AI DUE DILIGENCE

Defensible by design.

AI Due Diligence on every vendor in the database. Classification, dependency risk, data sovereignty, and three trust scores — the evidence your CIO, CISO, and board need to say yes.

30/50
AI DUE DILIGENCE

Salesforce

AI-AugmentedMedium risk
12345
1Classification
6/10
2Model Independence
5/10
3Learning Loop
5/10
4Data Sovereignty
8/10
5Dependency
6/10
WHY IT MATTERS

Demos answer "what."
Due Diligence answers "what if."

The architecture questions — who owns the model, what breaks under failure, where data flows, who maintains the AI — don't fit in a slide deck. That's where purchases go wrong. DD makes those questions first-class.

WHEN
Your provider changes terms
DD SURFACES
Dependency risk
Reveals how exposed you are to a single API partner — before contract, not after.
WHEN
Data crosses a border
DD SURFACES
Data sovereignty
Shows the data routing path up-front: what stays, what leaves, what can be contained.
WHEN
A model update breaks something
DD SURFACES
Learning Loop
Tells you the retraining posture: continuous, periodic, or frozen — and how much customer-side work each implies.
30/50
AI DUE DILIGENCE

ServiceNow

AI-AugmentedMedium risk
12345
1Classification
6/10
2Model Independence
6/10
3Learning Loop
5/10
4Data Sovereignty
7/10
5Dependency
6/10
THE QUALITATIVE DIMENSIONS

Four categorical judgments per vendor.

Four qualitative dimensions capture an AI vendor's architectural posture. Each has a defined rubric with fixed values; each links back to public evidence. Shape at right: ServiceNow — an ai-augmented workflow platform with hybrid model ownership and medium dependency risk.

  • Classification: AI-native / AI-augmented / AI-overlay
  • Dependency risk: low / medium / high with concrete detail
  • Model ownership: proprietary / fine-tuned / API-dependent / hybrid
  • Data moat: very-high / high / medium / low with description
See rubric
THE QUANTITATIVE TRUST SCORES

Three axes, one radar, a defensible shape.

Beyond the categorical dimensions, every profiled vendor gets three 1–10 scores. The three together produce the radar — a shape, not a total. Anthropic at right: high Independence and Sovereignty, deliberately lower Learning Loop because they don't train on customer API calls. That differential is the DD story a single score would hide.

  • Model Independence — independence from third-party LLMs
  • Learning Loop — how the product improves from usage
  • Data Sovereignty — customer control over data path
46/50
AI DUE DILIGENCE

Anthropic

AI-NativeLow risk
12345
1Classification
10/10
2Model Independence
9/10
3Learning Loop
8/10
4Data Sovereignty
9/10
5Dependency
10/10

Frequently asked

Comparisons tell you what a vendor says it does. DD tells you whether its AI story holds up — who owns the model, what happens if their provider changes, where your data goes.
How we score this