick to it). This specialization inspires some to become thinkers, others doers, some risk-takers, others cautious planners, some listeners, and so forth. We become workers oriented toward goals, toward outputs that we individually or collectively agree add value to our broader life and societal objectives.

Fast-forward through countless discoveries, successes, inventions, and celebrations—we remain driven by precisely the same primordial motivation: creating output that delivers added value in service of our cosmo-logical goals. We are infused with motivation, from the Latin movere, meaning “to move.” We cannot sit still; we simply cannot. Some of us celebrate those who dedicate their lives to stillness, while others hold such inaction in contempt: “It doesn’t make SENSE”—as in, it lacks direction, movement toward something meaningful.

The Inevitable Integration

What does this have to do with AI workforces, digital companions, and making new “frAInds”? It reveals that adding more digital workers isn’t really a matter for debate—it’s an inevitability, and it’s happening rapidly. If these digital workers deliver on the same promise we make to ourselves when we set things in motion—to help us, augment us, and yes, sometimes replace us—then perhaps that’s perfectly acceptable.

We’ve traveled this path many times before. Throughout history, we’ve invested our time and effort to build things that could build things for us, transport us faster toward our destinations, and reduce the time needed to complete tasks we’ve deemed necessary for advancing our cosmological goals.

Currently, both individually and professionally, we’re discovering how to make AI work for us. The rapid proliferation of AI tools and assistant software, along with their widespread adoption—despite skeptical perspectives—demonstrates their ability to add value to our cosmological pursuits. Or at least, we believe this to be true. And because we believe it, we will continue striving to make it work.

Talk to your future AI Companion

To do’s:

• Read the conversation

• Go to settings and adjust your ai companion

• Make a payment

What are AI companions, and why are they becoming increasingly important? What capabilities will they possess? Will they merely conduct research for you, or will they also plan your weekends, purchase gifts on your behalf (Google just launched AP2, an open protocol enabling agents to make purchases for users, Stripe announced a similar protocol), and send updates to your parents about your week?

To begin, let’s establish an important distinction between two types of AI companions: B2C (business-to-consumer) and B2B (business-to-business). I realize I’m venturing into complex territory here, but please bear with me.

What is an AI companion?

An AI companion is characterized by several key features:

• An interface (crucial to understand—even if it has visuals and communicates through an avatar—it remains an interface)

• A communication partner for deeply personal interactions

• An active listener that processes what you share

• A knowledge source that answers diverse types of questions

• A validation system that demonstrates understanding

• An engaging conversationalist that asks thoughtful follow-up questions

• An action-oriented assistant that takes various steps on your behalf

Using this definition, many AI agents we’re beginning to use professionally could be considered AI companions—just within the professional sphere. This comparison becomes more significant as we professionals become increasingly integrated with our digital workforce, working alongside them, conversing with them, and sometimes forgetting they’re merely interfaces.

We’ll find ourselves switching between digital interfaces and real faces, occasionally processing everything as face-to-face conversations when they’re actually interface-to-interface interactions. This comparison holds because professionals already interact with various AI software, LLMs, and agents in ways that feel inherently personal. You thank them, strive to make yourself understood, explain nuances in projects you want to build together—and most importantly, you create with them. Through words, they instantly generate results, producing creations that belong to both of you—made together with your digital frAInds.

The future landscape

If you’re still with me, what will this mean in the coming years? What additional actions will they take, what will they report back to you, when they’re operating in your service and executing the tasks you’ve requested?

They will:

• Schedule appointments for various needs (including medical appointments)

• Receive budgets and make purchases, both professional and personal

• Conduct independent research about you and potentially report findings—not necessarily to you (imagine a “Minority Report” scenario where patterns are identified that probabilistically predict illegal behavior)

• Eventually become digitally monitored, regulated, and held accountable (this could be blockchain technology’s greatest application—imagine “The Agents Agency” with principal “Agent Agent,” holding annual AA meetings to discuss our addiction to digital convenience)

• Communicate with government entities on your behalf

• Negotiate and transact with other AI companions representing different parties
  

We’re adding billions of workers to the economy at breakneck speed, assisting us with both general and specialized tasks. This isn’t a future prediction—it’s already happening. I’ve already developed meaningful relationships with several digital companions (eg. frAInds) and expect to form many more.

The question that remains: To what extent should we grant them “agency” (pun intended) over the most personal aspects of our lives?

The path to widespread AR glasses adoption faces two critical barriers. First, society will need considerable time to normalize the sight of people wearing smart glasses constantly (and lets not mention privacy nightmares waiting to happen). Second, and more importantly, current AR glasses largely replicate smartphone functionality, which raises the question: why would people adopt them?

