Inspiration
Small businesses often depend on one person who holds critical information in their head, messages, voice notes, and documents. When that person becomes unavailable, colleagues may not know which commitments matter, what was actually agreed upon, or which actions they are authorized to take. We wanted to explore how AI could help preserve that context and make business continuity safer.
What we're building
Siigo — STILL HERE is a business continuity system that transforms business evidence into actionable handoffs. It extracts operational claims from source material, checks quoted text against the original evidence, organizes commitments into a safe handoff, and applies policy checks to proposed actions. It also distinguishes machine-verified source text from commitments that still need human confirmation.
The project includes an Alexa-style simulated experience: an Alexa-formatted intent is sent to our own backend, which processes the request and returns a speech-ready response. This demonstrates a voice-assistant workflow without claiming a live Alexa+ service or physical Echo integration.
How we built it
We built a Python backend with FastAPI, a data layer for business evidence and operational records, a handoff compiler, a policy engine, and an interactive Next.js frontend. The workflow connects evidence ingestion and quote verification to continuity planning, financial authorization checks, and an audit trail. We also developed automated end-to-end, adversarial, extraction, and evaluation tests.
Challenges we faced
One of our biggest challenges was making the system distinguish a plausible AI-generated claim from text that can actually be traced to source evidence. We worked on exact quote verification, hash-based evidence integrity, actionable handoff classification, and explicit coverage gaps.
Another challenge was making the system demonstrate controlled autonomy rather than unrestricted automation. Our demo tests a ₹65,000 request against a ₹25,000 policy limit and contrasts it with a ₹18,000 action that passes the configured policy checks. The latter is recorded by the simulation; it is not a real bank transfer.
What we learned
We learned that useful AI for business operations requires more than generating a convincing summary. It needs traceable evidence, clear uncertainty, human oversight, explicit permissions, and predictable failure behavior. Our goal is not to replace the absent business owner, but to help the next person understand what is known, what needs attention, and what they are allowed to do.
Siigo (STILL HERE): Even when a key person is unavailable, the business can keep moving safely.
Log in or sign up for Devpost to join the conversation.