Slide 01

$40M Chasing the Past. $2M Building the Future. That Is Your Technology Budget.

CTO + CFO + Board

Core claim

Week one as the new CTO. $50M technology budget. $40M — 80% — goes to legacy maintenance. $2M — 4% — goes to new products. You are spending 80% chasing the past while competitors invest 100% in the future.

Chen's pricing engine: built in 2015. 52,000 lines. Costs $5.8M annually to maintain. Eight minutes with AI coding tools: 3,400 lines of working code. Annual cost: $375K versus $5.8M. 93.5% cost reduction on one service. Ninety days later: $32M freed from a $50M budget.

The math Four months of human work → six hours with AI. Test results: works for 99.4% of cases. The 0.6% edge cases cost $15K annually in manual reviews. Legacy cost: $5.8M.

Slide 02

Eight Minutes. 3,400 Lines. 93.5% Cost Reduction on One Service.

The proof

Chen's pricing engine 52,000 lines

Built 2015. Annual maintenance cost: $5.8M. Four months of human work to regenerate. "Whatever you do, don't propose a rewrite. We tried twice. Both failed."

AI regeneration 8 minutes

3,400 lines of working code. Handles 99.4% of cases. Misses edge cases like billionaires with Bitcoin collateral. Annual cost: $375K versus $5.8M.

Edge case cost $15K/yr

Roughly 50 weird cases a year handled manually. Previous cost of automating every edge case: $5.8M. Give away one $180K car per year. Still save $5.4M.

Cost us $15K annually in manual reviews versus $5.8M in perfect automation. At the macro level, I'll give away a free $180K car once a year if it means saving $5.4M.

What the CFO heard — and then asked: "How many services can you do this to?"

Slide 03

Month One: Three Services. Month Two: Eight More. Month Three: Done. $32M Freed.

Execution

Month 1

Three services

Pricing engine: 52,000 → 3,400 lines. Cost: $5.8M → $375K. Credit decisioning: 68,000 → 4,200 lines. Cost: $4.8M → $320K. Document generation: 45,000 → 2,800 lines. Cost: $3.2M → $280K. Three services. $13.8M → $975K.

Month 2

Eight more services

Total code: 1.2M lines → 180,000 lines (15% of original). Release cycle: quarterly → weekly. Legacy cost: $40M → $12M. The VP of Engineering who said "don't propose a rewrite" is now running the blitz.

Month 3

Final four services

Final code: 1.2M lines → 120,000 lines (10% of original). Legacy cost: $40M → $8M. Budget freed: $32M (64% of total). Edge cases encountered: 187. Manually handled: 187. Cost: $8,500.

Free cars given away 2 cars. $360K cost. $800K PR value. $420K in customer referral margin from those two edge cases. Net: +$860K. Previous cost of automating those edge cases perfectly: $26.4M.

Slide 04

The Shrine Is the Organizational Belief That Legacy Code Is Irreplaceable

Root cause

What the shrine costs

"Whatever you do, don't propose a rewrite. We tried twice. Both failed." That is not institutional wisdom. That is institutional fear encoded as policy.

Two failed rewrites did not prove rewriting is impossible. They proved that human-led rewrites at that scale, with that team, using those tools, failed. AI changes what is possible in eight minutes that previously required four months.

The real cost $40M per year in maintenance is the cost of worshipping at the shrine. It is not due diligence. It is institutional attachment to decisions made in 2015.

Permission to burn it

The new CTO asked for two things: permission to burn the shrine, and backing when they gave away two free cars.

The CEO's question was "what if things break?" The answer: "Then we handle them. We give white-glove service. If we accidentally approve a weird loan, we make him a customer for life. That is still cheaper than spending $40M maintaining perfection."

The frame shift Stop optimizing for zero edge cases. Optimize for total cost of delivery, including edge case handling. The math is not close.

Slide 05

What Would You Build If You Had $32M That Spent Last Year on Legacy Maintenance?

Strategic implication

Before: $50M budget

$40M (80%) — legacy maintenance. Keeping 1.2 million lines alive. Quarterly releases. Two failed rewrites as scars.
$8M (16%) — operations, infrastructure, security.
$2M (4%) — new products. That is your innovation budget. 4%.
Release cadence: quarterly. Every major change requires months of coordination and risk management.

After: $50M budget, 90 days later

$8M (16%) — lean maintenance on 120,000 lines. Weekly releases. Edge cases handled manually for $8,500/year.
$8M (16%) — operations, infrastructure, security.
$34M (68%) — new products, competitive features, market expansion. That is your new innovation budget.
Release cadence: weekly. The coordination overhead that required quarters now requires days.

Slide 06

How Much of Your Technology Budget Is Protecting Decisions Made Before ChatGPT Existed?

Decision close

The decision on your table

You do not need to run a 90-day blitz today. You need to ask what the experiment looks like — and whether you have the permission to run it.

Pick one service. The most expensive to maintain. Run eight minutes of AI regeneration on it. If it handles 95%+ of cases, you have your proof point. Present the math the way this CTO did: total cost of legacy, total cost of regeneration, total cost of edge case handling. Let the CFO run the calculator.

The VP of Engineering who said "don't propose a rewrite" became the person running the blitz. People follow results, not proposals. Run the experiment. Show the number. Ask for 90 days.