For a few weeks this year I was going to build a LinkedIn automation tool. Scrape a list of people who fit an ideal customer profile, write a connection request for each of them with an AI model, send it, and follow up on a schedule. The pitch was account safety. Everyone who uses these tools has a story about a restricted account, so the product would be the careful one: slow, human-paced, hard to detect.
I had a spec. I had a benchmark of the tools already selling this. I had a rough idea of pricing. And then I stopped, and I think the decision to stop is worth more than the tool would have been. This is the reasoning, written down while it is still fresh.
Why it was attractive
It is worth being honest about the pull, because it is real.
The demand is obvious. Every founder, every recruiter and every agency wants more conversations with the right people, and LinkedIn is where those people are. The category exists, the leaders in it are charging a subscription anyone can put on a company card, and people complain about the leaders constantly. “Safer than Dripify” is a positioning line that writes itself.
And the build was within reach. Scraping, an LLM for the copy, a queue with jitter on it, a browser that behaves like a person. None of it is hard for someone who does automation for a living. That is exactly what made it dangerous: the thing I am good at was pulling me toward a product I had not thought about properly.
The reason that mattered most
The whole product was a promise I could not keep.
Account safety was the value proposition. But the party that decides whether an account is safe is LinkedIn, and LinkedIn’s terms of service say, in plain words, that you may not use bots or automation on the platform. So the pitch is really this: pay me monthly, and I will help you break a rule, carefully, on a platform that can change how it enforces that rule any Tuesday it likes, without telling either of us.
That is not a product. That is insurance, sold by someone who does not control the risk. The first time LinkedIn tightens detection, every customer I have gets restricted in the same week, and every one of them was told by me that they would not be. The thing I was selling would be the thing that failed.
The other reasons, in order
Once I had seen the first one, the rest lined up behind it.
Churn is built into the product. A customer whose account gets restricted does not renew, and they do not blame LinkedIn. They blame the tool. Even a well-behaved tool loses its customers at whatever rate LinkedIn’s enforcement sets, and that rate is not published.
The support load is the wrong shape for one person. A restricted account is an emergency for the person it happens to. As a solo builder I would be the entire support desk for a stream of emergencies I could not fix, because the fix lives on LinkedIn’s side.
The legal picture is unsettled, and that is enough. I am not a lawyer, and I do not need to be one to notice that the people arguing about the legality of scraping LinkedIn are companies with legal departments. A one-person business in Devon does not want to be a test case, and it does not want a business model that depends on how the next case goes.
The market is already owned. Dripify, Expandi and a long tail behind them. My differentiator was “safer”, which is a claim a customer cannot verify until the day it turns out to be false. That is a weak position against incumbents with years of data on what gets accounts flagged.
The risk was asymmetric in the wrong direction. Best case, a modest subscription business in a crowded category. Worst case, a platform decision, a legal letter or a wave of restrictions that ends the business and costs my reputation with it. When the downside is that much bigger than the upside, the expected value is not the point.
The question I now ask first
All of this reduces to one question, and I wish I had asked it on day one:
Is the platform I am building on happy that I exist?
For LinkedIn automation the answer is no, and every other problem followed from that. It is also, I have realised, the same rule I already apply to client work. Every automation I ship for a business is built on APIs the vendor publishes and documents: Xero’s, HubSpot’s, Shopify’s, Google’s. Not because I am timid, but because a workflow that depends on a vendor not noticing is a workflow that is going to break at the worst possible moment, and I would rather not have that conversation with a client. I had somehow forgotten to apply the rule to my own product.
What I am building instead
The demand is still real, so the question became: where do people have business conversations that the platform actually wants automated?
The answer I landed on is Telegram and WhatsApp. Telegram’s Bot API and the WhatsApp Business Platform are both officially permissioned. The documentation is public, the rate limits are published, the messaging rules are explicit, and the platform’s position is not “please don’t”, it is “here is how”. You build in the open, and the thing you build does not evaporate when someone at the platform has a bad quarter.
So the spec I am writing now is for a B2B relationship CRM that lives natively on those two channels: contacts, context, follow-ups and nurture sequences, run through the official APIs, for businesses whose customers already talk to them there. It is at the spec stage, not the demo stage, and I will write it up properly when there is something to show rather than something to describe.
If you are weighing a similar build
Three things I would say to the version of me from a few months ago.
- Write down who controls the risk in your value proposition. If it is not you, you are selling insurance, and you are not an insurer.
- Benchmark the incumbents on the axis you cannot see. I compared features and pricing. I should have compared how they handle the week their customers all get restricted, because that is the business.
- Ask whether the platform wants you there. If the answer is no, the size of the market is not a reason to proceed. It is the reason the platform will eventually act.
None of this makes LinkedIn automation a bad business for the companies already in it. It makes it a bad business for me, right now, on my own. Knowing the difference is the whole job.