A working setup is only half of a successful rollout. This article suggests how to introduce AI Engine to your team so it becomes part of daily work. It is written for admins and sales managers.
Before go-live#
- Finish the checks. Complete Run a first test as a real user, in your team's language, on every app and record type where you placed the components.
- Start with a pilot group. A small group is enough, within your license's user limit (see below). Include at least one experienced rep who is sceptical, not only enthusiasts.
- Pick one task it should make easier. For most teams that is preparing for meetings or writing follow-up emails. Lead with that task instead of showing every feature.
- Prepare two or three prompt templates for your team's recurring work, for example an account review or a weekly pipeline brief. Custom actions written in your own business language are what make AI Engine feel built for your team. See Prompt templates.
- Name an owner on your side who answers questions and collects feedback.
Assign access#
- Assign AI Engine User to the pilot group. See Assign permission sets.
- Give the same users access to your AI provider credential, for example by assigning the permission set that holds the External Credential principal access. See Connect your AI provider.
- Make sure the components are on the pages the pilot group actually uses. See Add components to pages.
Keep the number of people with AI Engine permission sets within your license's user limit (up to three users during the free trial). Above the limit, AI actions stop with a "User limit exceeded" message until you remove assignments or get a license for more users from Cloud Connected. See License and trial.
The first session with users#
Keep it short and hands-on, on the users' own data:
- Ask one question each person can check from memory, for example "how many open opportunities do I have?".
- Walk through the one task you picked, on an account they know.
- Let them ask their own questions.
- Show the ? icon in the chat header. It switches the chat to Help mode, where questions about AI Engine itself are answered from the built-in guide.
What to tell users#
- In the chat and record actions, AI Engine works with the records and fields they can see in Salesforce, limited to the objects and fields the admin set up for it. If something is missing, it is usually a setting; tell the admin.
- Ask about one thing at a time. Questions that combine two objects are harder; splitting them gives better answers.
- Some actions take longer, especially Meet Prep with web search.
- AI can be wrong. Check anything before sending it to a customer. AI Engine drafts and the user decides: a drafted email goes out, and tasks, PDF files and Magic Snap records are created, only when the user confirms.
The first four weeks#
| Week | What to do |
|---|---|
| 1 | Open the Activity Monitor. See who has used AI Engine, and talk to the people in the pilot group who have not. |
| 2 | Collect the questions people ask repeatedly. Turn them into prompt templates. |
| 3 | Check in with the most active user and the sceptic. |
| 4 | Review the field scope in Agent Objects: add fields people need, remove fields nobody uses. Fewer fields keep requests smaller. |
Then extend access to the rest of the team.
Signs that need attention#
- Usage concentrated in one or two people.
- Everyone uses the same single action. The others may not fit their work yet, or nobody showed them.
- Usage drops suddenly. Ask what happened; there is usually a specific bad experience behind it.
See Activity Monitor for how to read usage data.