Hi and welcome to my weekly post! It’s been too long since we spoke last time, so why don’t you schedule a quick call with me here?
In the last post, I talked about my ideas for Slack apps. They were either in the AI space or focused on productivity or integrations. In this post, I cover how I thought about my elimination process and what was left after it.
💡Choosing one idea to commit to
On one hand, I don’t feel like any of my ideas so far are that amazing. Even the ones I get very excited about, when I read them the next day, I think to myself “That’s never going to work” or “Who needs that anyway?” On the other hand, real progress happens when you actually build something, show it to someone and iterate on the feedback. That’s why I’m sticking to my deadline for committing to an idea today.
I previously wrote about my checklist for selecting an idea. That checklist filters for 3 main things:
- Does the problem exist and is it worth solving?
- Do I know who has the problem?
- Can I imagine a solution that’s significantly better than existing ones?
So how do my ideas fare against the checklist?
Sudden death
My three main productivity ideas (reminders for un-responded messages, wrappers for other collaborative bots and cloning OOO messages from email to Slack) have died a sudden death.
The reason for dismissing all of them is that they don’t solve top problems. They are automations for things that don’t take that much time. So not really worth automating.
These were joined in their death by my ideas to integrate Surveymonkey and LinkedIn notifications to slack. LinkedIn’s API actually doesn’t allow subscribing to notifications from any user (which is the main use case I was considering). However, the honest reason for both is that I find it hard to get excited about bringing notifications from one place (API) to another (Slack).
Runner ups
Two other ideas I touched on briefly were about integrating data from Sales (e.g. Pipedrive) or Support (e.g. Zendesk) tools into stats that help teams stay engaged and motivated. Think data-driven leaderboards and competitions. Tools like these are nothing new - even on Slack. Even better, they are standard, proven tools in money-making functions (e.g. sales).
I’m putting a pin in these, and will come back to them in case I fail with…
Bringing GPT-3 to Slack
Of course, you could say I just found excuses to dismiss the other things and am jumping on the shiny new thing. And you’d be right to some extent. However, jumping on a growing trend is also a strategy. There are now almost 1 million (!!!) searches per day for “chat gpt”.
But millions are millions. And GPT is helping those millions solve a few problems already.
The way I’ve come to think about Chat GPT is as an “analyst on my team”. Some of the things it does well is come up with ideas from a prompt (e.g. what are the economic drivers of a clothing store?), synthesize information (e.g. what is the main message in an article?) and write copy/code (e.g. write a message to a customer saying we need to .
There’s even already a tool that uses GPT to answer business questions using frameworks (e.g. cost-benefit, strengths-weaknesses-opportunities-threats). Also, companies are already paying people to think about business questions using frameworks (e.g. management consultants).
However, one thing that I don’t like so much about using Chat GPT is having to go to their website (especially when it’s overloaded). I’d much rather stay somewhere where I already spend a lot of time: Slack. It’s a chat interface too, after all.
So my goal for the next few weeks is to build a prototype of a bot that can answer questions according to a framework chosen by the user. Beep boop (image generated by my other slackbot).
🗿How’s it going, Stefan
Thanks for asking! I’ve recovered from my 3rd cold this winter. Not being able to work at 100% made me feel depressed. And trying to work while sick probably made it harder to recover…
I’ve also been feeling more anxious in the past week. Putting a lot of pressure on myself to get this right.
I’m trying to counter this by focusing on what I can learn, and where I can have most fun. This strategy is somewhat effective.
➡️Next steps
- Write a Python script that can get a decent answer from OpenAI’s GPT3 to some generic business questions (e.g. “Write a cost benefit analysis for laying off 10% of our workforce”)
- Make a first version of a Slack bot that implements the above Python script
- Build a landing page for the bot
- Come up with a name for the bot