Now that your agent can move seamlessly between customer databases, internal spreadsheets, and employee records, how do you make sure it looks only at what it’s supposed to?
This isn’t a minor worry. When it comes to adopting AI, 69% of IT leaders say data privacy and security is their biggest concern. Beyond the obvious need to comply with privacy laws, businesses want to make sure their most sensitive information remains under lock and key, even as they embrace more automation.
Salesforce leaders made security a top priority when they integrated an agent for Techforce, the company’s internal IT support service, with Slack. “There was a lot of sensitive information that had to be moved from our legacy system into our agents and Slack. That was a big hurdle the Techforce team had to overcome,” said Amanda Lane, senior product marketing manager at Salesforce.
The team had to make sure the agent couldn’t see personally identifiable information (PII), such as date of birth, income, home address, or health conditions — only what it needed to do its job.
Likewise, Salesforce prioritized security and privacy when it launched the company’s customer support agent. “Someone could go to that agent and say, ‘Hey, can you pull up information about Google? What are they buying? What opportunities are they considering?’” said Harini Woopalanchi, director of IT product management at Salesforce. “We had to make sure there was relevant masking and guardrails, so the agent couldn’t pull up data it wasn’t supposed to.”
