Method
Is your Salesforce org ready for Agentforce? How we score it.
VanguardOps gives an org an Agentforce readiness score from 0 to 100, worked out by fixed rules from the org’s metadata. It scores up to six areas, then checks for an Agentforce license. No license means the verdict is Not ready, whatever the score. The whole method is on this page.
By VanguardOps, written by a Salesforce architect with 13+ years in enterprise orgs. Updated November 11, 2026.
Which areas are scored?
Six areas can be scored from metadata. Each gets 0 to 100, and the overall score is their weighted average. An area with nothing to measure, such as an org with no flows, is left out of the average instead of being counted as zero.
| Area | Weight | What is measured | How it scores |
|---|---|---|---|
| Field descriptions | Weight: 1.5 | What is measured: The share of custom fields that have a description or help text. If the org has no custom fields, all fields are counted. | How it scores: 100 minus the percentage of fields with neither. |
| Flow naming and size | Weight: 1 | What is measured: Flows whose API name isn't PascalCase or snake_case. A size check (more than 50 elements) is built in, but scans don't read flow element counts yet, so today only naming is scored. | How it scores: Each flow can lose half its credit for naming. The other half, for size, only applies once element counts are read. |
| Permission set hygiene | Weight: 1 | What is measured: Permission sets that nobody holds and that aren't in a permission set group, and permission sets with no description. | How it scores: Each permission set can lose half its credit for being unused and half for having no description. |
| Validation rule complexity | Weight: 1 | What is measured: Rules whose error message runs past 200 characters. Scans don't read rule formulas yet, so formula length and bracket nesting aren't judged today. | How it scores: 100 minus the percentage of rules that are complex. |
| Object relationships | Weight: 1 | What is measured: The share of objects with at least one lookup or master-detail field, so an agent can follow from one record to related ones. | How it scores: The percentage of objects that have one. |
| Apex invocable actions | Weight: 1.5 | What is measured: Whether Apex classes expose invocable methods an agent could call as actions. | How it scores: Only scored when the scan has direct evidence. Scans don't pull Apex class bodies today, so this area is usually reported as not checked. |
Why these areas?
An agent works from what the org tells it about itself. It reads field names and descriptions to decide which data answers a question, follows relationships to get from one record to the next, and acts through flows and Apex actions under a set of permissions. Where those are undocumented, tangled or loose, an agent guesses, and so does the person trying to work out what it did.
- Field descriptions carry the most weight. A field called Status_2__c with no description means nothing to an agent or to a new admin.
- Flow naming and size: flows are how agents take action. A consistently named, modest flow is one you can hand to an agent and reason about afterwards.
- Permission set hygiene: an agent runs with the access it is given. Unused and undescribed permission sets make it hard to give it the right access and no more.
- Validation rule complexity: when an agent’s update is rejected, the rule and its error message are all it has to go on.
- Object relationships: an agent answers “what’s open for this customer?” by following lookups. Objects with none are islands.
What do the verdicts mean?
| Score | Verdict | What it means |
|---|---|---|
| 90 to 100 | Verdict: Ready | What it means: The setup is in good shape, and an Agentforce license was found. |
| 75 to 89 | Verdict: Needs minor cleanup | What it means: A short list of fixes first. |
| 50 to 74 | Verdict: Needs significant work | What it means: Plan a cleanup project before a rollout date. |
| Below 50 | Verdict: Not ready | What it means: The setup would hold an agent back, or no Agentforce license was found. |
How does the license check change the verdict?
Tidy metadata doesn’t make an org ready if agents can’t be switched on. So after scoring, the scan looks at the org’s permission set licenses for Agentforce or Einstein generative AI.
- License found: the verdict follows the score.
- No license found: the verdict is Not ready, and the report says why. The score still shows how ready the setup is.
- Couldn’t be checked: the verdict is capped at Needs minor cleanup, with the note “We couldn’t confirm Agentforce licenses, so this isn’t a go-ahead.”
The scan also looks for a Data 360 license (Salesforce’s new name for Data Cloud) and for agents already defined in the org. A missing Data 360 license is reported as a finding; it doesn’t change the verdict.
What does the score not check?
A readiness score that claims to cover everything is worth less than one that says where it stops. This one stops here:
- Knowledge base. Whether there is enough published knowledge for an agent to answer from. Not scored. Reported as not checked.
- User telemetry. How people use the org today. Not scored. Reported as not checked.
- Agents you’ve already built. Topics, instructions, actions and test results are not inspected.
- The data in your records. VanguardOps never reads customer records, so duplicates, blanks and stale values are outside what it can see.
- Apex invocable actions, most of the time. Scans don’t pull Apex class bodies, so this area is usually listed as not checked.
A worked example
An org scores 60 on field descriptions, 80 on flows, 70 on permission sets, 90 on validation rules and 85 on object relationships. Apex actions couldn’t be checked.
- Weighted sum: 60 × 1.5 + 80 + 70 + 90 + 85 = 415
- Total weight: 1.5 + 1 + 1 + 1 + 1 = 5.5
- Score: 415 ÷ 5.5 = 75
- Verdict: Needs minor cleanup if an Agentforce license was found or couldn’t be checked. Not ready if none was found.
Common questions
- Is my Salesforce org ready for Agentforce?
- Two things have to be true. The org needs an Agentforce license, and its setup has to be something an agent can work with: fields that are described, objects that are related, automation that is named and not tangled, and access that is tidy. VanguardOps checks both from metadata and gives a 0 to 100 score with one of four verdicts.
- What is a good Agentforce readiness score?
- On the VanguardOps scale, 90 or above with an Agentforce license found is Ready. 75 to 89 needs minor cleanup, 50 to 74 needs significant work, and below 50 is not ready.
- Can an org score 95 and still be Not ready?
- Yes. If no Agentforce license is found in the org, the verdict is Not ready whatever the score, because agents can't be turned on. The score still tells you how ready the setup is for when they are.
- Does the readiness score use AI?
- No. The score is worked out by fixed rules. The same metadata always produces the same score.
- Does VanguardOps test the agents I've already built?
- No. It checks whether the org's setup is ready for agents. It does not inspect agent topics, instructions or actions, and it does not run test conversations.
- Does it read my customer data to score readiness?
- No. The score is built from metadata: field definitions, flows, permission sets, validation rules, object relationships and license records. It never opens a customer record, which also means it can't judge the quality of the data in those records.
- How do I get my org's score?
- Run the free scan. It needs no form and no credit card. The first screen shows your health score, total cost and top 3 findings; add an email to open the full report, which includes the Agentforce readiness score, the verdict and the findings behind it.
Think a weight or threshold is wrong? Email chris@vanguardops.ai. How scans handle your data is on the Security page.