How long lead generation takes, with and without AI
Building a lead list by hand is estimated at about eight hours. With AI it is about 26 minutes. Those figures come from a free public dataset, along with a more useful finding: which parts of the job people hand over completely, and which parts they refuse to.
Nobody has ever timed you building a prospect list, so here is the closest thing anyone has published. Anthropic releases a free dataset that estimates how long real working tasks take with and without AI, and four of the tasks it covers are the job you do every time you go looking for customers. Building a lead list comes out at about eight hours by hand and about 26 minutes with AI.
The more useful finding is underneath that. The same data shows which parts of the job people hand over completely, and which parts they will not let go of.
The short answer
Four jobs make up most of UK B2B prospecting. Here is the estimated time each takes alone, and with AI:
- Building the list: about 7.9 hours by hand, about 26 minutes with AI
- Qualifying what you found: about 4.7 hours, about 28 minutes
- Writing the outreach: about 5.7 hours, about 51 minutes
- Sales and revenue admin as a whole: about 5.75 hours, about 40 minutes
Add the first three together. A round of prospecting that ran to the better part of two working days comes back at about an hour and three quarters.
Where the numbers come from
They come from the Anthropic Economic Index, which classifies millions of real conversations against the US Labor Department's O*NET task list and estimates the time each task would take a person working alone. It is published free under a CC-BY licence. The figures here are from the release covering May 2026.
Two honest caveats before you quote them at anyone. The times are estimates drawn from classifying conversations, not stopwatch readings taken over someone's shoulder. And they describe people using a general assistant, so a tool built for one job can beat them, or fall short of them.
One more number matters for how you read all this: between 89% and 95% of these tasks were judged to be things a person could do perfectly well without AI. That is the point. This is an argument about hours, not about capability.
Describe the customer once. We check the record, the website and the email, and hand you the list with the reason on every row. 25 free to start.
The part people hand over completely
List building is the most fully delegated sales task in the dataset. Seven in ten of those conversations are handed over rather than worked through, and 58% are a single instruction followed by no further steering, which is the highest score of any sales task measured.
That is a real finding about how people feel, not just what they do. Nobody wants a conversation about building a list. They want to describe their customer once and get the names back.
Qualifying leads behaves the same way, handed over in 71% of conversations. Deciding which of 400 companies are worth an email is work people are glad to be rid of.
The part people refuse to hand over
Now the opposite. Marketing copy is worked through collaboratively in 68% of conversations, and handed over outright in only 25%.
Same people, often the same session, opposite behaviour. The finding is not that writing is harder than searching. It is that the list stays private and the email does not. One of them arrives in a stranger's inbox with your name on it.
If you take one thing from this post, take that split. Hand over the finding and the filtering. Keep your hands on what gets said.
What a finished job looks like
The dataset also records what the work produces, which is a good check on whether a tool is actually finishing the job:
- List building ends in a spreadsheet or data file 42% of the time
- Cold outreach ends in an email or message 53% of the time
- Qualification splits between a report, a written analysis, an email and a spreadsheet
None of those is a chat window. If you finish a prospecting session with advice on screen and nothing you can open tomorrow, the job is not done. A file is the deliverable.
This is arriving whether or not you adopt it
There is a second half to the dataset covering businesses that build AI into their own products, and it reads very differently from the chat side. Lead qualification is 1.95% of everything flowing through Anthropic's business API, up from 1.61% a month earlier. It is the single largest sales use.
Lead list building, sales intelligence and pulling data out of a CRM are all several times more common through the API than in chat. The gap is the tell: this is being built into products right now rather than typed by hand.
For a UK business, that has a practical edge to it. The competitor who replies to a new opportunity within the hour is increasingly one who has stopped doing the finding and the sorting by hand.
What to do with this
- Time your own version once. One list, honestly timed, including the tab-hopping. Most people are surprised by their own number.
- Hand over the finding and the filtering. This is where the eight hours actually sit, and where the data says people are most comfortable letting go.
- Keep the writing. Draft with help if you like, but read every line before it leaves.
- Judge tools on hours, not features. The question is how long your next 200 checked names take, and how much of that time is yours.
How Leadistry splits the same way
We built along the same line the data draws. Finding and filtering runs on its own; the words stay yours.
You describe the customer once, and the search runs against the live Companies House record with your trade, size and area applied. Set it to repeat and it keeps going without you. Every lead arrives checked, with a verified website and email, and you can see the audience before you spend a credit. Where a lead carries a reason to get in touch, a filing or a change on the public record, the reason arrives on the row.
Then it stops and hands over. Outreach is drafted for you and sent when you approve it, not before. That is not caution for its own sake, it is the 68% figure above, and we think the people in that number are right.
If you want the accuracy side of the same story, our accuracy page sets out what we check and what we will not claim. For the workflow end to end, see the UK B2B prospecting workflow on one platform, and for what the whole thing looks like as a job description rather than a feature list, meet your AI lead gen employee. If you would rather buy a list than build one, we compared the three routes in B2B email lists: buy, build or generate.
Common questions
How long does it take to build a B2B lead list?
About 7.9 hours working alone, and about 26 minutes with AI, on the estimates published in the Anthropic Economic Index. That is for the finding and listing itself, before anyone writes to a single name on it.
How long does it take to qualify a lead?
The same dataset estimates about 4.7 hours by hand against about 28 minutes with AI for lead qualification, which covers deciding which of the names you found are worth contacting.
Can AI do lead generation on its own?
It can do the finding and the sorting on its own, and the usage data shows that is exactly how people treat it: list building is handed over in one instruction more than half the time. The writing is different. People keep hold of that, and so should you.
Which parts of lead generation should you not automate?
The words. Marketing copy is worked through collaboratively in 68% of conversations, against 25% handed over outright, while list building is the opposite way round. The list does not carry your reputation to a stranger's inbox. The email does.
Where do these time figures come from?
The Anthropic Economic Index, a free dataset published under a CC-BY licence that maps millions of real Claude conversations to the US Labor Department's O*NET task list and estimates how long each task takes with and without AI. The figures here are from the release covering May 2026.
Are the figures measured or estimated?
Estimated. They come from classifying real conversations, not from timing people with a stopwatch, and they describe people using a general assistant rather than a specialist tool. Read them as the shape of the saving rather than a promise about your own week.