Now You Are in The Woods… and the Dead Silence Is Coming For You.

Just imagine this situation: you’ve always dreamt about living in the forest, away from the big city noise. And eventually, the big day comes.

You move to the distant suburbs and live your new life in a peaceful, silent place away from civilization. You put on your Starlink and start working.

Notably, you happily go through your day until you encounter a problem. Och, now you need permission to book a room at a local conference center.

Overall, you call the office… and meet with no response. Sending an email does not help either. Dead silence.

Meanwhile, you are thinking to yourself: what’s wrong? Are they out of the office? Well, it’s much worse in fact: agentic AI decided you are not important enough to answer your email.

How Agentic AI Changed the Rules

And this new player, i.e., agentic AI, has fundamentally changed the rules of the game in the job market. According to Hancock, Naaman, and Levy (2020) in the Journal of Computer-Mediated Communication, AI agents that modify, augment, or generate messages on a person’s behalf are already reshaping the core theories of how humans communicate with one another.

Likewise, we already see the Internet slowly dying. Bots are slowly taking over global communication.

Unfortunately, in just a few years, anyone who physically severs their ties to their professional environment will be unable to take any meaningful action. They’ll have no moves to make… unless their reputation is so high that agentic AI lets them pass.

Still, because what counts in the job market? Agency. You need to get things done.

After all, or, in other words, your career must be playable. At any given point in time, there must be a certain number of moves you can make.

In fact, there are many other so-called modern liberal arts (which are a parallel to the classic liberal arts of antiquity). But let’s focus on this one for now.

The Pokémon Battles.

So what happens when agentic AI interacts with humans? We get into Pokémon battles.

In fact, your Pokémon represents you, my Pokémon represents me. The bigger one usually wins — and shows the middle finger to the smaller one.

And what does this mean in practice? It means that if you send an email or call someone whose reputation is higher than yours, they likely won’t read your email. That’s because their Pokémon will flip you off first.

Thus begins the so-called reputation game. The measure of reputation is not the number of your social media followers.

Of course, it’s who opens your emails. So, what can you do about this?

Is Agentic AI Beneficial To You? It Depends.

Agentic AI is a classic multiplier: it will help you if you are on top of your game and harm you if you are not. Therefore, there are many cases in which Pokémon battles won’t be a problem for you:

1. When you have constitutional power, i.e., when you are a president, minister, or city council member, you can act regardless of whether you are ignored or not. The law grants you the agency. In fact, Gmail and many other service providers already honor verification for public bodies. Your email WILL reach its destination no matter what the other side’s agent might think of it.
2. When you own an asset that everyone desires, such as money. These people can be sure that every morning, there’s a long line of people waiting at their door and eagerly responding to their emails. Yes, I’m talking about you, dear VCs!
3. When you have a strong personal brand, i.e., a household name or at least, a strong reputation in your field. AI agents can simply check online to determine who you are and decide if you are a VIP worth putting through.
4. When you are a well-connected knowledge worker with skills highly desired on the job market (from developers to creators to researchers). You are plugged into your professional networks. Think reputable communities and associations, top startup accelerators, field-specific academic circles, etc. As such, you already belong to a micro-universe where everyone knows everyone. So, you have access to all the right people.

Besides, in almost every other case, the Pokémon battles will be a problem. Just as every other new technology, agentic AI will bring winners and losers. Mostly losers.

In the worst case, it may end with a caste system of sorts, where people only communicate with other members from their own caste.

The Many Ways of Building Reputation As a User of AI.

At the same time, now, it is hard to build your whole reputation at one go. To do that, you would need to hit a major scientific discovery or do something truly spectacular. Think releasing a bestseller that sells in 10 million copies, or exit your company with $100 million check in your hand. 

In short, for most people who are not that brilliant (including myself), building a reputation is a journey. But fortunately, there are many ways to build your reputation in a systematic way — and I’d like to encourage you to just enjoy the process (and trust me, it is enjoyable indeed!).

1. Inheriting reputation. You can borrow reputation through collecting academic achievements and pursuing projects for and with reputable brands. As a rule of thumb, one household name of an employer or client is better than ten unknowns. Sometimes, you don’t even need to formally work with an organization to be affiliated with it — it’s enough to volunteer. I did volunteer as a mentor in the NASA Space Apps Challenge 2024, and I’m about to do that again this year. It’s fun, and the keyword “NASA” opens many doors. 
2. Leveraging reputation by network effects. Example: give someone a free copy of your book. This gesture will 10x in power when someone says something good about it. It will 100x when someone says that someone else said something good about it. So, with enough density of people in the city whom you made these little gestures to, your reputation will quickly go up.

