Skip to main content
A good rollout starts small, shows results early and grows from there. This playbook walks you through four phases, from setup to company-wide use.

Phase 1: Prepare

Set up the organization before anyone else signs in.
  1. Map your structure. Create workspaces for departments or business units, and teams inside them. See Organizations, workspaces and teams.
  2. Choose your admins. Make one or two people Owners. Give team leads the Manager role so they can invite members and see their team’s usage. See Roles and permissions.
  3. Decide on data protection. If people will work with customer or employee data, ask BlackMountain to turn on PII enforcement. Review which models are hosted in the EU. See Data residency.
  4. Add company context. Write a short description of your company, its products and its language in the organization settings, so answers fit your business. See Organization settings.
  5. Check seats and limits. Make sure you have enough seats for the first phase, and agree on a spend limit. See Usage limits.
  6. Write simple guidelines. One page is enough: what data may go into 48°, which models to prefer, and who to ask for help.

Phase 2: Pilot

Start with one or two teams of 5 to 20 people who are curious and have clear, repeatable tasks.
  1. Invite the pilot group. See Inviting members.
  2. Pick three to five use cases. Choose tasks people do every week. See Identifying use cases.
  3. Build the first assistants. Turn each use case into an assistant with clear instructions and reference files. See Assistant Studio.
  4. Run a kickoff. Show a real task end to end in 30 minutes. Point people to the Quickstart.
  5. Collect feedback weekly. Ask what worked, what didn’t, and what people would build next.
If your organization has the Use Cases page, ask pilot users to run its guided tours. Each takes one to three minutes and runs in the real app.

Phase 3: Measure

After two to four weeks, review the pilot.
  • Adoption: open Control Tower and check Usage by User. Who uses 48° every week, and who stopped?
  • Cost: check Cost by Model and Insights. See Spend and cost per model.
  • Value: ask pilot users how much time each use case saves. A rough estimate per week is enough.
  • Gaps: collect the tasks that didn’t work well. They often need better instructions, a skill or an integration.

Phase 4: Expand

Roll out team by team, not all at once.
  1. Share what works. Make the best assistants, prompt templates and skills available to the next teams. See Visibility and sharing and Sharing skills.
  2. Name a champion per team. A colleague people can ask is more effective than any manual.
  3. Connect tools. Add integrations where teams need data from other systems. See MCP servers.
  4. Automate recurring work. Use scheduled tasks for reports and briefings that repeat.
  5. Review monthly. Check usage, cost and feedback once a month, and retire assistants nobody uses.

Common pitfalls

  • No clear task. “Try AI” leads to a few chats and then nothing. Start with a concrete task.
  • Too many assistants too early. Five good assistants beat fifty half-finished ones.
  • No guidelines. Unclear rules make people either avoid 48° or paste in data they shouldn’t.
  • No follow-up. Usage drops after the kickoff without regular check-ins.

Identifying use cases

Find the right tasks to start with.

Analytics

Measure adoption.

Inviting members

Add people to your organization.

Prompting basics

Share with new members.