
Is Your Data Ready for Agents? 5 Ways to Tell
Everyone loves a shiny new object, especially when that object is an AI agent. But before you plunge headfirst into an agent, let’s ask an unsexy question: Where do you stand with your data readiness?
The truth is, your agent will be only as good as your data. Think of your data like the foundation of your house. It’s what everything else is built on. Would you invest in a state-of-the-art kitchen if your foundation was crumbling or you had no foundation at all? Yet many companies rush to invest in agents without considering the state of their data. Or they implement an artificial intelligence (AI) agent only to find it doesn’t work as well as they’d like (again, see: data).
A recent survey shows that 90% of high-level data professionals believe company leaders are not paying enough attention to bad or inadequate data. Meanwhile, a report by IT consulting firm Capgemini found that fewer than one in five companies has a high level of data readiness, with only 9% fully prepared for the data integration and interoperability required for AI.
“An AI strategy without a data framework is just a wish list,” said Kuber Sharma, director of product marketing for Tableau at Salesforce. “Attempting to deploy AI agents without one leads to inconsistent results, security risks, and a lack of user trust.”
But how do you know if your data is good enough? Read on to see if your data is in shape for an AI agent — and where you are on the data maturity spectrum.

What is good enough data?
Before assessing your readiness, it helps to understand what it means for your data to be “good enough.”
The simple definition: It means that your data is high quality and usable enough for the agent to do its job. But doesn’t mean that your data needs to be perfect or that you need to clean all of your data — only what the agent needs to execute its tasks.
To do this, first identify the business problem you want an agent to solve. Then, identify the data the agent needs. If, for example, your agent will handle customer service, it probably needs data on your company’s products, shipping information, and return policies. But it won’t need your annual report or quarterly earnings readouts, or data about your workforce.
The data needs to be organized in a way that your agent can easily access and use. And you only need enough good data to get the agent started. For example, if you want your agent to handle questions about products to start, make sure it has well-organized product data. As you expand its tasks to handle returns and exchanges, you can then clean up the data for those policies. Working with an agent is an ongoing process; you can modify and make improvements over time.


What’s your agentic AI strategy?
Our playbook is your free guide to becoming an agentic enterprise. Learn about use cases, deployment, and AI skills, and download interactive worksheets for your team.


Data readiness questions you need to ask
When Sharma talks to customers, he helps them assess data readiness by sharing this checklist of questions:
1. Is your data unified and harmonized?
Put simply, is there one place where your agent can find all your data? Or is the data siloed and scattered across the company? “If your data is fragmented, the agent will deliver fragmented and inconsistent experiences,” Sharma said.
Your first order of business is to unify your data. Data Cloud, the Salesforce platform that gives agents instant access to all the data across your organization, can help. Data Cloud unifies data — whether it’s stored in Sales Cloud, Service Cloud, Marketing Cloud, Commerce Cloud, or external systems like data warehouses and data lakes — into a single, coherent profile.
But your data also needs to be “harmonized.” In other words, it needs to be in a standardized format. The United States, for example, might be referred to as “USA” in one place, but as the “U.S.” or “United States” in others. When you standardize formats across all your data, it becomes easier for your agent to understand and use.
2. Have you resolved identities and is the information up to date?
Even after all your data is unified, you may still have multiple entries for each individual customer. Maria Gonzalez, for example, could have several different profiles: one with the name of M. Gonzalez, another under her full name, and a third in which she’s identified by her email address only.
If you want data to be ready for an agent, these duplicate records need to be merged to create a single profile for each customer. Salesforce has a tool called Identity Resolution that can perform this task.
You also want to make sure everything’s up to date. Is Maria Gonzalez’s email correct? Do you know which products she’s using now? Your data team may need to check this manually. But clean, updated data helps your agent perform well. Old, incorrect data can lead to frustrating experiences for customers and unreliable outcomes like hallucinations.
3. Do you have clear data governance policies in place and is your data secure?
Customer trust is key to deploying an AI agent successfully. And to create trust, your agent needs clear rules. You should be in compliance with local regulations, such as the CCPA (California’s data protection law) or the GDPR (the European Union’s regulations). You also need a data governance framework, or a clear policy for data quality, security, and access.
An important part of this framework is making sure your data is secure. That’s one of the advantages of Data Cloud, which is secured by Salesforce’s Einstein Trust Layer; it automatically masks sensitive data and blocks information from being stored by external LLMs.
You also want your agent to only have access to the data it needs to do its job. An agent that’s scheduling appointments in a clinical setting, for example, should not have access to private or sensitive information about a patient’s health.
4. Can you activate your data in real time?
This is probably the most action-based of all Sharma’s questions. In simplest terms, it means, is your data ready for immediate use in workflows? Can you use it to power real-time action?
The only way to answer is to experiment. After you’ve set up your agent, feed it some data and see if it can take the action you want. For example, if you have a customer service agent, see if it can access your data about returns and issue a shipping label.
You may also want your agent to proactively make recommendations to customers. Sharma pointed to a few large, successful companies that do this well. “When I order a matcha latte at Starbucks on my phone, I immediately get a message, ‘Have you tried the matcha cookie as well?’’’ he said. “Or when you order shoes on Amazon, they usually follow up with a ‘You forgot socks, dude.’” If you want your agent to do the same, load it with the data it needs and see if it prompts customers with recommendations.
5. Have you established a feedback loop?
As much as you might want to “set and forget” your agent, well… you can’t. Agents need humans in the loop to make sure they’re doing their job. Ask yourself: What systems do you have in place to test your agent? How can you monitor your agent’s performance and make changes if necessary?
Salesforce’s Agentforce Command Center can help with this. It’s a first-of-its kind observability tool designed to evaluate AI agent activity and help you figure out where you need to make improvements. But no matter which tool you use, you need some type of system in place to check if your agent has the data it needs and is accessing it correctly.
Salesforce’s AI Readiness Assessment tool can also give you a good idea of whether you’re ready for an agent.


Download our Essential Data Sources worksheet
For each data source your agents will need, score its readiness from 1 (needs significant improvement) to 3 (well established). Identify an action to prepare this data for implementing AI agents, and assign an accountability owner.


Where are you on the data maturity spectrum?
At the same time you’re going through the checklist, you can see where you are on the data maturity spectrum with this list suggested by Sharma. Here’s each stage of readiness on the road to deploying an agent:
Once your data is clean and organized, you’ve got a solid foundation. So go ahead and make plans for that state-of-the-art kitchen — er, AI agent. It’s not always fun doing the unsexy legwork first. But the payoff, a high-performing AI agent, is worth it in the end.
Images by Aleona Pollauf/Salesforce.