What does AI-ready data mean for Agentforce?
Salesforce's data readiness module scores data across several dimensions. The ones that matter most for a first agent are below.
| Dimension | Question to ask | Agent-ready looks like |
|---|---|---|
| Availability | Is the data present where the agent can reach it? | Structured and unstructured data in a trusted system with a defined update cadence |
| Accessibility | Does the agent see only what it should? | Access limited by record- and field-level security and sensitivity policies |
| Completeness | Are the needed fields filled in? | A complete customer profile for the outcome you want |
| Consistency | Is the same thing stored the same way everywhere? | Standardized values across systems and objects |
| Correctness | Has the data been validated? | Ways to detect and flag bad data before it reaches AI outputs |
| Recency | Is it updated often enough? | Monitoring for stale or missing data |
| Uniqueness | Could duplicates cause a wrong answer? | Identity resolution linking related records |
| Sensitivity | Is sensitive data protected? | Policies enforced so agents access only what is appropriate |
What are the data preparation steps?
Salesforce's Trailhead module on Data 360 for Agentforce sets out three steps: identify the data, locate and connect it, and define the processing it needs.
- Identify what data the use case needs: CRM records such as leads or cases, structured or unstructured data, and any files to upload.
- Locate the source and decide how to connect it. Data in an external warehouse can use a zero-copy data federation connector.
- Outline processing: fix quality issues, and run identity resolution if records need to be unified into profiles.
How does Agentforce use knowledge articles and files?
Through data libraries. Salesforce offers two main data source types for a library: knowledge articles and file uploads. The agent's reasoning then cites the sources it used.
For richer retrieval, Data 360 can build search indexes with retrievers that prompt templates call, which agent actions then use.
What Salesforce data management work pays off first?
Deduplication, required-field cleanup on the objects your agent touches, and a permissions review. These are the same habits that make reports trustworthy.
Our Salesforce data management FAQ covers the everyday side of keeping Salesforce data clean.
- Merge duplicate accounts, contacts and leads, and add duplicate rules so they stay merged.
- Standardize picklists and formats (for example, one opt-in value, not three).
- Archive or retire outdated knowledge articles; publish the ones customers ask about.
- Give the agent's user the least access it needs.
- Set an owner and a review cadence for each data source the agent uses.
How does the Einstein Trust Layer protect your data?
Secure data retrieval keeps your permissions and access controls in force when grounding prompts. Salesforce also describes zero data retention with third-party LLMs, masking of sensitive data before prompts are sent, and toxicity detection.
Frequently asked questions
Do we need Data 360 before using Agentforce?
Not always. Data libraries can ground an agent in knowledge articles or uploaded files. Data 360 helps when the agent needs unified data across systems or retrieval over large unstructured content. Check licensing with Salesforce.
How long does data preparation take?
It depends on how many objects the use case touches and how messy they are. Scoping one use case narrowly keeps the clean-up manageable.
Can you clean our Salesforce data?
Yes. Deduplication, field clean-up, permissions and knowledge tidy-ups are a standard part of our Agentforce readiness work.
Need help with Agentforce?
We help small and mid-size businesses choose a first Agentforce use case, prepare their Salesforce data, build and test the agent and train the team. See our Salesforce Agentforce consulting page, or our Salesforce guides and FAQs.