Enhancing Digital Fund Discovery Through Intelligent Search
Smart Site Search and Advisory Discovery Architecture for a Leading Asset Management Company (AMC)
A Search Experience that Inhibited Fund Discovery
The AMC website relied on legacy keyword-based search that created severe friction across four distinct dimensions, reducing session duration and suppressing digital fund discovery:
Original search field accepting only exact fund names or ticker codes.
- Keyword Inflexibility: The search field only processed inputs that perfectly matched a fund's formal registered name or ticker symbol.
- Zero-Result Abandonment: Investors searching with natural language or broad asset classes were met with an abrupt "No content found" message.
- Discovery Wall: Without fuzzy matching or synonym resolution, non-expert retail buyers could not locate funds they intended to evaluate.
Conversational query returning generic results with no contextual answers.
- Unparsed Advisory Intent: Goal-oriented questions like "What is the right age to start investing?" failed to produce relevant educational guidance.
- Absence of Synthesis: The engine could not extract key takeaways or summarize fund prospectuses, returning blank screens instead of answers.
- Tax Planning Blindspots: Queries about 80C tax-saving funds defaulted to disconnected lists without explaining qualifying conditions.
Critical investor guides, scheme documents, and educational blogs remained undiscovered.
- Isolated CMS Repositories: Scheme brochures, fact sheets, fund detail pages, and educational blogs lived in segregated silos.
- Disjointed Query Routing: Searching for a specific fund topic failed to surface the corresponding PDF guide or beginner article.
- Underutilized Collateral: High-value compliance and advisory collateral generated by research teams received minimal digital circulation.
High bounce rates caused by an unguided, transactional search flow.
- Session Dead Ends: The purely transactional search flow abruptly terminated user sessions upon query failure with no recovery path.
- Context Evaporation: Moving across pages wiped previous search queries and filters, forcing visitors to start over.
- Missed Investor Conversion: The platform had no mechanism to nudge high-intent visitors toward account registration or advisor consultation.
Transitioning to Intelligent Advisory Discovery
Powered by Origamis AI DocuAgent (Generative AI + RAG), the discovery engine was structured around four foundational pillars to drive engagement, maximize content ROI, and generate qualified investor leads:
New "Ask Anything..." input providing direct 2–3 sentence answers before listing links.
- Conversational Reframing: Replacing passive "Search..." with "Ask anything..." transforms a database lookup into an advisory dialog.
- Synthesized Direct Answers: Every query produces a concise 2–3 sentence generative summary directly addressing the investor's core question.
- Confidence First: Retail investors receive clear explanations of investment concepts before navigating into detailed product tables.
Categorized link buckets: Documents (PDFs), Webpages (Fund Details), and Blogs (Education).
- Tri-Category Partitioning: Results are systematically categorized into Documents (PDFs), Webpages (Fund Details), and Blogs (Insights).
- Curated Link Density: Each category displays 3–5 targeted links, eliminating cognitive overload while providing depth.
- Multi-Format Discovery: Enables investors to download compliance documents or read beginner guides without leaving the search flow.
Persistent chat session drawer maintaining context as users navigate between AMC pages.
- Multi-Turn Memory: The search transforms into an unobtrusive drawer that stays open as users navigate across funds and calculators.
- Session Context Continuity: The LLM retains past queries and visited pages to provide personalized, cumulative advisory insights.
- Seamless Minimization: Users can collapse and expand the drawer at any point without losing their research thread.
Timed CTA prompts spaced to every 3rd answer to prevent user fatigue.
- Cadence-Controlled Prompts: Registration and login nudges are strictly spaced to every third answer, establishing value before asking for action.
- Contextual Value Proposition: CTAs highlight personalized portfolio tracking and saved comparison lists rather than generic sign-ups.
- Intelligent Fallback: When an exact match is missing, the system suggests related investment strategies rather than an empty state.
Before and After: Interface and Discovery Breakdown
A step-by-step examination of the interface progression from a legacy search modal to an intelligent, multi-format discovery surface:
The Original Search Modal
The baseline experience: keyword-dependent with no guidance.
The baseline experience: keyword-dependent, with no guidance, suggested queries, or conversational intelligence.
The Exact Match Trap
Zero results returned for natural language conversational queries.
Demonstration of zero-result state from a natural language query on the original keyword system.
The Dead End
Unrecoverable empty state showing "No content found".
Unrecoverable empty state showing "No content found", leaving users stranded with no recovery path.
"Ask Anything..." Advisory Modal
Natural language input prompt and contextual suggestion chips.
Redesigned search modal with natural language input prompt and contextual suggestion chips for common investment questions.
Direct Contextual Answer Generation
Generative LLM synthesis delivering a 2–3 sentence direct answer.
Generative LLM synthesis delivering a 2–3 sentence direct answer addressing the user's specific query.
Three-Category Classified Results
Partitioned knowledge: Documents, Webpages, and Blogs.
Partitioned knowledge results categorized into Documents (PDF guides), Webpages (Fund Details), and Blogs (Educational Insights).
Persistent Chat Session Drawer
Persistent side-drawer maintaining multi-turn context across pages.
Persistent side-drawer maintaining multi-turn context and visited pages as users browse across AMC pages.
Strategic CTA Nudge & Fallback Recovery
Timed register/login prompt spaced to every third answer.
Timed register/login prompt spaced to every third answer, paired with intelligent best-matched fallback suggestions when exact matches are unavailable.
The redesign delivered verified improvements across discovery success, content engagement, and digital lead generation within the first release cycle:
Search Success Rate
Users successfully finding relevant funds, documents, or advisory answers in a single search session.
Search-to-Product View Rate
Users transitioning directly from search answers to detailed fund product pages and SIP calculators.
Bounce Rate from Search
Significant reduction in immediate visitor drop-offs following conversational investment queries.
Investor Resource Downloads
Three-fold increase in downloads of investor guides, fund fact sheets, and educational market blogs.
The Power of Framing ("Ask Anything...")
Changing the search field from a passive "Search..." to "Ask anything..." was a subtle but fundamental UX intervention that guided retail investors toward natural conversational inquiry, instantly lowering the barrier to entry.
Search is an Advisory Surface, Not a Utility
By treating the search results modal as an advisory discovery surface rather than a mechanical index, we transformed user intent into high-intent engagement, positioning the brand as a trusted wealth partner.
Designing the Failure State First
The greatest impact came from solving the worst user experience: the dead-end empty state. By ensuring every outcome included direct fallback recommendations, best-matched collateral, and guided next steps, we eliminated drop-offs.
Cadence Beats Aggression in Conversion
Lead generation nudges only work when trust is established. Spacing registration prompts to every third answer allowed visitors to experience tangible value first, maximizing conversion while preventing visitor fatigue.