This is an excellent and extremely practical interview.

Am I correct that this podcast in not on Apple podcasts? It would be convenient to listen to it there.

Expand full comment

I really liked this interview.

I'm in the middle of implementing GenAI in my company and access to internal data sources is the first thing on the map - like most other companies, our information is produced in troves (and poor quality) and then very difficult to operate (we use Atlassian tools, SFDC, SharePoint, etc. ). Like new sales people trying to happen what happened on an account or project managers willing to get a quick update on a current project.

But very few talk about that and ChatGPT itself will not get you there (not to mention the enterprise user management issues, which are also rarely discussed). Maybe a POC with PDFs loaded in a customGPT; but the next stage is to connect live data, and it becomes way more difficult!

I just wanted to share a couple of things:

- OpenAI is working on this and promised us (we are en enterprise customer) live data connectors in the near future. We'll see.

- We found an open source product, Danswer, that seems to do the job quite well. We are prototyping it.

- a French product, Dust (dust.tt) offers the same type of features.

- I think that there is a bunch of other companies specialised on just that. I have found this one https://www.searchblox.com/products/ and this one https://www.glean.com/solutions/all-teams.

However Cohere seems really interesting and you cannot really discard a company founded by a guy who co-wrote a seminal paper at 20 :-)

All in all it's good to focus on the real problems mid-sized companies willing to deploy GenAi internally actually have.

Thanks for that Azeem.

Expand full comment