How a Bank Used AI to Flag Outdated Knowledge Base Content

The Challenge

The bank’s internal knowledge base held the procedures, policies and help content that people across the organisation relied on. Knowledge management, internal communications, IT and operations teams all had a stake in it, and much of its value lay in the items that were used most often.

Those frequently used items were also the ones that could quietly lose their accuracy. A procedure might change, a policy might be revised or a help article might no longer match how things were done, while the old version stayed in place. Users could then be pointed in the wrong direction, and the quality of self-service suffered. Without a consistent check, it was hard to tell which of these items had fallen out of date.

The Solution

With cognipeer Studio, our team configured an AI agent that reviewed the content in the bank’s knowledge base for signs that it might be out of date. It analysed each item’s last update date, how often it was accessed, the feedback users had left and signals that it conflicted with other content.

Based on these signals, the agent identified the items that needed updating and created a revision request for the relevant content owner. Owners received a request tied to a specific item and decided what to change themselves. The content stayed under the control of the people responsible for it, while the agent brought the items that needed revision to their attention.

Implementation

1. Chose the signals to analyse. We set out the indicators that pointed to outdated content: last update date, access frequency, user feedback and conflicts with other content.
2. Configured content analysis. The agent assessed knowledge base items against these signals to identify those in need of revision.
3. Linked items to owners. Each item flagged for updating was connected to the content owner responsible for it.
4. Raised revision requests. The agent created a revision request for the relevant owner, who reviewed the content and made the update.

Business Impact

The knowledge base became more current. Outdated procedures, policies and help content were identified through a consistent set of signals, and the owners responsible were asked to revise them. Users were less likely to be sent in the wrong direction by old information, and the quality of self-service improved for the people who turned to the knowledge base for answers.

• More up-to-date institutional knowledge across procedures, policies and help content.
• Fewer cases of users being misdirected by outdated information.
• Better self-service quality for the people using the knowledge base.
• Revisions made by content owners, prompted by item-specific requests.

Technology Behind the Solution