CASE STUDY

How We Saved a Swiss Bank €5M

4-Week AI Vendor Due Diligence That Prevented a Costly Mistake

€8M
Contract at Risk
€5M+
Hidden Costs Found
4 Weeks
Analysis Time
€40,000
Our Fee

The Situation

Client: Top 10 Swiss Private Bank (name confidential)

Context: 2 weeks from signing an €8M, 3-year contract with a prominent AI vendor

The Vendor's Promises:

  • • Revolutionary AI-powered wealth management platform
  • • 40% efficiency gains in portfolio management
  • • "Turnkey" solution requiring minimal IT resources
  • • Full regulatory compliance out of the box
  • • ROI positive within 18 months

Red Flag: The CTO called us saying: "Something feels wrong, but I can't articulate it to the board."

Our Analysis Process

1

Week 1-2: Technical Deep Dive

  • • Reviewed 2,000+ pages of technical documentation
  • • Tested AI models with real (anonymized) client data
  • • Mapped integration requirements to existing systems
  • • Interviewed vendor's technical team
2

Week 3: Hidden Cost Discovery

  • • Found 23 "integration points" each requiring custom development
  • • Discovered need for 8 additional full-time specialists
  • • Identified €2M annual licensing for "optional" but required features
  • • Uncovered data migration costs of €800k not in original quote
3

Week 4: Performance Validation

  • • Vendor's "40% efficiency" based on perfect lab conditions
  • • Real-world testing showed 8% improvement at best
  • • Model showed significant bias against small portfolios (60% of clients)
  • • Compliance features didn't meet Swiss regulatory requirements

Critical Findings

  1. True Total Cost: €14M over 3 years (not €8M)
  2. Vendor Lock-in: Proprietary data format would cost €3M to migrate away
  3. Compliance Gaps: Model explanations didn't meet FINMA requirements
  4. Performance Reality: 8% efficiency gain vs 40% promised
  5. Hidden Dependencies: Required complete data architecture overhaul
  6. Team Requirements: 8 FTEs needed, not 2 as promised

The Outcome

Bank's Decision: Contract cancelled based on our findings

Alternative Path Taken:

  • • Built internal AI capability with €2M investment
  • • Hired 3 AI engineers (not 8 specialists)
  • • Achieved 15% efficiency improvement (better than vendor's real 8%)
  • • Maintained full control of data and models
  • • Ensured complete regulatory compliance
  • • ROI positive in 14 months

Total Savings: €5M+ in hidden costs and overruns

Key Lessons

What Banks Should Ask AI Vendors:

  • • Show me production deployments at similar-sized institutions
  • • What's the TOTAL cost including integration, training, and maintenance?
  • • Can I export my data in standard formats anytime?
  • • How does your model handle regulatory explainability?
  • • What happens when I need features not in the base product?

Red Flags to Watch For:

  • • "Proprietary" everything (data formats, APIs, models)
  • • Vague integration timelines ("typically 3-6 months")
  • • No clear data governance framework
  • • References only from much larger/smaller institutions
  • • "AI" that's really just rules engines

Is Your AI Vendor Hiding Costs?

Our 4-week due diligence process reveals what vendors don't want you to know.

Average hidden costs found: 65% above initial quote
Success rate: Saved 87% of clients from bad deals

Contact Marco Gruppo for AI Due Diligence:
marco@gruppomarco.net | calendly.com/marco-gruppomarco/30min

← Back to Home