Heading 1
Heading 2
Heading 3
Heading 4
Heading 5
Heading 6
Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat. Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur.
Block quote
Ordered list
- Item 1
- Item 2
- Item 3
Unordered list
- Item A
- Item B
- Item C
Bold text
Emphasis
Superscript
Subscript
The decision on the table is the order of work. Start with a customer-facing agent and you test your data, your permissions and your governance in public. Start with employees and you find the same gaps where a person catches them. The groundwork you lay in those first months, from field-level permissions to a trained admin team, carries over to every AI use case that follows, so it keeps its value even if the first pilot is paused.
Check what you already license
Many Salesforce editions already include predictive Einstein features, such as lead scoring, opportunity insights, case classification or engagement scoring in Marketing Cloud. What you have depends on your editions and contract, so the first step is an inventory rather than a demo. Switching on a feature you already pay for is the lowest-risk AI project you will run.
Those predictive features also test your data. A lead score that looks random usually points to missing or inconsistent fields, and that is worth knowing before any generative model starts writing from the same records. When to choose predictive or generative AI is covered in predictive and generative AI compared.
Clean the data the AI will read
Every Salesforce AI feature answers from your records, so their quality sets the ceiling. Duplicate accounts, empty fields and outdated knowledge articles turn into wrong scores and confident but wrong summaries. Fix the objects your first use case reads, and leave the rest for later.
For some use cases, that means unifying data across systems in Data Cloud (Data 360), the Salesforce layer that builds one customer profile from every channel. The case for that layer is set out in Data Cloud as a customer foundation. For a first use case that only reads data already in Sales Cloud or Service Cloud, you can start without it.
Start generative AI with employees
The safest first step with generative AI is a task an employee reviews before anything reaches a customer. Prompt Builder lets an admin write a reusable prompt template that pulls in record data, such as an account's open opportunities or a case history, and returns a summary or a draft. The employee reads it, edits it and decides what to send. Good first templates are call preparation, case summaries and follow-up emails.
The next step is an assistant inside Salesforce. Einstein Copilot (Agentforce) was the name for the conversational assistant that answers questions and runs actions from the screen your people already work in. Salesforce has since folded it into Agentforce, the platform it now uses for all its AI agents. The principle stays the same: the assistant works with the permissions of the person using it, so your sharing model decides what it can see.
