Using AI to generate leads

[szalek]

Have any of you used openai bots to search the internet. I don't mean usually sending a link to ai to tell me something about a particular site. Just about searching the internet with the purpose of getting information about potential customers. I used to search for companies manually. I searched based on the information I needed (keywords) and collected data about the company plus contact details which I entered into excel. I could then use this for activities that were designed to turn the collected data into customers. However, this was very time-consuming.

Was it possible to configure gpt in the paid version for this?
Have any of you used or are using such a solution?

 

Dislike ads? Remove them and support the forum: Subscribe to Fastlane Insiders.

[sdbrownlie]

Have any of you used openai bots to search the internet. I don't mean usually sending a link to ai to tell me something about a particular site. Just about searching the internet with the purpose of getting information about potential customers. I used to search for companies manually. I searched based on the information I needed (keywords) and collected data about the company plus contact details which I entered into excel. I could then use this for activities that were designed to turn the collected data into customers. However, this was very time-consuming.

Was it possible to configure gpt in the paid version for this?
Have any of you used or are using such a solution?

Yeah it's more of a using the AI within a script than using a GPT though if you're doing it at scale - GPTs can only perform actions for a 'little while' so you'd have to sit there typing 'continue' over and over as it found like 10 sites at a time or something.

You could probably make it do it quite well if you just are doing this for custom outreach though with just a GPT and you are working with small numbers. You can give a GPT api access to hunter as well as giving it guidance which pages to pick to find contact/owner/about pages etc to research on etc etc. But if you are doing something massive scale for email outreach or something you'll want to code something up yourself that can handle big batches.

Some tips if you do code something up for yourself:

* hunter.io still has better emails for some businesses than they have listed on their websites so paying for that and hooking up the API is a good idea
* tools like modal (they give $30 free credit - enough to scrape tens of thousands of sites) are handy for speed - it'll run your script on 1,000 'machines' at once so scraping 10,000 sites takes a couple minutes instead of hours on some cheap vps running 8 threads or something.

 

[StrikingViper69]

Lol AI.

You can do this with a simple script to scrape data. People have been doing it for years.

 

[sdbrownlie]

Lol AI.

You can do this with a simple script to scrape data. People have been doing it for years.

True I've had a script since 2010 when I started my SEO business but AI has added lots of functionality (personalizing outreach for me etc) that wasn't possible before. It also keeps coming up with new angles to search for etc etc which sure humans can do and keyword tools can do but it's increased the yield from scrapes massively too because it comes up with stuff that my team weren't and saves them hours of time.

It's also able to do things like prepare for calls for you - one of the things we do for new linkbuilding clients is look at all their existing content, the type of site, and assess a whole bunch of stuff for 'suitability' so we know going into the call how hard the site is going to be to work with etc. The AI can do all of that based on our rules/categories/systems which saves half an hour for every booked appointment we have.

Lots of ways for AI to enhance our old processes - even if our old processes were great and are working just fine. Why not save some time, make things more efficient. It's not like these APIs cost much money at all. I'll happily bin $300/month to OpenAI to save 40-50 hours of productive hours for me and my team to do something else with.

I probably haven't even scratched the surface on everything it can do to improve our old process. We're working on having it reply to classify emails we get back in and reply to all the simple to handle ones - that's probably a time saving of 8+ hours/day for our team etc which would be huge compared to the smaller savings listed above.

 

Dislike ads? Remove them and support the forum: Subscribe to Fastlane Insiders.

[szalek]

Yeah it's more of a using the AI within a script than using a GPT though if you're doing it at scale - GPTs can only perform actions for a 'little while' so you'd have to sit there typing 'continue' over and over as it found like 10 sites at a time or something.

You could probably make it do it quite well if you just are doing this for custom outreach though with just a GPT and you are working with small numbers. You can give a GPT api access to hunter as well as giving it guidance which pages to pick to find contact/owner/about pages etc to research on etc etc. But if you are doing something massive scale for email outreach or something you'll want to code something up yourself that can handle big batches.

Some tips if you do code something up for yourself:

* hunter.io still has better emails for some businesses than they have listed on their websites so paying for that and hooking up the API is a good idea
* tools like modal (they give $30 free credit - enough to scrape tens of thousands of sites) are handy for speed - it'll run your script on 1,000 'machines' at once so scraping 10,000 sites takes a couple minutes instead of hours on some cheap vps running 8 threads or something.

what I am trying to achieve with ai is to search the most popular job portal in my country "pracuj.pl". After entering a key phrase, I get more than 1000 results of which 20 are interesting. Searching such a portal takes more than 2 hours. I once had a programme that searched this portal but it was removed. This programme actually selected 20 interesting offers out of 1000 and generated results in a csv file. For this kind of problem, I was thinking of using AI

 

[sdbrownlie]

what I am trying to achieve with ai is to search the most popular job portal in my country "pracuj.pl". After entering a key phrase, I get more than 1000 results of which 20 are interesting. Searching such a portal takes more than 2 hours. I once had a programme that searched this portal but it was removed. This programme actually selected 20 interesting offers out of 1000 and generated results in a csv file. For this kind of problem, I was thinking of using AI

While you could use an LLM / AI for that I think it would be a bit like firing a nuke to sink a small fishing boat unless the jobs are very difficult to categorize/tell apart using more basic NLP techniques or libraries.

I used the most basic set up possible on an example UK jobs site just to keep my post in English on an English forum!

Interim Financial Accountant

Grant Thornton is recruiting for this Interim Financial Accountant vacancy on Reed.co.uk, the UK's #1 job site.

www.reed.co.uk

And it is able to extract this for each job:

Job Title: Interim Financial Accountant London - Reed.co.
Salary: £400.00 - £500.00 per day
Keywords Found: financial: 16 accounting: 14 reporting: 11 public: 9 sector: 9 work: 8 team: 7 community: 6 clients: 6 support: 6

If that kind of information was enough for each job for you to filter to your '20 most interesting' then a fairly standard web scraping set up would do the job for you without needing to bust out the AI!

 

[szalek]

While you could use an LLM / AI for that I think it would be a bit like firing a nuke to sink a small fishing boat unless the jobs are very difficult to categorize/tell apart using more basic NLP techniques or libraries.

I used the most basic set up possible on an example UK jobs site just to keep my post in English on an English forum!

Interim Financial Accountant

Grant Thornton is recruiting for this Interim Financial Accountant vacancy on Reed.co.uk, the UK's #1 job site.

www.reed.co.uk

And it is able to extract this for each job:

Job Title: Interim Financial Accountant London - Reed.co.
Salary: £400.00 - £500.00 per day
Keywords Found: financial: 16 accounting: 14 reporting: 11 public: 9 sector: 9 work: 8 team: 7 community: 6 clients: 6 support: 6

If that kind of information was enough for each job for you to filter to your '20 most interesting' then a fairly standard web scraping set up would do the job for you without needing to bust out the AI!

Do you do web scraping manually or do you have a script or code for this?

 

Dislike ads? Remove them and support the forum: Subscribe to Fastlane Insiders.

[sdbrownlie]

Do you do web scraping manually or do you have a script or code for this?

Script/code. I have a few for different purposes. Usually able customise one of those up for the job in hand - that's why I was able to do a quick look at what could be grabbed off a job site without much time spent. Improvements ofc if it was a real project would be figuring out what speed they'd allow before blocking/adding proxies if needed (ie it's so slow to make it useless) and improving the data extraction (eg using a better nlp library to grab the keywords, looking at other structured data on the page if any as well as just the salary range and so on).