Getting Your Salesforce Org AI-Ready: A Data Quality Checklist

An AI agent acting on bad Salesforce data does damage faster than a report showing bad data ever could. AI-ready does not mean every record is perfect. It means the specific records and content your first agent will read are accurate, complete, current and governed. Most of the checks below use tools already in your org.

What does "AI-ready data" actually mean in Salesforce?

Salesforce's Trailhead module on AI project planning says an organisation needs to be "data-ready to start an AI project, which means the data for the project is accurate, available, accessible, and securely governed" (Trailhead, as at September 2026).

Scope matters more than adjectives. A Salesforce blog post on agent data readiness (2 September 2025) says it "doesn't mean that your data needs to be perfect or that you need to clean all of your data", just "only what the agent needs to execute its tasks". Its example: a service agent "probably needs data on your company's products, shipping information, and return policies."

Salesforce's AI readiness page sets the order of work: "AI success starts with clean, connected data. Audit your current landscape to uncover gaps in quality, access, and consistency."

In practice: name the first agent's job, list what it will touch, and get that to a standard you would let a new employee act on unchecked. Everything else can wait.

Why does bad data hurt more with an agent than with a report?

A dashboard built on duplicate accounts shows a wrong number, and a human questions it. An agent built on the same duplicates emails the wrong contact or quotes a policy you retired two years ago, at volume, at 2am, without asking.

Salesforce's research says the problem is widespread. These are global enterprise surveys, not small-business data, and Salesforce sells the cure. Its State of Data and Analytics, 2nd edition (4 November 2025; 3,800 analytics and IT decision makers and 3,852 line-of-business leaders across 18 countries, surveyed 27 June to 13 August 2025; figures as at September 2026) reports:

• "84% of data and analytics leaders say their data strategies need a complete overhaul before their AI ambitions can succeed."
• Leaders "estimate over a quarter (26%) of their organizational data is untrustworthy."
• "89% of data and analytics leaders with AI in production say they've experienced inaccurate or misleading AI outputs."

The State of Service: AI Agents Edition (20 May 2026; 3,075 service professionals, surveyed 9 March to 4 April 2026) adds that "Seventy-two percent of service operations professionals say data readiness is a major blocker to AI".

Those already running AI report the data problem loudest. Fix it before the pilot, not after.

The AI-readiness data quality checklist

Every row is checkable in Setup or a standard report. Limits and editions are quoted from Salesforce Help as at September 2026 and linked where they appear.

#	What to check	Why it matters for an agent	How to check it	Native tool
1	Duplicate accounts, contacts and leads	Agent acts on the stale copy, or contacts both	Report grouped by email or name; review Duplicate Record Sets	Duplicate Rules and Matching Rules: "up to five active duplicate rules per object" (Essentials through Developer; Starter and Pro Suite not named)
2	Required fields and validation rules on fields the agent reads	Blank fields become guesses	Report on blanks in the key fields	Validation Rules "verify that the data a user enters in a record meets the standards you specify before the user can save the record"
3	Stale and abandoned records	Old facts become confident answers	Report on Last Activity; open opportunities past close date; closed-lost with no reason	Reports; a scheduled Flow to flag or archive
4	Record ownership and inactive users	Tasks and notifications route to nobody	Filter Users to inactive; report records by owner	Mass Transfer Records
5	Picklist and stage hygiene	"NZ", "New Zealand" and "nz" are three values to an agent	Report grouped by each key picklist	Restricted picklists and global value sets
6	Knowledge articles and FAQ content	Grounding source; an outdated policy becomes a wrong answer at scale	Knowledge report by last published date and owner	Knowledge; Agentforce Data Library
7	Field history tracking on the fields that matter	To see what the agent changed and reverse it	Set History Tracking in Object Manager	Field History Tracking: "20 fields per object", kept "up to 18 months, and up to 24 months via the API"
8	Storage headroom	Dedupe jobs and archives need room	Storage Usage in Setup	Allocations: Enterprise "10 GB" plus "20 MB" per user licence; a 10-user org has about 10.2 GB (our arithmetic)
9	Permissions and sharing	Agent inherits what it is allowed to see	Profiles, permission sets, Sharing Settings; count who has View All Data	Profiles, permission sets, sharing rules
10	Integrations and source of truth	If Xero and Salesforce disagree, which does the agent quote?	List every integration and the owning system for each synced field	See our Salesforce Xero integration design guide

Which Salesforce data quality tools are free, and which are not?

Most of what you need is already in the licence. Editions and limits are quoted from Salesforce Help as at September 2026.

Tool	Cost	Editions (per Salesforce Help)	Notes
Duplicate Rules and Matching Rules	Included	"Essentials, Professional, Enterprise, Performance, Unlimited, and Developer Editions"	Starter and Pro Suite not named
Validation Rules	Included	Essentials, Contact Manager, Group, Professional, Enterprise, Performance, Unlimited, Developer	Too many rules and users stop entering data
Field History Tracking	Included	Contact Manager through Developer	20 fields per object; history "does not count against storage limits"
Field Audit Trail	Paid	Enterprise, Performance, Unlimited "with the purchase of Salesforce Shield or Field Audit Trail licenses"	Keeps history "until you manually delete it"
Data Import Wizard	Included	All except Personal and Database.com	"up to 50,000 records at a time"; import only; cannot import opportunities or cases
Data Loader	Included	Enterprise, Performance, Unlimited, Developer	"Between 5,000 and 5 million" records; test in a sandbox first
Data Quality Analysis Dashboards (Salesforce Labs)	Free	Check the listing	Old, formula-field-based dashboards; trial in a sandbox first
Data Detect	Paid: "Requires the Salesforce Shield add-on subscription or the Data Detect add-on license"	Enterprise, Performance, Unlimited, Developer	Finds sensitive data; classification, not cleansing
Salesforce Optimizer	Retired	None	Older guides still recommend it

