(This column originally appeared in Forbes)
I’m searching for a part-time administrative assistant for my business. I placed an ad on Craigslist (and don’t judge – Craigslist has actually been a pretty reliable place for my job searches).
Fortunately, I wasn’t disappointed.
To date, I’ve received more than 30 decent resumes. That was the easy part. But unfortunately, now starts the real work: sorting through these documents to figure out the best candidates.
Maybe, I thought to myself, I could take advantage of Google’s recently released note-taking application called NotebookLM? I read about it and was intrigued: an AI assistant that could help me figure out the best candidate! Let’s do it, I thought to myself.
NotebookLM promises some pretty amazing stuff. It’s designed to be a mini large language model, but rather than using all the world’s data it just provides a conversational way to get insights only on the data you provide it. If the application does what it promises I was ready to abandon my current note-taking application and use it to help me answer specific questions about my business…like finding the best candidate for a job based on 30+ resumes.
Unfortunately, I was disappointed.
To its credit, NotebookLM was able to grab the resumes from my Google Drive and allow me to add notes. It’s designed to generate summaries of each file and allow me to search across multiple notes and files within a notebook, all in that conversational GenAI way that that my current note applications don’t allow, or even come close to replicating. I’m told the application will also enable me to create summaries and blog posts using the files I’ve uploaded and the notes created which can certainly be a productivity tool. I can also share my notebooks with others.
All of this is great. But the reality is that NotebookLM failed at helping me with my resume project. As it stands, it’s still not a usable tool for my business. Why?
For starters, the data you can provide it is still very limited. It will only accept pdfs and copied text and not spreadsheets, documents or even URLs from a website. It’s not a dealbreaker – I could convert the Word and Doc files into pdfs (or copy them into my notebook) – but the time this took was onerous. It also didn’t allow me to upload all files at once so I was forced to do this tedious task one-by-one – another annoying time suck. There are still data limitations and although you can search and have conversations with data included in notes and files within a notebook it won’t do the same across multiple notebooks.
Unfortunately its analysis of the 30+ resumes that I (finally!) uploaded wasn’t helpful either. I asked NotebookLM to list the candidates that had financial experience, show me those that had college degrees or evaluate based on a candidate’s experience with technology firms and the answers were either incomplete, inaccurate or basic regurgitations of the resumes without any type of insights or help. I get it: there’s only so much you can do with the data provided. But these would not be difficult requests if I made them to a human assistant.
What do I want from a great AI-powered note taking application? It’s what most businesspeople would want: the ability to ask questions of all the notebooks and get intuitive answers like an assistant would provide. For my resume project who is the strongest candidate? Who has the most skills in one area compared to the other candidates? Who’s got the most experience? Who has the most education? When the GenAI bot hits a wall because there’s “not enough information” what more information is needed and how can the application get it for me?
Am I asking too much of this application? I don’t think so. I’m doing this all manually. Because NotebookLM doesn’t do the job I have to do the job for it.
I’m filling out a spreadsheet that includes each candidate and their qualifications and my observations from these resumes which I’m sharing with my team so they can more quickly evaluate. This is a mundane task that can easily be automated by a good AI note taking application. I think that NotebookLM will ultimately do these kinds of things. But it isn’t happening yet. Like every other AI product offered by big tech so far, it’s more “wait and see what will happen!” instead of “this is happening now!”
To be clear, this is an experimental product. Google says this all over the place with an offer to “try it now,” not “buy it now.” Its landing page literally quotes tech site The Verge’s conclusion that the platform is the “messy beginning of something great.”
I couldn’t have said it better.
