Every AI conversation eventually needs a human answer

AI is getting very good at answering questions. But inside a company, some of the most important questions can't be answered by a model. 

That's why we're excited to bring Polly into Slackbot through MCP. Now, while working with Slackbot, you can move naturally from asking AI for information to asking your team for input, without leaving the conversation or switching tools.

From AI insight to human input

Imagine you're prepping Wednesday's all hands with Slackbot. You might ask it to pull the focus areas for this week's session, summarize what shipped, or draft the agenda. Eventually, though, you need the questions people actually want answered.

Instead of leaving that workflow to go set something up somewhere else, Slackbot can use Polly. A plain poll would give you three buttons. Polly gives you a moderated, anonymous Q&A with upvoting, reminders, and results you can read back in the same conversation.

Or maybe you're reviewing a launch. Slackbot can help summarize what happened, but you may still want to know how the team experienced it. Ask it to send a quick pulse on how confident people feel about the launch. 

The AI helps you think. Polly helps you ask, track results, and read the room. 

This is the way we've always thought about work

Polly started with a simple idea: getting input from people shouldn't require pulling them out of where they're already working. That meant bringing polls, surveys, Q&A, feedback, and team programs directly into Slack rather than sending everyone another link or asking them to adopt another workflow.

AI changes the interface, but not that principle. Increasingly, the place where work begins may be a conversation with an AI assistant. You ask it to research something. Summarize a thread. Draft a plan. Analyze a problem. Prepare a meeting. And then you need your coworkers. Gathering that human input should feel like a natural continuation of the same workflow.

AI shouldn't create another layer of work

AI is most useful when it removes steps, not when it adds another workflow to learn. As assistants become a more natural starting point for work, people shouldn't have to leave the conversation, remember which tool does what, and manually stitch the process together.

The better model is for the tools your team already relies on to become capabilities AI can call when they're needed. That's what MCP makes possible.

That's the direction we find most compelling. AI as an interface to the tools and workflows your team already trusts, rather than another layer sitting beside them.

The loop gets more interesting from here

Creating a poll is only the beginning. The real opportunity is closing the loop between AI and people.

Polly meets Slackbot