Using AI to personalise outreach to a list of 200 companies

You have a list of 200 companies and a tool that can write 200 personal opening lines before lunch. The temptation is to let it. Resist it for most of them. You can personalise to 200, but not equally. Spend real reading on the best few, a lighter touch on the middle, and an honest plain message on the rest. One rule holds throughout: a personal line must come from something you have read. Where that would need a guess, stop and send the plain version.

Why not treat all 200 the same way?

Time, not the tool, is the limit. An assistant makes 200 lines in minutes. Checking 200 lines takes you far longer, and unchecked lines are where the damage happens: the wrong plant, the wrong product, a compliment on something another company did.

Tiers also make your results readable. When one tier replies more, you learn something. When all 200 got the same message, you only learn that the list replied or did not.

What does the working sheet look like?

One row per company, with these columns:

Company | City | Tier A/B/C | Decision-maker (name, role) |
Source link | The line I wrote | Checked by me (Y/N) | Sent on | Reply

"Source link" is the column that keeps you honest. If a line has no link in that cell, it is deleted before it is sent.

How do I sort the tiers?

Use two questions: does this company buy what you sell, and do you know who decides? Both yes puts it in tier A, which for illustration might be twenty companies. One yes puts it in B, perhaps sixty. The rest are C. The sizes are yours to choose. What matters is that effort falls as fit falls. Our post on building a qualified B2B prospect list goes into sorting.

Tier A. Open the company's own site. Paste its About and Products text into the assistant and ask for a plain list of what it makes. Choose one observation yourself and write the line. Ask the tool to tighten, nothing more. The closing question is yours.

Tier B. Paste one page of the company's site and ask for a draft with a single line about its work. Read each draft beside the page. Delete whatever goes beyond the page.

Tier C. Use a template with the company name and trade. Do not ask a tool to invent a personal line. A message that says "I am writing to every Udaipur hotel group with more than one property" is honest about why the person is getting it.

What went wrong in Udaipur?

Imagine, as an invented example, a reseller of payroll and attendance software in Udaipur who wants to reach 200 hotels and resorts across Rajasthan. The sales head had an assistant write personal openings from company names alone.

One draft praised a hotel for its "new spa wing". The property had no spa. Another congratulated a group on an expansion into Goa that had been announced by a different chain with a similar name. A reply arrived from a general manager: "We have no spa. Who told you that?" By then the sales head had sent forty such lines.

The sales head stopped and rebuilt the batch in tiers. For the twenty hotel groups that mattered most, the sales head read each website and wrote the line. One read: "Your site lists three properties and a staff careers page. How do you track attendance across them?" For sixty, the sales head pasted one page of each site and checked the draft against it. For the remaining hundred and twenty, a short template was used naming the trade and the region. A short apology went to every manager who had received a wrong line, with thanks to the one who replied.

Replies came from all three tiers. Tier C also brought two "please remove us" answers, which the sales head honoured at once. The replies that led to calls were mostly from the first tier, but the numbers were small.

What does a checked line look like?

Short, true and about their work. Never about their size or success.

"Your site lists three properties and a careers page." Good: it names something visible on their own page.

"I was impressed by your growth." Poor: nobody can check it, and it says nothing.

Anything about a plant, a client, a date or a facility that you have not read. Wrong, and the kind that costs you the account.

When should you stop personalising?

At the first line that needs a guess. When the company has little public text, since the tool will fill the gap. And when you notice you are polishing the line more than thinking about the buyer's problem.

Some lists do not suit this at all. If your buyers are individual shop owners who do not read email, a visit or a call will beat any line. If the trade is regulated, take claims about what a company is licensed to do only from its own pages.

Keep personal details out of the tool. Paste public company text, never a contact list with phone numbers. Rules on unsolicited email and messages differ by channel, so ask your own adviser before you send. Our AI use policy post gives the limits.

What if I get a line wrong?

Someone will reply to say so. Thank the person, correct the sheet, and look for the same mistake elsewhere in the batch. One wrong line found early is cheap. The same line found after forty sends is not.

Send tier A first

Send the best twenty by hand, so you can adjust before the rest. Record every reply by tier in the sheet. If tier A and tier C reply about equally, the personal effort is not paying, and the time can go elsewhere.

Write the follow-up for each tier before the first send, so the second message does not have to be improvised. AI prospect research covers the reading behind tier A, outreach is one part of the work under marketing, and to plan your own list, tell us what you need.