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 where the first AI budget and data effort go. Predictive models are usually lighter to run and easier to measure, because each one produces a score you can test against real outcomes. Generative agents can take work off people's desks, but they need clean knowledge, guardrails and someone who reviews their output. Choose the wrong type for the job and you pay for a capability your data cannot support yet.
Predictive AI in Salesforce scores, ranks and forecasts
Predictive AI looks at historical data and estimates what comes next. In Salesforce it appears as Einstein features: lead and opportunity scoring in Sales Cloud, Einstein Engagement Scoring and Engagement Frequency in Marketing Cloud, send time optimisation for email, and product recommendations in commerce. Salesforce launched Einstein as predictive AI for sales and marketing, and these features still carry the Einstein name.
The value is easy to test. A score either predicts the outcome better than your current rule or it does not, and you can check that on last quarter's data before anyone changes a process. Predictive models do need volume: enough closed deals, sends or orders for a pattern to exist. Below that level, a well-written business rule does the same job at no extra cost.
Generative AI in Salesforce writes, summarises and acts
Generative AI produces new content: an account summary before a meeting, a draft reply to a customer, a product description or a campaign brief. In Salesforce, that capability started as Einstein GPT (Agentforce) on the Einstein 1 Platform (Agentforce 360 Platform). Salesforce has since brought its generative and agent capabilities together under Agentforce, and the naming history is covered in the story of Einstein GPT.
Agents go a step further than drafting. They take actions, such as updating a case or booking a meeting, within limits you set. That makes them useful for work that follows written rules. It also makes their mistakes more visible, because an agent can act before a person has read what it wrote.
The two types fail in different ways
A predictive model that gets it wrong produces a bad score, and a sales representative or marketer usually filters it out before it matters. A generative agent that gets it wrong says something to a customer. That difference sets the governance each one needs. Predictive AI needs periodic checks on accuracy. Generative AI needs rules on what it may say and do, plus someone reviewing a sample of its output every week.
Measurement differs too. You judge a predictive feature by how well its scores match outcomes, such as conversion by score band. You judge a generative feature by the work it removes: cases resolved without a person, handling time, drafts accepted without edits. In our marketing automation work, with campaigns that reach two million recipients, send time and frequency decisions are a clear example of where predictive scores pay off, because small gains repeat across every send.
