Put Your Superhuman Go Custom Agents to Work in Slack, a Salesforce Company
by David Driscoll, Software Engineer
Think about the best teammate you have. You don’t have to hand them the same task list every morning. They know their role expectations, how your team operates, and when they should jump in without being told. You trust them to read the room.
That’s the bar for a useful AI agent. Superhuman Go lets you build custom agents around the work that’s specific to you and your team. You give an agent a job, show it once how you want that job done, and connect it to the tools and information it needs.
One agent might learn how your sales team decides which accounts are worth pursuing. Another might know the exact taxonomy your product team uses to sort customer feedback. A third might know that your refund policy lives in one specific internal wiki and not in the six outdated docs that also mention refunds.
Describing what an agent does is only part of the job, though. You also have to decide when it should do it. That’s where triggers come in.
Agents that don’t wait to be asked
Most AI still starts with a prompt. You notice a task AI could help with, navigate to the tool, re-explain the context it doesn’t have, make a request, and bring the response back to wherever you were working.
Custom agents in Superhuman Go can specialize in the unique work you and your team do. You define the job once, connect the context, then decide what actually sets the agent in motion. That might be a schedule or something you’re writing, or it could be an event happening somewhere your team already works.
For a lot of teams, that somewhere is Slack. A customer drops feedback in a channel. A rep asks whether anyone has context on an account before a call. Someone wants to know whether a feature shipped to enterprise or only premium plans. Every one of those messages requires a follow-up action, and today a person has to notice it and follow through. With triggers, the message itself can kick off the work without adding to your coworker’s already full plate.
How Slack triggers work
Slack triggers for Superhuman Go let your agents work inside the conversations in public channels that your team is already having. There are two ways to set them up:
The first is to mention an agent when you need it. Add it to the channels where it’s useful, and anyone can @mention it. Picture a rep prepping for a call who pulls up a thread about the account. Instead of leaving Slack to check the CRM, read notes from their last check-in, and review the previous quarter’s deal history, they ask:
@account-prioritizer, what should I know about Acme before this call?
The agent searches against the tools you’ve connected and provides an answer directly in the thread:
Acme opened three support tickets last month, their champion moved to a new role in June, and the account renews in ninety days.
The rep never leaves the conversation, and everyone else who’s working on the deal can see the answer, too.
The second way is to let an agent monitor a channel and respond automatically when a message matches a keyword or phrase you’ve set. For example, if you’ve set up a trigger for your company’s family leave benefits, whenever a message includes “parental leave,” the agent can surface the policy information and associated resources and even tag a point of contact as an FYI in case the sender has further questions.
Give an agent one job to own
Agents perform much better when they’re given a specific job to do rather than when they’re asked to be jacks of all trades and masters of none.
Consider a #customer-feedback channel where sales and customer support post what they’re hearing. Right now, someone manually combs through that feedback and surfaces it to the product team. A customer feedback agent can take on that task. Teach it how your team weighs feedback so it knows a churn-risk complaint from an enterprise account outranks a nice-to-have from a trial user. When feedback lands, it tags the message, files it in the right place, and prioritizes what’s worth reviewing.
Another example is an internal channel where the same handful of questions come up constantly: how to expense conference fees, when the next release freeze is, and which template the security review wants. Build an agent that knows where those answers live, drop it in the channel, and let people mention it. The questions and answers stay in Slack, so the next person who needs help can either scroll or ask a new question.
The same approach works across teams because you control the context, the triggers, and the assignment. Marketing can build an agent that catches campaign ideas before they scroll out of view. Ops can build one that routes intakes to the right owner. Two teams can build agents that seem identical from the outside and behave nothing alike, because you decide what each one can see, what rules it follows, and what triggers it to act.
Why it matters that the work is visible
Putting agents in Slack changes who sees the work. A lot of AI use today happens in 1:1 conversations. You ask, you get an answer, you carry it back to your team and paraphrase it, and everyone else takes your word for what the AI actually said.
Slack is multiplayer, so the team watches the agent get called and sees its answer sitting right under the message that prompted it. They can correct it, build on it, or run with it. The agent becomes part of the conversation. And since it’s a Superhuman Go agent, it brings the same instructions and connected tools into the rest of your team’s work, wherever that may be.
Custom agents aren’t about handing everyone another chatbot. The real shift is that the way your team already works—the judgment calls, the routing, the “someone should really look at this” comments—can start running on its own. You decide what an agent should do, give it the tools and context to do it well, and choose what triggers it. After that, a question or a passing comment in a channel is enough to move the next thing forward.
Start building custom agents in Slack today with Superhuman Go for Slack.
