Inspiration

I'm currently doing my MBA, and we regularly see senior executives come to our class as guest speakers. Through dozens of case studies and conversations with these leaders, it is clear that the difference between good and great executives is the ability to make high-quality decisions at speed.

The rate of decision-making is often the difference between a company that thrives and one that falls behind. When we saw this gap in executives waiting weeks for siloed analysis before they could act, we knew it was a tool worth building.

YouTube Video Link (3min pitch and demo)

https://youtu.be/fg36D14WesY

What it does

Every major CEO decision, such as margin recovery, market entry, and capital allocation, requires weeks of siloed analysis from the CFO, COO, and strategy team. CEO OS collapses that into minutes.

It runs a six-stage AI workflow that plans the investigation, gathers evidence, executes parallel financial, operational, and strategic analyses, then synthesises a board-ready memo with the top three decisions, a 90-day action plan, and the trade-offs between functions.

How we built it

We started by ideating with Claude as a sparring partner to write the product spec. We validated it with a real executive, whose feedback led me to add the Concept Navigator and then use Claude Code to build the backend (FastAPI + 6-stage pipeline) and frontend (single HTML file) in one shot. After integrating Z.ai's GLM-5.1 APIs, we aligned the final submission with the hackathon's five judging criteria.

Challenges we ran into

  1. API latency: Heavier analyst stages sometimes hung on Z.ai's servers. Solved with a 90-second timeout, automatic retry, and a cached demo fallback.
  2. API credits: Initial "insufficient balance" errors required figuring out that the coding package uses a different endpoint — a workaround that cost build time.
  3. Depth vs. time: With more executive feedback, the prompts and output could be richer. But this is a meaningful step up from my previous work, and I'm heading in the right direction.

Accomplishments that we're proud of

We managed to complete a high-quality submission for the hackathon without any delay

What we learned

  1. We learned to build with Z.ai's GLM-5.1 and orchestrate a multi-stage agent pipeline.
  2. We also learned to use AI tools in concert with Claude for ideation and code, Perplexity for research, Z.ai for the core product. Finally, this submission was significantly more advanced than my previous hackathon, and shipping it gave me real confidence as a developer.

What's next for CEO OS - Decision Intelligence

  1. Enable PDF and document upload so CEOs can feed in full annual reports, board packs, and earnings transcripts directly — eliminating manual copy-paste and unlocking GLM-5.1's 200K context window.
  2. Build connections with Slack, Oracle ERP, Salesforce, and other business tools so that the data CEO OS receives is accurate and up to date.
  3. Customise the tool so that the CFO, COO, and other members of the management team can build their own decision intelligence for their respective business functions and take high-quality, fast decisions at scale.

Built With

  • claudecode
  • github
  • mobbin
  • netlify
  • perplexity
  • python
  • railway
  • vercel
  • vscode
  • z.ai
Share this project:

Updates