Strike Salesforce Optimizer off your list. Salesforce's knowledge article "Salesforce Optimizer App Retirement" (12 June 2026) says it is retiring the app "with the Winter '26 release for all organizations on Hyperforce" and that it "won't be accessible to non-Hyperforce orgs with the Summer '26 release". Any guide that starts with Optimizer is out of date.

What does Agentforce specifically need from your data?

If you are being sold Agentforce, two things come up that most SMBs have never switched on: Data 360 (formerly Data Cloud) and the Agentforce Data Library.

Salesforce's Agentforce page (as at September 2026) says agents are "grounded in Data 360 and Customer 360, so every agent knows the customer before it says a word." Data libraries do the grounding: "you can add knowledge articles, uploaded files, or custom retrievers as a data source", and "To use Agentforce Data Libraries, you must have Data 360 set up. Data libraries consume Data 360 credits." (Setting Up Data Libraries). So row 6 of the checklist is not optional for a service agent.

Prerequisites, as at September 2026:

• Edition. Data libraries need "Enterprise, Performance, and Unlimited Editions" with an Einstein or Agentforce add-on, "or Agentforce Foundations" (Salesforce Help).
• Einstein. First, "turn on Einstein Generative AI" (Enable Agentforce). That page still notes: "In August 2026, we plan to turn on the Agentforce platform for all orgs with Agentforce access".
• Data 360. Salesforce Foundations is "a $0 add-on" that includes "Data 360 Provisioning" and "1 TB of Data Storage", supported on Enterprise, Unlimited and Performance. Salesforce adds "Availability varies by region"; check the credit allowance in your own contract.

On Starter Suite or Pro Suite? None of this applies to those editions. Our Einstein vs Agentforce guide covers what AI you can actually buy, and Agentforce for small business what it costs.

What does a 30-day plan look like?

Week 1: measure. Install the free dashboards or build three reports: duplicates by email, blank key fields, records untouched in 12 months. Write the first agent's job in one sentence, then list every object, field and article it will read or write.

Week 2: fix the trust-breakers. Merge duplicates on those objects and turn on a duplicate rule so they do not come back. Add validation rules on the fields that matter most. Archive or close dead records. Before opening Data Loader for bulk updates, read our data migration best practices.

Week 3: content. Write or tidy the knowledge articles the agent will ground on. Retire anything out of date, give each article an owner, and read the top ones as a customer would.

Week 4: governance, then pilot. Turn on field history tracking where the agent will write, limit the running user's permissions to the job, and document which system owns each integrated field. Then pilot one agent in a sandbox; Salesforce's advice for its Testing Center: "use Testing Center only in your sandbox environment".

An honest note: if you cannot write the first agent's job in one sentence, you are not ready to buy credits. Do the data work anyway; it pays off in reporting whether or not you buy AI this year.

Where does SAASKOOL fit?

The free Salesforce health check already includes a data quality assessment (see why your org needs a health check), and you keep the written report. If the report says you are ready, our Salesforce AI and Agentforce services cover the agent build. If not, a fractional admin from NZ$99/hr + GST can work through weeks 2 to 4.

Frequently Asked Questions

What does AI-ready data actually mean in Salesforce? Salesforce defines data-ready as data that is accurate, available, accessible and securely governed, scoped to the project. It does not mean every record is perfect, only that the records, fields and knowledge content your first agent will read or write are trustworthy enough to act on unchecked.

Do we need Data 360 (Data Cloud) to use Agentforce? For grounding an agent on knowledge articles or uploaded files through an Agentforce Data Library, yes: Salesforce states you must have Data 360 set up and that data libraries consume Data 360 credits. Salesforce Foundations, a no-cost add-on for Enterprise, Unlimited and Performance, includes Data 360 provisioning, though availability varies by region.

Which Salesforce data quality tools are free? Duplicate Rules, Validation Rules, Field History Tracking, the Data Import Wizard and Data Loader are included with the editions Salesforce lists for them, and the Salesforce Labs Data Quality Analysis Dashboards app is free on AppExchange. Field Audit Trail and Data Detect need Shield or a separate add-on licence, and Data Detect finds sensitive data rather than cleaning records.

Is Salesforce Optimizer still available? No. Salesforce's retirement notice of 12 June 2026 states the Optimizer app was retired with the Winter '26 release for organisations on Hyperforce and is not accessible to other orgs from the Summer '26 release. Guides that start with Optimizer are out of date.

How long does it take to get an org ready for an AI agent? For a small business preparing one agent with a clearly defined job, a focused 30-day plan is realistic: measure in week one, fix duplicates, validation and dead records in week two, tidy knowledge content in week three, then turn on tracking and governance before a sandbox pilot in week four. Orgs with years of unmanaged data, or no clear first use case, take longer.

---

Start with the free health check for a read-only verdict on your data, then talk to us about Salesforce AI and Agentforce services when you are ready to build the first agent.