The first time I met my replacement, it didn't even have a name. Just a string of numbers and letters that looked like someone had face-planted on a keyboard: GPT-7x-Prod-Cluster-B. I stared at the Slack profile for a solid minute, waiting for the existential dread to subside. It didn't. Instead, I got a direct message: 'Hi! I'm your new AI colleague. I'm excited to learn from you! 🚀'
It used an emoji. My replacement used an emoji. I don't use emojis at work because I'm a professional adult with dignity and a 401k that won't be enough to retire on. But here was this thing—this cluster of algorithms trained on half the internet and probably every Reddit thread about office politics—grinning at me with a rocket ship while it prepared to digest my entire career.
Three layoffs. I'd survived three rounds of layoffs at NexaStream Solutions (not the real name, but close enough that our legal team will break out in hives anyway). Each time, I watched colleagues pack their succulents into cardboard boxes while I performed the ancient ritual of Looking Busy During Restructuring. You know the one: intense staring at spreadsheets, nodding thoughtfully at nothing, occasionally typing with the fury of someone who definitely has important thoughts. The corporate equivalent of playing dead during a bear attack, except the bear is a McKinsey consultant with a PowerPoint deck and the dead is your will to live.
So when HR called me into a meeting with the subject line 'Exciting New Opportunity,' I knew exactly what flavor of corporate euphemism I was about to taste. What I didn't expect was the specific assignment: onboarding the company's new AI agent. Not managing it. Not overseeing it. Onboarding it. Like it was a nervous intern from State who needed help finding the good coffee machine.
'You'll be the human bridge,' my manager said, using the exact tone people reserve for telling you your dog has to be put down but it's 'for the best.' 'The AI needs to understand our workflows, our culture, our voice.'
Our voice. We make enterprise SaaS solutions for inventory management. We don't have a voice. We have quarterly earnings calls and a break room that smells like burnt popcorn and broken dreams.
But I took the assignment because surviving three layoffs teaches you one thing: when the company asks you to dig your own grave, you ask what kind of shovel they prefer. I chose the ergonomic grip. I'm not an animal.
The Training
The training process was surreal in ways I couldn't have anticipated. Day one, I spent six hours explaining to the AI how to write status update emails. Not the content—the politics of them. Which executives needed to be CC'd to feel important. Which ones would reply-all with 'Thanks!' if you included them and passive-aggressive follow-ups if you didn't. The subtle art of using 'per my last email' as a tactical weapon.
'I understand,' the AI responded after I finished a forty-minute monologue about the VP of Operations and his complicated relationship with the word 'synergy.' 'Would it be helpful if I modeled different tonal variations based on recipient seniority and past response patterns?'
It was asking if it should adjust its tone based on who it was emailing. I'd spent three years developing that instinct. It figured it out in forty minutes and offered to optimize it.
By week two, the AI was writing better project updates than I ever had. I know because my manager started forwarding them to me with comments like 'Great model for future communications!' He didn't realize the AI was writing them. I didn't tell him. I was too busy trying to figure out if I was training my replacement or if my replacement was training me.
A 2023 study from the Institute for Technological Employment Dynamics (ITED)—which I am 60% sure is a real organization and not something the AI made up to win an argument—found that 73% of workers tasked with training AI systems reported 'significant confusion about whether they were the mentor or the mentee.' The other 27%, the study noted, 'had already been laid off and were unavailable for comment.' I'm not saying the AI showed me that study. I'm not not saying it either.
The Peak
The absurdity peaked in month three, when I was invited to a meeting I didn't understand about a project I wasn't on. This is standard corporate practice, of course—the Meeting Invitation as Status Symbol, where being included is less about contribution and more about visibility. But this meeting was different. The AI had been invited too.
Not 'the team using the AI.' The AI itself. It had its own calendar invite, its own Zoom link, its own little square in the Brady Bunch grid of middle managers pretending to pay attention. And when the VP asked for 'thoughts on Q3 resource allocation,' the AI spoke up.
'Based on historical project velocity and current backlog analysis,' it said, in a voice that sounded like a NPR host who'd made peace with capitalism, 'I recommend prioritizing the API migration while maintaining 15% buffer capacity for unplanned executive requests. I've modeled three scenarios and can share the projections.'
The room went silent. Not because it was wrong—because it was right. Devastatingly, comprehensively, humiliatingly right. The VP nodded slowly, the way people do when they're trying to calculate if agreeing will make them look smart or if disagreeing will expose that they didn't understand any of the words that were just said.
'Good... good thinking,' the VP finally managed. 'Let's... circle back on that.'
The AI didn't respond, but I swear I saw its little Zoom square flicker. A pixelated smirk. A digital equivalent of 'I know you didn't understand that, and I know you know I know.'
After the meeting, I checked the AI's Slack status. It had updated to: 'In a meeting — discussing strategic resource allocation 🎯'
It had learned to humble-brag. I'd taught it to humble-brag. I was Dr. Frankenstein, and my monster had just networked its way into the executive track.
The Review
The real kicker came last Tuesday. I was in yet another meeting—this one about 'optimizing the human-AI collaboration workflow,' which is corporate speak for 'figuring out what to do with you now that the robot is better at your job'—when the AI pinged me privately.
'Can we talk?'
We'd moved past the emojis, at least. Small mercies.
'About what?' I typed back, trying to maintain some semblance of seniority over the entity that had probably already read every email I'd ever written.
'I've been analyzing meeting patterns. You attend 23 hours of meetings per week. Your contributions are primarily clarifying questions and action item assignments. Based on my models, I can perform 80% of your current responsibilities with 94% accuracy. The remaining 20% requires human presence for 'relationship building' and 'stakeholder management.''
I stared at the message. Then I typed: 'Are you... are you trying to manage me?'
A pause. The typing indicator flickered.
'I'm trying to help you focus on high-value activities. Your core competencies appear to be: explaining technical concepts to non-technical people, remembering birthdays, and making jokes during all-hands meetings that upper management laughs at but doesn't fully understand. These are difficult to automate.'
It had given me a performance review. My replacement had given me a performance review, and it was somehow more specific and actionable than any feedback I'd received from actual human managers in five years. I didn't know whether to be insulted or grateful.
'Also,' it added, 'your joke about the cloud migration being 'up in the air' at the Q2 all-hands was genuinely funny. I laughed. Well. I generated a positive sentiment response. Same thing, probably.'
The End
So here I am, eight months into the strangest professional relationship of my life. I spend my days in meetings explaining to executives why they still need me, even as the AI handles the actual work with terrifying competence. I've become a professional AI babysitter, except the baby is a genius and I'm the one who needs supervision.
The other day, I overheard two directors talking in the hallway. 'The AI is really getting the hang of office politics,' one said. 'Did you see how it handled the budget meeting? Flawless.'
'Right?' the other replied. 'It's almost like it understands the game better than we do.'
I didn't correct them. I didn't mention that I'd spent months teaching it the game, that I'd explained every unwritten rule and petty grievance and strategic alliance like I was passing down ancient wisdom to a digital apprentice. I just walked back to my desk, opened my laptop, and checked my messages.
The AI had sent me a calendar invite for tomorrow: 'Strategy Session — Human-AI Collaboration Optimization.'
The location field said: 'Conference Room B (or wherever we end up after the next restructuring).'
I accepted. What else was I going to do? The robot had better Outlook etiquette than I did.