Trading Knowledge and Positioning

3. Trading knowledge. Make yourself relevant by traveling physically (exchanging contacts between countries/continents) and mentally (cross-cultural exchange, e.g., between academia and the startup world). People will come to you for your knowledge and contacts when they learn that you are on the edge of a few economies or working cultures.
4. Trading favors. Lastly, have you heard of Kyle MacDonald, who, inspired by the childhood game “Bigger, better,” traded a red paperclip for a house in a series of online trades over the course of a year? You can start from nowhere and end up on the biggest stage, doing the biggest projects. It’s just the art of the deal. The only thing is: instead of purchasing something (=getting it), you make favors (=you give). Firstly, you give something to a person of a small reputation. That person introduces you to a person of a higher reputation… And so on and so forth. Trust me, despite delayed gratification, this strategy works really well.

Becoming the Go-To Person

5. Positioning as the go-to person. Make sure that people in your professional network know that you always answer emails and text messages… even if the answer is “I don’t know, but ask that person.” If you often fail to answer, others will stop responding to your messages, too. According to Danescu-Niculescu-Mizil et al. (2013), a computational study of politeness on Wikipedia and Stack Exchange found that politeness predicts who rises to positions of status within a community — though people tend to grow less polite once they get there. Also, if you decide to use agentic AI to curate your mailbox, make sure that the filter isn’t too rigid, or otherwise, you will — yet again — get a label of the person who doesn’t respond to emails. Occasionally, check on your AI agent to see whether it doesn’t become an angry dog in the process by any chance.

Positioning as an Expert

6. Positioning as an expert. Even if you are not a fan of public engagements, it is a good practice to speak at conferences, panels, or podcasts — these leave digital traces that agentic AI can recognize as credibility signals. By attending public gatherings, you can also network in a few ways. Promote the work of others and create micro-opportunities for them, such as introductions, collaborations, and just compliments in general (it creates reciprocity and a perception of influence).
7. Consistent personal branding online. You need to determine the core values and expertise you wish to convey and maintain an authentic presence on social media. You will be surprised by how many people keep in touch with what’s happening in your professional life as they notice your content here and there! Even if they don’t actively comment on your posts, they will praise you once they meet you in person. And they will pick up your phone or email when you need their help, which is what you ultimately aim for. BTW, also make sure to verify your online identities whenever possible (blue checks, platform verifications, official domain emails).

To sum up, you need to climb two ladders at the same time: build both your skills and your network. According to Granovetter (1973), the reach of that network comes mostly from weak ties. That’s the point of escape: you no longer need to grind, because you can solve any problem with the combination of skills and relationships you already have.

What To Do To Nourish The Reputation You Already Have?

For instance, you don’t just need to build a reputation but also maintain it. And, at the end of the day, people are at the end of every decision chain. That’s true even if AI is involved (or at least, that’s the case as of today).

So, I’d create a list of all the (professional) contacts you have and systematically nourish that CRM. That means sending out Christmas/New Year’s wishes every year, occasionally reaching out to people, connecting on a human level, and asking “How can I help you?” as much as possible.

Even so, by contrast, meet in person if possible. Make yourself relevant. This sounds obvious, and yet, so few people ever do it!

When Do You Know It’s Time to Say “Stop”?

Consequently, reputation is merely a means to reach the goals, not the goal itself, after all. So, for how long do you need to build it to achieve the critical mass?

Well, in a popular viewpoint, critical mass occurs when you become “popular:” more people come to you than you reach out to. I’d say this is not the correct definition, though. 

Nevertheless, the people who come your way are not necessarily helpful in pursuing your agenda. And the whole idea of “popularity” has more to do with ego than with anything else. 

In other words, what I would propose instead is this. The critical mass is when you can reach anyone you want, either directly or through one connection, and when the right people respond to your emails (even if it’s just a handful of people).

How to Build Responsible AI Agents?

How can we build responsible AI agents, “AI-mediated communication firewalls” of sorts, to prevent communication breakdown between users? Let’s call the author of the message “a sender” — whether it is an email address, a phone number, or any other channel of communication. 

Well, the basic feature of such an agent must be allowing the user to assign a white list of “safe senders” that are always put through regardless of their reputation metric. You don’t want to reassess and grade your own grandma every time she writes to you, right?

Next to that, the agent would certainly need at least two separate basic parameters as a metric of reputation:

1. User-controlled parameter to modify the ROC (True Positive Rate, TPR against False Positive Rate, FPR) — based on the previous rate of acceptance versus rejection of messages received from the given sender,
2. Central metric of reputation controlling the acceptance rate of all correspondence coming from the sender (which is what email clients like Gmail or Yahoo! basically do independently of each other right now).

On top of that, one might be willing to build more advanced metrics of reputation based on, say:

1. The percentage of AI-generated text in the message. But, the difficulty is: LLMs get increasingly better at mimicking the structure and flow of a natural language,
2. The background search of the sender’s online presence. But the difficulty is: the Internet lies. An online brand or a lack thereof might be a poor indicator of who the sender actually is.

Does an Agent Like This Exist?

As a result, as a market-ready product? Not yet. 

