At the start of 2024, the main Salesforce AI risk was not moving too slowly. It was funding a collection of disconnected pilots that could never survive production controls.

A strong 90-day roadmap narrows the problem, creates reusable foundations and produces evidence about value, safety and adoption. It should end with an informed scale decision—not a theatrical demo.

Cloud Group point of view

Choose the use case where better context and faster action matter more than model novelty. The first initiative should be valuable enough to measure, bounded enough to govern and representative enough to teach the next three teams.

A practical playbook

The strongest next step is narrow enough to govern and useful enough to produce evidence. We would structure the work around these moves:

  1. Days 1–20: rank use cases by value, feasibility, risk and learning value.
  2. Days 21–40: define the data contract, action boundary and evaluation set.
  3. Days 41–70: build the thinnest end-to-end experience with real permissions.
  4. Days 71–85: run controlled user trials and failure analysis.
  5. Days 86–90: present scale, redesign or stop recommendations with evidence.

The architecture and operating implication

Create only the foundations the selected use case exercises: a trusted context source, reusable actions, traceable prompts and an evaluation pipeline. Avoid launching a broad data program or enterprise chatbot before the outcome proves which capabilities are truly required.

Measure what changes

Model activity is not a business result. Track a small set of indicators that connect behavior to accountable work:

  • Minutes saved or conversion improved at the workflow level
  • Accepted outputs without material correction
  • Failure rate by known scenario
  • Cost per successful outcome

Ninety days is enough to change the quality of the decision. It is not enough to finish the AI transformation—and that distinction keeps the program honest.

Primary sources

This field note is grounded in the product and market context available at the time of publication.