4 must-have AI tools for knowledge workers
Everyone together now, to the tune of "My Favorite Things": Claude's context window and GPT-four-point-oh / Copilot web search and Fathom-dot-video...
If you feel overwhelmed by keeping track of what’s happening in the AI tools market, you’re in good company. It’s literally my job and I still feel constantly behind. But to make your job easier, here are the top 4 tools that I recommend practically any knowledge worker start using, right now.
ChatGPT-4.0 for problem-solving, strategy, brainstorming, and writing
Claude.ai (only available in the US and UK for now) for similar purposes to ChatGPT, but when you need a vastly larger context window
Perplexity Copilot for web research
A transcription tool: Either a notetaker app that integrates with video calls, like Fathom.video or Otter.ai, or a voice memo transcriber like Whisper or Otter.
The additional tools I would recommend for a specific person or team depend on your existing productivity stack (do you use Microsoft Office or Google Docs?) as well as your job function (do you analyze data? make graphics for social media posts?). But I have yet to meet many knowledge workers who this set of tools can’t help dramatically.
Why these four?
Claude and ChatGPT-4
Claude and ChatGPT-4 are general purpose Large Language Models that do not have access to the live internet. They probably know less than you about whatever 2-4 fields you are most expert in, but they know more about almost every other domain of human knowledge than you do. If you give them the right context and feedback, they are also very good writers across a wide range of tasks — clearly inferior to the best human writers at any particular type of task, but better than most people who write as a significant part of their job. And a lot faster (and cheaper).
They’ll also both lie to you/make shit up (called “hallucinations”), so you have to stay on your toes.
I debated whether to list them both vs saying you could pick one — but they are both core parts of my own workflow for different purposes, and they seem to both be useful to the people I’m coaching and training. You can think of each of them as a staff member working for you who has different strengths and weaknesses, and you just have to learn what types of tasks it’s best to delegate to each of them.
At a high level, my summary is:
Reasons to use Claude instead of ChatGPT-4:
Massive context window (~75K words — i.e., you can feed it an entire medium-length book) is a gamechanger for many tasks
Handles text and PDF attachments with ease
Concise default writing style, which I personally appreciate in many contexts
Reasons to use ChatGPT-4 instead of Claude:
Generally just slightly smarter, more strategic, and more creative than Claude, in my experience. If I don’t need the big context window, I default to ChatGPT.
Custom instructions save a small amount of time for many tasks.
Sharing functionality
For advanced users, the ecosystem of plug-ins and features like Advanced Data Analysis can be very valuable.
Perplexity Copilot
Perplexity Copilot is essentially an alternative to Google. It’s based on GPT-4 but is internet-connected and cites its sources. This is kind of what Bing and Bard promised to be, but Perplexity Copilot is way better than the other two at a lot of stuff.
My rule of thumb is that I use Perplexity Copilot anytime I think that Googling and clicking through links to find the right information or source will take more than 3 minutes. As a result, I spend probably 50% less time Googling. (And yes, this is an enormous potential challenge to Google’s business model!)
Perplexity Copilot chats are also shareable, so I can easily show you some of the ways I use it:
Tech support for Macbook issue (btw, it was correct, there was a hardware problem)
Make sure you turn on Copilot! Otherwise Perplexity is just another skin on top of ChatGPT-4. You get 5 free queries every few hours, or you can upgrade to get what is in practice unlimited queries.
Voice-to-text tools
Ok, I kind of cheated. This is actually a couple different types of tool lumped under one. For anyone who has regular video calls, I recommend an AI meeting notetaker — specifically, for Zoom calls I happen to like Fathom.video’s UX the best of any tool I’ve tried so far.
You could also try Otter.ai or Fireflies.ai, or any number of other competitors. Before you settle on one, make sure you are comfortable with the terms of service and the way your data could be used. And always make sure other participants in your calls know that you are being recorded.
Once you have one of these tools, you no longer need to take real-time meeting notes. My favorite workflow is to copy the transcript of the call into Claude and then ask Claude questions like “What did I promise I would do after this call?” or “List the other person’s main concerns about X.”
If you don’t have lots of virtual meetings, or your virtual meetings are too sensitive to be recorded, then here are two other voice-to-text tools and user flows you might be interested in:
Have a chat outloud with Inflection’s Pi chatbot on your iPhone. Can be a good way to use the time while driving! If you’re a verbal processor, talking to Pi is a great way to problem-solve without needing to find a person to talk with.
Voice memo transcription: Let’s say you meet with a donor. Instead of spending 20 min writing up notes once you get back to the office, take a voice memo in the taxi on the way back. Then have Otter.ai or another transcription app transcribe it for you, and have Claude clean the notes up into a useable format.
AI Impact Lab runs workshops on Claude, ChatGPT, and Perplexity Copilot. Reach out if you’re interested in hiring us to come support your staff in learning how to leverage these tools effectively!