Inspiration Enterprises spend millions annually on competitive intelligence teams—manually scanning news, sites, earnings calls, and social media to avoid surprises. It's slow, reactive, and inconsistent. We're replacing that with an AI system that runs 24/7 and never forgets.
What it does Generates high-fidelity, executive-ready reports on competitors—tracking moves, messaging shifts, partnerships, launches, and white space. Each report identifies strategic strengths and areas where your brand can differentiate or go on offense. All external data; no internal integration required.
How we built it
Multi-agent architecture: An orchestrator agent (Opus via AWS Bedrock) decomposes the research plan.
Sub-agents handle vertical tasks—Tavily for targeted web search, Gemini for summarizing long-context documents like press releases or transcripts.
Fully model-agnostic. Agents collaborate and hand off work across different LLMs—proving interop works in real-world pipelines.
Challenges
Context limits: stitching long docs, handling multi-step reasoning over fragmented sources.
Structuring unstructured data from the web into clean, analyzable evidence.
Avoiding shallow summaries—forcing the system to generate true insight, not just paraphrase.
Accomplishments Built a working system that delivers structured, accurate, and often non-obvious insights. It's slow, but reliable—proving the multi-agent setup works end to end.
What we learned When coordinated well, multi-agent systems outperform monolithic prompts—especially for large, complex, reasoning-heavy tasks.
What's next We're turning this into a continuously running background task—auto-generating competitive briefings before exec meetings, board decks, or strategy offsites. Ultimately: an always-on AI analyst who knows your market better than your team does.
Log in or sign up for Devpost to join the conversation.