Jai T.

Nikhat I. it's both! Identification when setting it up then deploying it as a Prioritization workflow. 1st is useless without the second

Clay may also be retaining companies with zero matched employees. Ex. People data labs has 76m companies, 22m have at least 1 employee in their data set. Clay had 44m last I checked, so my guess is both subsidiaries/m&a + companies with no employees

Main thing that makes it good vs bad is really good context on messaging framework and evaluation of that content - no shortcut vs doing a bunch of examples and fixing it, tracking the change and then updating skill to match the changes you made. Takes 30-40 examples in my experience (generate 5, review + edit, generate 5 and so on)

Yeah fully automated at this point - Claude code. Skills are the correct route but 90% of it is setting up MCPs and APIs and making sure it all routes correctly. Cowork might be able to do it, doubtful you could get it to work via web. Needs local files and clis make it 100x easier

Allie Harrison highly recommend throwing Claude code at the problem (not Claude, not cowork). Give it a list of wins and losses and dates and it'll go find and build a mini explainable ML model. Can share a skill or talk from a customer about it if that's helpful!

I'd look into Claude code + the api, then having it create a script to do what you wants (test v small, host in n8n/trigger)

We just did a bunch of direct api integrations and that's working well for our volume/icps/business model.

Travis R. both will likely work! Have seen more people use fullenrich but better quality on bettercontact for us b2b saas. Any concerns about committing to one provider for everything? Feels like “best” changes every few months.

Very cool Ugo C.! How are the marketplace products priced? Product guy in me is curious how this happens in the first place!

Are you using the hubspot native calendar booking links? Think other platforms/integrations (ex. Calendar) add a property.