Inside Cloudflare’s fix for AI’s data problem

This article summarizes an episode of AI Engineer’s video series featuring Justin Joyce, manager of Cloudflare.

Image credit: Shutterstock

The real value of AI agents is capturing a company’s undocumented knowledge and enforcing it in a reusable format. Justin Joyce, a strategy manager at Cloudflare, warns that standard processes fail when important information gets stuck in manual reports.

To fix these slowdowns, companies must treat data as a product, help beginners perform like experts, and set controls to prevent errors.

Manual data silos choke business velocity

Without a structured product, unstructured information forces experts into a cycle of administrative work while managers wait for outdated answers.

This reliance on spreadsheets degrades efficiency across several areas:

• The expert gap: Inconsistent preparation between novice and top salespeople leads to unpredictable revenue forecasts and an uneven customer experience.
• Hidden costs: Shifting regional cases into spreadsheets masks the operational expenses of running the business.
• Execution delays: Sales representatives waste time searching for account details instead of making customer decisions.

Structured workflows standardize sales intelligence

To eliminate execution delays and treat data as a product, teams must adopt the following delivery mechanisms:

• Convert rules into reusable analysis. Translate business questions into searchable files for non-technical employees.
• Push weekly insights directly to leaders. Eliminate the need for executives to hunt for updates in dashboards.
• Deliver self-service context. Provide salespeople with a digital space for account background.

Unifying these delivery methods ensures sales and marketing teams operate from the same performance data without friction. Joyce explains that “with self-service, you allow [teams] to pull data when they need it,” while pushing insights directly to leaders.

Centralized instruction control prevents automated mistakes

While self-service systems accelerate access to intelligence, AI agents will automate errors faster than humans without a control layer.

Companies must translate institutional knowledge into skill files by establishing the following oversight protocols:

• Attach definitions to fields so models interpret every request the same way.
• Program the system to answer inquiries automatically before escalating to workers.
• Establish approval processes before releasing AI instructions to the company.
• Combine overlapping commands to prevent models from offering conflicting advice.

AI offers an operational advantage by distributing knowledge, but only if an organization exercises discipline over how that information is stored, reviewed, and deployed.