AWS explains how Norton Revamp turns live web results into AI-written posts

Norton Revamp uses current web results to help professionals decide what to post, tying AI-generated content to their career goals. A November 3 AWS case study explains how Gen Digital built that system—and reports faster software development alongside a substantial review burden: checking AI output consumed 20–30% of the team’s time.

The disclosure concerns an existing product, not a new release. Gen says Norton Revamp launched in March 2026. The professional reputation platform combines personalized news, content coaching and goal recommendations. The account was co-written by Gen executives Iskander Sanchez-Rola and Lei Gu, alongside AWS staff.

A career goal becomes a search query

The pipeline starts when someone sets or updates a professional goal. Amazon Nova generates search queries using that goal, recent activity and engagement patterns. Amazon Bedrock AgentCore Web Search then returns article snippets, URLs, titles and publication dates. Those results supply current information for the model, rather than leaving it to rely on its training alone.

Retrieval is not the final selection step. Gen’s own system ranks results for relevance, recency and source credibility before passing them to Amazon Nova Lite. The model uses that material to generate suggestions intended to match the user’s voice. Engagement data feeds back into later query generation, influencing what the system searches for next.

Users move through an Analyze, Refine, Post coaching flow for LinkedIn and X. Search also informs weekly themes and milestones, with goal recommendations drawing on the past four weeks of activity. Gen says the personalized news feed refreshes each session, while some search results are cached for periods matched to their freshness requirements.

Managed retrieval, custom source judgment

Gen says it evaluated custom crawlers, third-party news services and static knowledge bases. Its objections were maintenance and freshness for crawlers, additional contracts and external data transfers for news services, and stale information in static repositories. It chose AWS’s managed search service to avoid operating its own retrieval infrastructure and keep queries within AWS.

AWS’s June 17 announcement already identified Norton Revamp as a user of the tool. The November account adds the implementation detail: Gen built its own credibility-scoring layer and used engagement feedback to refine retrieval. Gen says relevance improved over the first 90 days, but provides no numerical relevance score.

Faster implementation still needed human review

The team separated AI used inside the product from AI used to build it. Humans owned architecture, security reviews, business logic and personalization algorithms. AI supported code generation, documentation, test scaffolding, refactoring and interface components. With Claude, the team says it produced at least 10 design variants per concept and reduced design-to-prototype time from weeks to hours.

Those gains came with review work, not a handoff of responsibility. Gen says its security team reviewed AI-generated recommendations and retained final decisions. Testing covered search freshness, relevance to user goals, generated content and engagement tracking. Automated test generation helped the team reach 85% code coverage—the share of code exercised by tests.

Gen reports 99.9% uptime since launch and says it monitors AI content quality and user satisfaction in real time. The speed and availability figures are company-reported results in a co-authored AWS case study, rather than an independent evaluation.