Here’s Why your RAG sucks: ⬇️ · Brendan's AI Community

Pratik Kadam

Jun 6 • General discussion 💬

Here’s Why your RAG sucks: ⬇️

Cause when you do RAG the value it provides is derived by “ Time saved from finding the answer. “

Which is Hard to measure and hard to sell.

We won’t see RAG Agents as a Question and Answer Systems but more of Report Generation Systems.

Where as when you sell the Report, it is now a Decision- Making tool that allows you to allocate your resources.

Here’s an example:

We might pay an employee $100/hr and when it uses the RAG Q&A system, it might save him hours of work depend upon the task.

We assume 5 hours saved = $500 saved.

Now when you are a company which what’s to hire, let’s say 5 engineers ( ~ 100k+/per engineer).

We got 20 candidates with their resumes, interview transcripts and all their data. Every company have their Values, SOP’s, a template of hiring process ,etc.

So know the RAG Report Generation system will go through all the candidates data and pick the top 5 engineers which have Values, resumes ,etc. according to your company and generate a report.

Now if that report is generated by an person, he get’s paid around 20% of Engineer’s salary, cause he maid a decision.

That's like 20% of $500k+

Now that single RAG Report generation system has lot more value than Q&A system.

And note that RAG Report generation system is just a For Loop (multiple questions asked to Q&A system) but with proper Template Format, SOP of Hiring process.

This not only will save time but also analyzed entire business data and find a person which will the best for your company.

Also I am working on a similar project but let me know if you want to build something like this, I am open for 1 last client this month.

Thanks :)

2 comments