Intro

Spoke.ai is a tool that helps product people (PMs, POs, Engineers) launch products faster using AI. In a nutshell, the product is an inbox that centralizes all of the notifications from the users’ tool stack (Slack, Jira, Notion, Confluence, Figma Linear, Miro, or Github) and by using Artificial Intelligence, it does 3 core things:

1. Generates conversation summaries so they won’t have to waste time reading tons of messages.
2. Prioritizes and labels conversations so users can get a glance at the type of conversations they have pending.
3. Suggests the most relevant actions that someone has to take based on how a conversation is developing (for example if someone reports a bug or flags a feature improvement that has to be developed soon, then Spoke suggests creating a ticket out of it).

Here is a quick look at how its final state looks👇. Don’t worry, I will tell you how we got to this final state.

A quick run-through of how the app worked in its final state

A quick run-through of how the app worked in its final state

To use the app, the users would have to start by authorizing and connecting their Slack workspace with Spoke.ai so the app could capture the incoming conversations from there. Even the conversations coming from other tools like Jira, Notion, Linear etc were coming via their Slack workspace if they had those tools connected to their workspace already.

Finally, users had to add their most important Slack channels to their Priority List on Spoke. This way, only conversations coming from those channels would be processed by the app, therefore allowing the user to stay focused only on the channels that matter most to them.

Problem

Back then, Spoke was a web app that users could only access via their browsers and the goal of the product was to help users cut through Slack’s noise. That is because people have constantly reported that there’s usually a lot of communication mess in there and it can distract them from their focus time.

However, the web app format did not solve the problem. Users would still go back to Slack because either Spoke.ai could not fit into their workflows, or they didn’t fully trust the AI-generated summaries and suggested actions so they always had to go back to Slack for validation.