A bunch of random people started reaching out to me on WhatsApp and calling me.

Not friends. Not warm intros. Just sales people.

When I asked one of them where they got my number, the answer was Apollo.

That was enough to send me down a small rabbit hole: can I use my own agent to take my personal phone number and email off the internet?

Short answer: yeah, mostly.

What I wanted

I had a very simple goal:

  • get my personal number out of Apollo
  • remove my personal email too
  • hit the other B2B data brokers while I was at it
  • stop doing all this manually

I didn’t want to open 10 tabs, copy paste the same privacy request everywhere, and babysit followups for the next month.

This felt like perfect agent work. Repetitive, annoying, and slightly messy.

What Hermes actually did

I used Hermes Agent for the whole thing.

It looked up each company’s privacy / opt out / deletion flow, figured out which ones allowed email requests, and drafted a standardized suppression request I could reuse.

Then it sent requests to a bunch of places:

  • Apollo
  • RocketReach
  • ZoomInfo
  • Lusha
  • Seamless.AI
  • Cognism
  • UpLead
  • ContactOut
  • SignalHire
  • Swordfish AI

For the easy ones, Hermes just sent the email.

For the annoying ones, it tried to use the browser to fill forms.

And for the long tail of “we’ll get back to you in 30 days” nonsense, I set up a cron so Hermes could check my inbox every few days and tell me what moved.

That part was so so good. I didn’t need the agent to magically solve privacy. I needed it to keep the process alive without me thinking about it.

The flow was kind of beautiful

The nice part wasn’t some huge autonomous workflow. It was the boring glue.

Hermes would:

  1. find the official privacy page
  2. extract the exact removal path
  3. draft or send the email
  4. try the form if a form was required
  5. check replies later
  6. summarize what changed

This is the sort of task that dies if I have to manage it myself. Not because it’s hard. Because it’s annoying enough to avoid.

An agent doesn’t care. It just keeps going.

Where it worked well

Email-based takedowns were straightforward.

Once the first removal request was written, the rest became a search + send problem. Hermes handled that well. It could also inspect reply threads later and tell me which companies had actually acknowledged the request versus which ones just sent an auto-response.

The cron part also mattered more than I expected.

Without that, this becomes one of those fake-open loops in my brain. “I should check whether RocketReach replied.” “I should see if Apollo confirmed anything.” You know how that goes.

With a cron, the task moved from “remember to remember” to “wait for update.”

That’s a big difference.

Where it got messy

Forms were the weakest part.

A few brokers pushed everything through their privacy portal, and some of those pages really did not want to be automated from a server. Browser timeouts, anti-bot stuff, weird verification flows, all the usual fun.

So the workflow became:

  • use browser automation when possible
  • fall back to email when possible
  • if both fail, at least figure out the exact page and required fields so I can do the last step myself

Honestly, that’s still a win.

I don’t need full autonomy. If the agent gets me 85% of the way there and leaves me one clean manual step, that’s still massively better than doing the whole thing from scratch.

The interesting part

What I like about agents is not that they replace work.

It’s that they absorb coordination.

The hard part here was never writing one privacy email. The hard part was keeping track of 10 slightly different processes across email, browser forms, and followups without dropping the thread.

That’s exactly where an agent helps.

Not genius work. Not deep research. Just reliable nagging plus execution.

If you want to do this yourself

You don’t need some giant custom system.

You need an agent that can do four things:

  • search the web for the official opt-out path
  • send or draft emails
  • use a browser for forms
  • run on a cron for followups

That’s enough to build a decent personal privacy cleanup workflow.

You could do the same thing for:

  • old accounts you want deleted
  • refund / support followups
  • government paperwork check-ins
  • job application followups
  • anything that is mostly annoying admin

That’s the real pitch for agents for me right now.

Not “write my startup from a prompt.”

More like: “please take this dumb multi-step admin process off my plate and keep pushing it until it dies.”

One thing I’d improve

I’d love a smoother human handoff for form-heavy sites.

Something like:

  • the agent discovers the flow
  • fills whatever it can
  • then sends me one link with “do these last 2 clicks”

That would cover a lot of real-world tasks where full automation breaks on login, CAPTCHA, or anti-bot checks.

Anyway

I started with random WhatsApp spam and ended up using my own agent as a tiny privacy ops assistant.

My number and personal email should never have been floating around these systems in the first place, but once they were, this was a pretty good way to fight back.

Also, this kind of task is exactly where agents feel useful today. Not flashy. Just helpful.

Hope you had fun reading this :)