Build an AI-Native Go-to-Market Platform
Topcoder
📌 About the Event
AI-native go-to-market platform the kind of system that’s emerging as a new category in B2B software. The thesis is simple and well-documented in the market: the average sales rep spends roughly 30% of their time actually selling. The other 70% disappears into account research, contact research, CRM updates, call prep, note-taking, follow-ups, forecast updates, quoting, and context switching across 15+ disconnected tools. The category we’re entering rejects that status quo. Instead of being yet another tool a rep has to log into, the platform we’re building is the operating layer where sales work happens — with AI agents doing real work alongside the human, on top of a unified data model that knows everything about every account, contact, signal, and conversation.
ℹ️ Why this challenge exists
This is the first challenge in a series that will build out this platform over time. That framing matters enormously for how you should approach it.
We are not asking you to build a demo. We are not asking you to build a polished front-end with mocked API responses behind it. We are not asking you to wrap an LLM call in a chat UI and call it a CRM. Submissions that take a spec-to-AI-tool shortcut tend to be obvious to judges, and they will not be competitive here.
What we are asking you to build is the foundation on which everything else gets built. Think of yourself as the founding engineer and founding product architect on day one of a startup that intends to compete in this category. Every decision you make in this challenge – your data model, your service boundaries, your agent orchestration approach, your integration framework, your extensibility model – will either accelerate or strangle every challenge that comes after this one. The winning submission is the one we can keep building on for the next 12 months without ripping it out.
🚀Requirements
What this iteration must deliver
This first iteration has two parts, and both have to be real:
1. The platform backbone. A unified data model designed for GTM data (accounts, contacts, opportunities, activities, signals – at minimum), real persistence (not in-memory, not JSON files), a multi-tenant-ready foundation, an ingestion pathway for external data, and an agent orchestration layer capable of running real LLM-driven workflows with tool use, structured outputs, and traceability. This is the layer that makes us a platform instead of an app.
2. One vertical slice, built deep: Prospect. A complete, working Prospect experience that exercises the backbone end-to-end – sourcing accounts that match an ICP, enriching contacts, capturing intent signals, generating context-aware value hypotheses, and producing personalised outreach. The Prospect slice is not a side feature; it is the proof that the backbone is real. If the backbone is well-designed, building Prospect on top of it should feel natural – and the next challenge (adding Engage, then Manage, then Operate) should feel like extending the same system, not starting over.