Rewriting the migration playbook…with a little help from a friend
How do you shift a hodgepodge legacy library of 3,000 knowledge base articles into a new ITSM system without the manual grind? (Hint: AI.)
THE SITUATION
In the early days of the telephone, you couldn’t just call someone—you spoke to an operator, and they’d manually connect your line to the line of the person you wanted to reach. The new technology was better than its predecessor, but it was still time- and labor-intensive, so for a while it was hard to scale.
Fast-forward a century and a half, and a similar dynamic was at play at a global health insurance provider. The company was migrating to ServiceNow as part of an IT service management transformation and needed to bring along its knowledge base of some 3,000 articles too. While the always-on digital library was better than its predecessor, articles were still written and maintained manually and spanned an organic hodgepodge of platforms and file types. This library—like many important but non-urgent tasks—rarely received consistent attention; managing and accessing it had become more time- and labor-intensive than it needed to be, while the lack of standardization made it hard to scale.
The good news was that the ServiceNow implementation would provide standardization and scale; the less good news was that each article would first need to be normalized to fit its requirements. That meant aligning the different ways of handling article fields (like title, author, or category) to ensure usability. It meant mapping disparate article types (like FAQs, policies, or how-tos) to distinct templates to affirm proper categorization. Applying consistent HTML formatting to boost article readability and user experience. Adding summaries and keywords to improve library search and discoverability.
And to fit the new system, each of those 3,000 articles needed something slightly different.
Cold comfort to the company’s operations leaders, but this is a common challenge; many large organizations using cloud-based platforms face similar difficulties with comprehensive, accessible knowledge repositories. But just as automated switchboards meant no more operators, Generative artificial intelligence (GenAI) can mean automating traditionally manual processes into automated, scalable operations.
Enter Deloitte’s ServiceNow-certified AI Team.
CLEANER ARTICLES. BETTER SEARCH. FASTER SUPPORT. HAPPIER TEAMS.
THE SOLVE
ServiceNow does offer an AI-enabled capability for knowledge articles, but in practice it’s limited—and it wasn’t built to handle the kind of high-volume article migration in question. Given the scale of the move and the constraints of the target environment, the team recommended Sage, Deloitte’s purpose-built knowledge-article AI agent, to carry this workflow end to end and help ensure the migration is both efficient and consistent.
Once deployed on the company’s cloud, Sage used GenAI to automatically process thousands of legacy articles, then modify them. The agent analyzed and assessed each article’s content, in some cases rewrote portions, then placed them into the proper article-type template. But that’s not all. Sage standardized formatting by converting all articles to HTML, and improved search functionality and discoverability by generating new metadata—unique summaries and assigned keywords—for each article. And following behind for quality control: an additional custom-built AI validation agent, Valerie, vetting Sage’s work before passing it to the humans on the client side for final review.
The kicker? Estimates projected that if the company migrated these articles using the old, manual approaches, they’d take 60–90 minutes each.
In real life and in real time, Sage came in at under three minutes each.
BUSINESS-CRITICAL MIGRATION ACCELERATED 30X WITH AI-POWERED PRECISION
THE IMPACT
Fast isn’t useful if it’s sloppy, but more than 98% of articles processed cleanly on the first pass, which meant the company could trust what was being migrated, with the accelerated processing time reducing months of manual work into something that could (and did) get done. In short order, the people accountable for knowledge base articles could work on things more impactful than copying, pasting, and migrating content all day.
The difference for employee “customers” was immediate, too: They could find what they needed quickly, instead of hunting and pecking the old way. The AI-generated summaries and keywords made articles clearer and more concise, so people spent less time searching and more time solving problems—the kind of improvement that shows up in daily work, with small moments saved adding up across the whole organization.
But the real payoff came from how the library connected to the rest of the organization’s operations once it was in ServiceNow. IT support teams, for instance, could now pull up relevant knowledge articles right in the middle of working a case—something they couldn’t do before. And because the solution integrated with the company’s existing Azure setup, it created a foundation that could scale. More articles in the future? Not a problem. No need to proportionally increase the manual effort.
And no need to call an operator.
AN AI ASSIST TRANSFORMS MONTHS OF MANUAL WORK INTO DAYS
