Show me
Original data

How long personalized outreach actually takes

Producing personalized outreach takes about two hours per contact, even with AI assistance. For 250 contacts that is roughly 500 hours, or three months of one person working full time. The AI is not the bottleneck. The bottleneck is that the work is serial and has to be repeated for every individual.

The breakdown

This is what it takes one person to produce a full set of outreach for a single contact, with AI assistance throughout and templates and hosting already set up. It is the optimistic version.

Research and verify the person: profile, recent posts, company site, contact details
10 min
Build the brief and decide the angle for this specific person
5 min
Five emails with five subject lines, each making a different argument, each reviewed
30 min
Four LinkedIn messages, all different, none echoing the emails
15 min
A personalized web page: content, assembly, unique URL, publish, verify
30 min
A personalized video: script, render, upload, verify
20 min
Final quality check across every asset
10 min
Total
~2 hrs

Bar widths show each task's share of the ~2-hour total.

*These times come from producing this work one contact at a time across Leeluu's own pipeline of 3,500+ B2B contacts, before the process was automated. They are our own estimates from doing the work, not an industry benchmark.

What that means at volume

Producing personalized outreach for 250 contacts, one contact at a time: 2 hours per contact, 500 hours total, 3 months of one person full time, $20,000 to $25,000 in labor.

The part the numbers do not capture

Quality does not hold across 250 repetitions. In practice, by around the fortieth contact people begin reusing prompts and phrasing. By the eightieth the drift is visible: the messages start sounding like each other rather than like the individuals they are addressed to.

Which is why, in most organizations, this work does not happen at all. Teams default to a segment-level template because the individual-level version is not achievable one prompt at a time.

Why AI has not solved this

The writing was never the slow part. The slow part is everything around it: opening each profile, reading the recent posts, feeding the right context into the model, prompting separately for each asset, checking the output against what is actually true, assembling it, hosting it, and verifying it works.

That work is serial. It does not compress by adding a better model.

Leeluu was built to remove the serial part.

If you have a list and no 500 hours, see how it works.