Everything you'd hire a partner for, in one console.

Talk to your org. It reads the org before it answers.

AI Command is more than a chatbot on top of Salesforce. It reads your real org, writes the SOQL, checks security, and puts every write behind an approval step.

• Answers questions about your data. "How many" becomes COUNT(), "per stage" becomes GROUP BY, "this quarter" becomes THIS_QUARTER.
• Shows who can see what. "Who has modify all data" is treated as a security question, not an update.
• Remembers the conversation. "Which are they?" follows on from the last answer. Threads come back when you reload.
• Hand a chat to a colleague. They continue the same thread with their own org access. Workspace members only, no public links.
• Builds behind a gate. Generate → code review → validate against the org → you approve.
• Refuses unsafe requests. "How do I delete a field?" gets an answer, not a deletion.

Tested against 202 real phrasings on every build.

From a requirements document to deployed Salesforce, with a source for every requirement.

Attach the BRD in the command box. Nova recognises it and asks before it builds anything. Every requirement comes back linked to the page and line it came from.

• Any format. Word and Excel are converted to text. PDFs and scans are read with vision.
• Checked against your org. A BRD says "Applicant". Nova checks whether you already have Applicant__c, or whether standard Contact already covers it.
• Coverage banner. Every requirement is either mapped to metadata or set aside with a reason. None are silently dropped.
• Asks instead of guessing. It gives concrete options with a recommended default, and you send your answers once.
• Private by default. Objects holding personal data start with private sharing.
• Send the plan for review. A colleague can check it before it's built.

Cost checks run before any AI call. Uploading the same file twice reuses the first session.

It learns your org, and checks its answers against it.

Each org gets its own long-term memory. Scans and everyday work feed it, and every AI generation starts from it. It improves as you use it and drops lessons that go stale.

• Knowledge. Metadata and past work, searchable by meaning. It finds "the field that tracks customer tier" without the API name.
• Structure. A map of which objects, fields, Flows and Apex connect, so Nova knows what a change will affect.
• Memories. Lessons like "this org calls customers Members" are consolidated daily. Unused ones fade.
• Output verifier. Generated output is checked against the real org structure before you see it.
• Not model training. It retrieves and remembers for your org only, inside your workspace.