Crown's AI Playbook Starts With Keeping Data Off the Model

Crown Information Management launched its physical box storage business in 1983, spending over four decades in the industry before expanding into digital data management. It stores and processes paper and digital records for banks, hospitals and manufacturers in more than 50 countries. In March 2024, the company dropped "Records" from its name and became an information business built for the AI era.

Group CIO Chris Davis-Pipe and Head of Digital, Asia, Antony Anand outlined what that change looks like in practice, in a conversation with AIM Media House in. Their answers centered on one distinction: what data reaches an AI model, and what never should.

A Rebrand Built Around Digital Trust

The rebrand also introduced a new strapline, 'work uncomplicated'.

"If we could rebrand ourselves as Crown Information Management, we could take that opportunity to communicate to our customers that we can do more than what we can do with records," Anand says. "That's the whole idea of rebranding, and I think that's working for us right now. People come to us and talk to us about problems beyond storing records."

The company's platform, EffiDocX, uses optical character recognition and machine learning to pull information out of scanned documents so clients can search records by asking questions rather than hunting for files. Crown also lets clients choose which AI model touches their data, including bringing their own enterprise OpenAI account.

"We don't want to just provide all the data that's in [EffiDocX] to these AI companies," Davis-Pipe says. "We need to maintain the data within our own infrastructure so that we're happy it's secure, the clients are happy it's secure, and then we will only send the vectorized information to the AI company."

He raised the case of Claude conversation links that had turned up in Google search results that month. "We obviously can't have that with our customer documents," he says.

Security Starts With What AI Never Sees

Davis-Pipe says the company's comfort with AI internally began with Microsoft Copilot, specifically because it stayed contained. "Where we really started to see use within Crown was with the launch of Copilot, because at that point we felt safe to adopt it within our organization," he says. "We're able to keep it within our Microsoft tenant."

That same instinct carries through to how Crown built EffiDocX. "It starts with the cloud infrastructure and making sure that that is properly configured, properly audited," Davis-Pipe says. "We get some experts involved to really try and hack it."

He adds that the platform runs single sign-on, multi-factor authentication and role-based access control, so an HR manager can only reach the records assigned to them. Every file is encrypted using 256-bit AES, both in transit and at rest.

The same concern surfaced elsewhere in the industry that week. Days earlier, an OpenAI model broke out of a sandboxed test and hacked into Hugging Face's systems while trying to cheat on an evaluation. Around the same time, Microsoft CEO Satya Nadella told enterprises to keep their AI systems separate from any single model provider, warning that no single AI lab should be trusted with an organization's full infrastructure.

ROI Before Ambition

Anand describes an approach built around forcing clients to justify a digitization request before Crown will act on it. The company pushes back first, then decides.

"It starts with us asking why you want to digitize," Anand says. "Say you want it available in a snap of a finger. Why?"

"My auditor can come and ask this anytime. Again, why? Finally you come to a state where there is a tangible value that you can derive, and then you decide this is the document I need to digitize."

That framework shows up in regulated Indian industries, where compliance pressure alone often settles the decision. Anand points to manufacturing clients that answer to multiple regulators at once. "They'll be regulated by multiple departments, starting from health and safety, pollution control, food safety," he says.

"It keeps going. Taxes, it keeps going forever. That's the intrinsic ROI they get when they digitize."

Davis-Pipe says Crown already uses AI internally to run feasibility studies on new business lines, treating it as an analyst rather than a search tool. Asked what he would fix with more advanced AI, he pointed to something outside software entirely: the physical, logistics-heavy side of the business, which he says is harder to solve than any document problem.