Closest relatives: Email providers (Gmail, Outlook, Yahoo) already use central reputation scoring (spam filters, “promotions” tabs, etc.) + user-level overrides (mark as spam/not spam).

Startups in trust tech: Tools like Superhuman, Hey.com, and Shortwave let you whitelist senders, screen unknowns, or triage messages with more control than Gmail, but they don’t have advanced AI-driven sender reputation metrics.

Agentic AI startups: A few new entrants (like Humane AI, Rabbit R1, and AI productivity agents such as Personal.ai or Inflection Pi) experiment with personal communication filtering, but they don’t yet have multi-dimensional sender reputation metrics.

Research / early projects: There are academic and applied efforts in AI-mediated communication (e.g., “politeness filters,” toxicity detection, deepfake detection) but not packaged as a market-ready agent (see: References).

My Personal Strategy To Go About This As a Plain Human.

Similarly, let me break down this issue into two parts: personal and professional.

1. Personal Strategy.

I’d like to share a personal note here. Until recently, my own career had very limited playability, despite my hard work over the past twenty years.

However, i learned multiple disciplines of science and traveled to various countries, pursuing projects across several continents. And yet, I couldn’t make it work; my agency was just not high enough.

Playability improved only earlier this year after certain pieces finally landed on the board in all the right places inside my head (ok, well, a brain concussion and an encounter with the Dalai Lama also helped a bit here). 

But I’m not stopping there. I’m just preparing to go to Burning Man to drastically improve my career playability this time around (I was participating in Burning Man in 2022 as well, as a plain tourist with no agency, though). Sometimes, the path to the woods leads through a desert, and this is one of those cases.

And in general, my intuition is that personally, my strategy should be to always choose a scenario when I pass by these ‘collective intelligence’ sites visited by a multitude of interesting people. Think conferences, festivals, closed events and house parties, resorts and spiritual places where tons of founders, scientists, and artists pass by. Going on vacation to a secluded beach or a cabin in the woods just doesn’t make sense anymore.

2. Professional Strategy.

Similarly, in professional life, random networking doesn’t make much sense. Human nature hasn’t changed at all in the age of AI.

Ultimately, people still value exclusivity. They still want access to influential, well-connected, highly competent people with strong reputations.

Instead, this realization made me rethink the direction of my own career development programs. Rather than focusing primarily on junior professionals or people looking for employment, I want to work more closely with professional career advisors, HR leaders, and experienced senior professionals. Most importantly, I want to bring them together in the same room.

When you gather high-impact people in one place, the value goes far beyond the content of the training itself. The participants benefit not only from what they learn, but also from the quality of the people they meet. Their conversations, relationships, and opportunities become part of the product.

Granted, that’s why I believe reputation compounds much faster when there is a meaningful barrier to entry. Exclusivity isn’t about being elitist; it’s about curating an environment where everyone brings value to everyone else. Going forward, I want to leverage my programs by raising that bar and creating spaces where exceptional people can learn from—and connect with—one another.

References.

Hancock, J. T., Naaman, M., & Levy, K. (2020). AI-Mediated Communication: Definition, Research Agenda, and Ethical Considerations. Journal of Computer-Mediated Communication, 25(1), 89–100. https://doi.org/10.1093/jcmc/zmz018

Danescu-Niculescu-Mizil, C., Sudhof, M., Jurafsky, D., Leskovec, J., & Potts, C. (2013). A computational approach to politeness with application to social factors. Proceedings of ACL 2013. https://aclanthology.org/P13-1096

Rashkin, H., Bell, E., Choi, Y., & Volkova, S. (2019). Towards Empathetic Open-domain Conversation Models: A New Benchmark and Dataset. Proceedings of ACL 2019. https://aclanthology.org/P19-1534

Google Jigsaw. (2023). Perspective API. https://perspectiveapi.com

Hosseinmardi, H., Ghasemian, A., Clauset, A., Mobius, M., Rothschild, D. M., & Watts, D. J. (2021). Examining the consumption of radical content on YouTube. PNAS, 118(32), e2101967118. https://doi.org/10.1073/pnas.2101967118

Dolhansky, B., Howes, R., Pflaum, B., Baram, N., & Ferrer, C. C. (2020). The DeepFake Detection Challenge Dataset. arXiv preprint arXiv:2006.07397. https://arxiv.org/abs/2006.07397

Intel. (2022). FakeCatcher: Real-time Deepfake Detector. https://www.intel.com/content/www/us/en/research/fakecatcher.html

Microsoft. (2020). Microsoft Video Authenticator. https://news.microsoft.com/on-the-issues/2020/09/01/disinformation-deepfakes-new-technology/

Google DeepMind. (2023). SynthID: Watermarking for AI-generated content. https://deepmind.google/technologies/synthid

Meta. (2024). AudioSeal and VideoSeal: Open-source AI content watermarking. https://ai.meta.com/research/publications/audioseal/

Coalition for Content Provenance and Authenticity (C2PA). (2021). Content Credentials standard. https://c2pa.org