Hey ​@kenzofong,

This is interesting. Thanks for sharing. Is Omni able to provide the data source as well?

Mike, Consultant @ Automatic Nation 
YouTube Channel 

Yes, you’ll always see both the entity name and the source website/URL for each record.

That sounds great, thanks ​@kenzofong!

Mike, Consultant @ Automatic Nation 
YouTube Channel 

Hello. I love this feature and have tried to use it to do some research on a very niche industry. My prompt is very simple and asks for company name, email addresses of decision makers, and mailing address for volumetric concrete mixing operations in the US and Canada. Omni has been “finding data” for nearly 20 minutes now. It’s unclear if this is a normal amount of time for a query like this, if I am just not being patient enough, or if the query itself is simply too big of an ask. I asked ChatGPT to do deep research on this same topic and it gave me a list (I need to vet it, so TBD on accuracy) within 8 minutes. I’d like to compare the two outputs and would like to know how much “thinking time” from Omni we should allocate for queries like this. Thanks!

Thanks for the thoughtful question! The amount of time Omni takes really depends on two factors: 1) how complex the request is and 2) how large the universe of possible entities might be. A broad query like “all volumetric concrete mixing operations in the US and Canada with decision-maker emails and mailing addresses” is quite intensive, since it requires scanning a wide surface of the web and then enriching with multiple data fields. That can take considerably longer than a simpler lookup.

To get a sense of quality and turnaround, I’d recommend narrowing the scope first. For example, try:

“Find all company names, email addresses of decision makers, and mailing addresses for volumetric concrete mixing operations in California.”

Smaller, more targeted regions typically return faster, and you can then compare Omni’s results side by side with other providers before scaling up to larger geographies.

So you said it’s free to use for now right? but any thoughts on what pricing might be because I use other ai tools to find leads

So one trouble spot I’m running into is basically trying to do a web table based off a pre-existing data set. 

I’ve got a large (large) dataset of film titles a local org has available for rent, but record wise all they have is title and MPAA rating. I’d like to use a single AI action to scrape IMDB for the title and populate film poster, description, and length. This is similar to the web table function (a single query/action populating multiple fields) but it seems like Omni/agent fields can only populate a single field, while the web table is only for new data. Am I totally missing something?

We’re still figuring out pricing for features like this. Curious, how do you use AI tools today to find leads, and which ones are worth paying for in your workflow? Are there any features that would make you switch or stick with one tool over another?

Depending on how many films you have in your dataset, you can do this using web tables by adding the list of titles to your prompt and then making sure the tool is selected. For example:

“Find all descriptions, posters and runtimes for these movies: Phase One • Iron Man (2008) • The Incredible Hulk (2008) • Iron Man 2 (2010) • Thor (2011) • Captain America: The First Avenger (2011) • The Avengers (2012) Phase Two • Iron Man 3 (2013) • Thor: The Dark World (2013) • Captain America: The Winter Soldier (2014) • Guardians of the Galaxy (2014) • Avengers: Age of Ultron (2015) • Ant-Man (2015)”

This returns a web table with the information you’re looking for like below:

Haha, oh no, I mean large. 22K odd records in the database.

Definitely aware this would burn through monthly AI credit limits, but I’m wanting to figure out a way to run batches. Even running multiple agents they don’t seem to chain together correctly. For instance I added an agent to populate a URL of the title’s IMDB page, and then another directly instructed to use that URL for the poster, but I find instances where the URL is correct but the poster is an entirely different movie. 

So I use EXA via API and so far I am happy with it I then export it to my google sheet so I can do all sorts of automations I want to do but recently I am not really happy with there results to looking for some better options 

Thanks for sharing. Would love to chat more, will send you a DM for follow up if that’s okay.

Yes, that’s a lot of records that would overload the prompt input. Makes sense as a use case and we’ll look into this. In the meantime, you should be able to at least find poster URLs with the new web image search in Field Agents.

Cool! I see this as the other side of the web table concept. It’s great to get an initial dataset, but after some time you want to refresh that data set rather than make a whole new one.  

I can, but there are two downsides I see here.

• I would presume it uses more credits (or at least more compute) for an agent field to scrape the web for the imdb url, and then for another agent to scrape the web for the poster, instead of being able to be pointed directly where it needs to go from the initial scrape.
• It introduces a lot more variability and potential for incorrectness or hallucination to run these agents in parallel rather than in sequence. 

Anywho, just feedback and thoughts 

Depends on the complexity of the query. With web tables you can do very complex queries that do not have a straightforward answer/source on the web which is more computationally complex and as a result can take more time/credits.

I would probably still do this sequently (i.e. build out a base table with web table before augmenting with field agents). Finding a specific URL from IMDB once you have some movie meta data is a way less open ended and as a result faster, less error prone and credit wise should also be reasonable.