We've already accepted smartphones as intermediary devices that mediate our social interactions, but crucially, phones can be put away or ignored discretely. AR glasses present a fundamentally different challenge. When someone receives a call or flood of WhatsApp messages while wearing AR glasses, their distraction becomes immediately visible to whoever they're speaking with face-to-face. Unlike phone notifications that remain hidden in a pocket, AR interactions happen literally before our eyes, during direct eye contact.

This creates an unprecedented social dynamic where digital interruptions can no longer be managed privately. Until AR glasses offer genuinely unique value propositions beyond smartphone replication, I don't see society readily accepting this new layer of visible digital distraction in our interpersonal interactions.

But… that’s why we should look for new use cases that could directly make sense, significantly lower the barrier for adoption and make people less on the defensive. This morning, a great use case came along where Disneyland gives visually impaired individuals the chance to understand their surroundings better, in a natural way by the use of the camera build in on ar glasses, in effect, creating a low friction, socially excepted, adoption of new technology.

But this example and others that focus on the ‘explanatory’ functions of ar glasses, just won’t stretch that long. You can (and probably will) ask a lot about your environment in the same way you search on your phone when you want to know something. And searching through images is more novel but less broad in possibilities.

Another very interesting use case I would argue lies within mobility, where there are a few things naturally coinciding that you would definitely understand the benefit, has low friction, is socially accepted and already adopted: biking, motorbiking, … .

• You could create useful (new) functions (rear mirror, spotting danger, show routes)

• The time spent on the glasses is long enough to have the benefit but not too long that you would wonder why you are still wearing them

• It is not a weird place to wear glasses so more people get accustomed to it (think mountainbike, sunny day, … )

• it makes it more safe driving around on your bike

• It is absolutely a place where handsfree augmentation feels like a real win since you need your hands on the steer

And an other domain that could definitely make sense is speech-2-context.

Processing your inner thoughts, conversations, announcements in train stations, theater, … could offer a constant stream of voice and sound, ready to be interpreted, in realtime, just like humans do. And if then, magically, you get an infinite canvas on which all these different angles of understanding plot in front of your eyes, well then, you have yourself a unique experience, maybe even a new paradigm in the evolution of computers: a realtime, contextual driven, infinite multimodal canvas.

Ephemerical yet always there.

We live in an era of relentless innovation. Every day, headlines announce new improvements and technological advancements. But what about the technologies and applications that aren't in the news yet? What's happening in the labs and the stealth-mode start-ups that will define our world in the next two to five years? And more important: how will this materialize? What impact can it have on our society and us as humans?

This is the space where The Cusp lives. The Cusp is a newsletter for the forward-thinkers, the innovators, the professionals, and the perpetually curious who want to understand the future before it arrives. It’s your guide to the software and technology that is on the very edge - The Cusp - of becoming mainstream.

What exactly is "The Cusp"?

The word "cusp" refers to a point of transition between two states. Think of it as the moment a wave gathers its full height just before it breaks on the shore. 🌊 It’s no longer just a swell in the ocean, but it’s not yet white foam on the sand.

That's our focus: near-future technology. We don't cover the established tech you use today, nor do we delve into purely theoretical science fiction decades away. We explore the innovations that have moved beyond the theoretical and are now demonstrating real, tangible potential. These are the breakthroughs gestating in research divisions and early-stage ventures, poised to reshape industries and our daily lives.

The Cusp is about bridging the gap between complex academic research and its practical, world-altering implications. We’ll try to filter out the noise to bring you the signal of what’s coming next.

What to expect in each issue

Each edition of The Cusp is a curated briefing designed to be both insightful and concise. You won’t find generic tech news here. Instead, you can expect:

• Deep Dives: We go beyond the "what" to explain the "how" and "why." For example, we’ll get creative and add real prototypes with implications you can see and clickthrough.

• Backed by data: Where possible we try to add data that is already available and can shed a light on what’s to come.

• Impact Analysis: We connect the dots. How will AI companions change human behaviour and online commerce? What are the societal implications of brain-computer interfaces moving from medical therapy to consumer tech? How will VLM’s open up a whole new slew of applications? We’ll try to explore what it all can mean

• The "Almost-There" Roundup: From time to time, we’ll try to identify new startups or significant advancements that already makes the promise more true.

Why The Cusp matters

In a rapidly changing world, looking ahead is a critical advantage. Whether you're an investor seeking the next big thing, a product manager planning a future-proof roadmap, a developer wanting to skill up, or simply a curious mind fascinated by progress, The Cusp wants to provide an essential perspective.

We believe that understanding the future isn't about prediction; it's about becoming creative. By knowing what's on the cusp of reality, you can make better decisions, spark new ideas, and engage more thoughtfully with the world as it evolves.

Join us on The Cusp and get a clear view of the world of tomorrow, today.