← Daniel Clark
02 / 08 Slack Swarms & Agent Integrations

Sign in to edit

Get the deep dive

The full story behind Slack Swarms & Agent Integrations: the research, the design decisions and how the team got there. Leave your details and I’ll send it to you.

Used only to send you the deep dive. Never shared.

Deep dive requests

    02 · Slack · MuleSoft · Swarms

    Slack Swarms & Agent Integrations

    Bring the experts to the problem. Then bring the agents and the contextual data.

    01

    User problem

    A support case has been escalated. The order was stuck and the shipping data is split across the order system, Service Cloud and the dispatch database, each in their own ERP. The case bounces from tier to tier while the customer waits, and every Slack bot can only see its own app

    02

    My role

    Design and architect how a Slack swarm integrates across every system, so the team resolves the support case together in a single channel and the customer’s data from each system comes to them.

    03

    How I led

    The first prototypes in 2021. The swarm was the unit of work and has now evolved into the whole organization’s AI headless strategy

    01

    Pull the order into the swarm

    One command brought the order and the case into the channel, with the experts who could fix it.

    02

    One bot for the whole workflow

    One bot followed the process across every system, so the swarm acted from the message: reassign, update, escalate.

    03

    Admin and IT in the same swarm

    The admin builds the flow; IT approves it and hears about failures in the channel.

    04

    Agents join the swarm

    An action that runs from a message can run for an agent too. The screen is optional.

    04

    Business outcome

    Swarming became how Salesforce runs support across headless experiences

    01

    Swarms

    Days-to-close improved 26% and same-day resolution 19% with swarming in Slack.1

    02

    Swarm agents

    Engineers @mention Hive in the thread. The wait for help went from hours to under a minute, saving 30 to 40 minutes a case.3

    03

    Shipped

    Composer launched in 2021 with a Slack connector; Flow in Slack followed in 2022.46

    Bring the experts to the problem.One bot per workflow, not per app.Agents join the swarm.

    What the teams shipped

    Service, sales, admin Composer Bot (concept, 2021)

    One bot for a whole workflow; pull an order into a swarm.

    Service agent Swarm in Slack

    Start a swarm on a case; the right experts join in Slack.

    Salesforce Help: swarm in Slack from Slack ↗
    Admin MuleSoft Composer for Slack

    A Slack message can start an integration, or be a step in one.

    MuleSoft Documentation: Composer Slack connector reference ↗
    Admin Flow in Slack

    Automations that post and run inside Slack.

    Salesforce Admins: introducing Flow in Slack ↗
    Support engineer Hive, a swarm agent

    @Hive in a thread; a sourced answer in 20 to 30 seconds.

    MuleSoft Blog: Hive, MuleSoft Support’s swarm AI experience ↗

    Sources

    1. Slack: Salesforce’s support team resolves cases 26% faster with Slack2 September 2021
    2. Salesforce Help: swarm in Slack from Slack
    3. MuleSoft Blog: Hive, MuleSoft Support’s swarm AI experience4 June 2026
    4. Salesforce: MuleSoft Composer launch10 March 2021
    5. MuleSoft Documentation: Composer Slack connector reference
    6. Salesforce Admins: introducing Flow in Slack2022