The Proof Is in the P&L.
+15pts. Net margin expansion. From 60% to 75%.
Every metric on this page comes from Trilogy — a real company with real customers, real code, and real revenue. Nothing hypothetical. Nothing projected.
The Numbers That Changed the Argument
Same people, four times the output.
from 60% to 75%
Infrastructure spend cut by a factor of six.
Customers answered faster, not worse.
Running three SaaS brands and $110M in revenue.
The cleanest the codebase has been in company history.
How 40 People Outperform Companies 10x Their Size
Where They Started
Trilogy was a successful software company with $110M in revenue, but it carried the weight of decades: a bloated codebase, manual processes, and a team stretched thin across three SaaS brands. AI was being discussed, but no one was deploying it systematically. The CEO decided that would change — starting with herself.
The Turning Point
In April 2023, the CEO launched Weds.ai — a weekly, company-wide AI hackathon where every employee builds with AI. Not a training program. Not an optional workshop. A mandatory, hands-on session every single week. The message was clear: AI is not a department initiative. It is how this company operates now.
What They Implemented
Weds.ai
Weekly AI hackathons — 52 sessions per year — that turned every employee into an AI practitioner. Ideas from these sessions became production systems.
Second Brains
AI assistants trained on institutional knowledge, replacing tribal expertise that lived in people's heads. When someone left, the knowledge stayed.
AI-Native Support
Customer support rebuilt from scratch with AI handling 80% of inbound volume. Faster responses, maintained quality, freed the human team for complex escalations.
Codebase Compression
3 million lines of legacy code compressed to 125,000 through AI-assisted rewrites. Cleaner, faster, and more maintainable — with the lowest bug count in company history.
AI-Augmented Exec Ops
Board prep, investor updates, and strategic analysis shifted to AI-drafted, human-reviewed workflows. Weeks of assembly reduced to hours.
What Surprised Them
Culture got stronger.
Rather than replacing people, AI gave the team leverage. Employees who embraced it became dramatically more productive. Morale improved because people stopped doing repetitive work.
Compounding was faster than expected.
Each AI system made the next one easier to build. Second Brains fed into support automation. Support automation surfaced product insights. The flywheel effect accelerated with every deployment.
Biggest resistance came from middle management, not the bottom.
Individual contributors adapted quickly. The friction came from managers whose roles were built around coordination — exactly the work AI replaced first.
Before AI vs. After AI
| Dimension | Before | After |
|---|---|---|
| AI usage | Optional, ad hoc experimentation | Company-wide, daily, non-negotiable |
| Revenue / headcount | Baseline | 4x baseline |
| Codebase | 3 million lines of legacy code | 125,000 lines, AI-rewritten |
| AWS costs | Baseline infrastructure spend | 1/6th of original spend |
| Support | Fully staffed human support team | 80% handled by AI agents |
| Board prep | Weeks of manual assembly | AI-drafted, human-reviewed |
| Product specs | Manual requirements gathering | AI-generated first drafts in hours |
| Bug backlog | Growing year over year | 34-year low |
| Team size | ~40 employees | ~40 employees |
| Approach | Pilot programs and committees | CEO-led, full-company adoption |
From the Inside Out.
Total consultants hired: zero. Total committees formed: zero.
Department by Department
Engineering
AI rewrote and compressed 3 million lines of legacy code down to 125,000 lines. The remaining code is cleaner, faster, and more maintainable than anything the team had shipped in three decades. Bug counts hit historic lows. Deployment speed increased dramatically.
Support
Customer support was rebuilt from scratch with AI handling the majority of inbound volume. Response times dropped. Customer satisfaction held steady or improved. The human support team shifted from answering repetitive questions to handling complex escalations that actually require judgment.
Product
Product specifications that used to take weeks of cross-functional meetings are now AI-generated first drafts reviewed and refined by humans. The product team spends less time documenting and more time deciding. Iteration cycles shortened from months to weeks.
Executive
Board presentations, investor updates, and strategic analyses are now AI-drafted and human-reviewed. The executive team spends less time assembling information and more time acting on it. Decision-making accelerated across the entire leadership layer.
These Results Were Not Inevitable. They Were Led.
Every number on this page traces back to one decision: a CEO who stopped waiting for consensus and started building.