Future CoS (Chief of Staff)

Your agentic generative UI— always prepared, always insightful. Helps founders and the C‑suite accelerate strategic decision‑making, built on Agentic Loop, RAG, System Thinking, HIL Governance.

Comment

Inspiration

Startup founders, CEOs, and C-suite executives are constantly overwhelmed by operational noise. They spend hours hopping between fragmented tools—checking bank accounts, active CRM sales pipelines, calendar appointments, competitor pricing pages, and team task boards. Future CoS was inspired by the need for a unified AI Chief of Staff that acts as a proactive partner. It runs background loops to synthesize cross-functional data, detects anomalies in real-time, models the mathematical impact of financial decisions, and outputs clean strategic logs.

What it does

Future CoS is an executive AI copilot that unifies workspace management, scenario modeling, and business reporting into a single text-driven canvas:

  • Executive Daily Briefing: Compiles cash runway, sales pipelines, and calendar meetings into a morning summary, highlighting critical alerts.
  • Scenario Simulator ("What-If" Analysis): Evaluates corporate pivots (e.g., hiring headcount, pricing shifts) and calculates exact runway impacts over 12 months in a secure sandbox.
  • Secure Document Vault & RAG: Allows secure uploads of pitch decks and sheets, enabling executives to query internal knowledge bases.
  • Market Intel Digest: Monograms competitor pricing changes and caches competitor data in Firestore to minimize search token waste.
  • Board & Investor Reporting: Automatically compiles board-level performance reports and downloads them as clean Markdown files.
  • Decision Logs & Strategic Ledger: Records strategic choices with documented rationale, confidence scores, and assumptions on an immutable, SHA-256 styled ledger.

How we built it

We developed the application as a distributed hybrid-runtime system:

  • Core Frontend Canvas: React and Next.js, implementing the Sleek design system tokens (60% surface background, 30% near-black text, 10% accent blue). It utilizes Inter for typography and maintains a strict metronomic 8pt spacing grid.
  • AI Agent Engine: FastAPI and Python LangGraph, driven by Gemini 2.5 Flash for high-speed multi-tool reasoning, planning, and self-reflection loops.
  • Data Layer: Dual-mode storage using Firebase Firestore for production state synchronization, alongside automated local fixture databases for offline local development.
  • Deployment: Automated Docker builds pushed and deployed as parallel containerized services on Google Cloud Run.

Challenges we ran into

  • Local Emulator Dependencies: Operating system constraints on local Java JRE prevented starting the Firestore emulator locally. We resolved this by building a dual-mode Firebase helper that connects to the emulator in dev mode while providing mock fallback fixtures.
  • Markdown Clutter in Generative UI: The LLM agent initially outputted redundant markdown descriptions alongside visual cards. We fixed this by introducing Rule 14 to the system prompts, forcing single-sentence intros.
  • Asymmetric Bento Grid Alignment: Sizing 5 pilar cards created uneven rows. We introduced a 6th pillar (Secure Document Vault) and reformatted the grid sizes (Row 1: 2+1, Row 2: 1+1+1, Row 3: 3 full-width) to achieve perfect layout symmetry.

Accomplishments that we're proud of

  • Dynamic Generative UI: Interactive task lists, download cards, and pricing feeds render dynamically in the chat stream.
  • High-Speed Compilation: The entire frontend codebase compiles with 100% type safety.
  • Production Deployment: Parallel backend and frontend container services run smoothly on GCP Cloud Run.

What we learned

  • Visual Precision: Executive tools require minimal layouts. Adhering to the Sleek 60-30-10 discipline reduces cognitive load and aids fast reading.
  • State Alignment: Dynamic card actions (like checkbox toggles or date adjustments) must map directly back to database records to ensure consistency.

What's next for Future CoS (Chief of Staff)

  • Live Bank & CRM Integration: Connecting directly to Plaid and HubSpot APIs for automated real-time ingestion.
  • Multiplayer Boardroom Ledgers: Enabling co-founders to review, comment, and sign off on decision logs.

Built With

Share this project:

Updates