The Business Challenge
The business situation the organisation faced before the project began.
Competitor and market research is slow and scattered across many sources, with duplicated and conflicting information. AI-generated summaries make it faster but often cannot show where a claim came from, so teams cannot audit or trust the insights they act on.
Why AI Was the Right Solution
Why AI was the right approach — and what alternatives were considered.
Manual research across many sources is slow, and keyword search cannot extract or compare claims. LLMs can read and summarise large volumes of material, but on their own they can state things without evidence. Combining AI extraction with mandatory citation links gives the speed of AI while keeping every claim verifiable.
How the Solution Was Delivered
Discovery, design, development, testing and rollout — the journey, not the tooling.
- 1Defined the core problem: AI research summaries are fast but hard to verify.
- 2Designed a pipeline to ingest sources, normalise them and remove duplicates.
- 3Added claim extraction so findings are broken into checkable statements.
- 4Linked every insight to its supporting citations.
- 5Built the interface for competitor analysis and opportunity detection.
- 6Made embeddings and live search optional so the platform runs with or without them, and packaged it with Docker.
Key Technical & Architecture Decisions
Architecture, model selection, workflow and trade-offs.
Citation-first design: no insight is stored without its supporting sources.
PostgreSQL with pgvector so structured data and semantic search live in one database.
Groq for fast LLM processing.
Optional embeddings and Brave Search live search, so the platform stays usable with minimal configuration.
React and TypeScript front end with a FastAPI back end, packaged with Docker.
Challenges & How They Were Solved
Obstacles hit along the way and how they were overcome.
Duplicate and overlapping sources: handled with normalisation and deduplication before analysis.
Unverifiable AI output: solved by requiring every insight to link to citations.
Different deployment needs: optional components and a Docker configuration keep setup flexible.
User Adoption & Change Management
How users responded — training, change management and feedback loops.
Trust is the main barrier to using AI for research. Showing the evidence behind each insight lets users check the work quickly and rely on it, rather than re-researching AI conclusions themselves.
Business Outcomes & Impact
The measurable outcomes from the underlying AI Deployment — with the story behind them.
Every generated insight is linked to supporting citations, giving fully auditable research output.
The story behind the numbers
Research produced with XENOM is auditable: anyone reviewing an insight can see exactly which sources support it. Deduplication and claim extraction turn scattered material into a structured knowledge base, which makes competitor and opportunity analysis faster to review and easier to defend when decisions are made on it.
Lessons Learned
What surprised the team, what worked well, and what would be done differently.
- Citation-first design makes AI research trustworthy.
- Deduplication should happen before analysis, not after.
- One database for structured and vector data simplifies the stack.
- Optional components keep deployment flexible.
Future Opportunities
Where this solution could go next.
Alerts when new sources change an existing insight.
Scheduled monitoring of competitors.
Confidence scoring based on the number and quality of supporting sources.
Export of evidence-backed reports.
Professional Reflections
Guidance for another builder tackling a similar problem.
When AI is used for research, design for verification from the start. Make citations mandatory, deduplicate early, and keep optional components optional so the core product stays simple to run.