AI Platform Selection — Governance Framework
AI Platform Selection Decision Tree
01
The decision tree
1. Validation
2. Audience
3. Compliance
4. Data
5. Complexity
6. Governance
7. Result
STOP
REVIEW
APPROVED
flowchart LR
%% ===== Stage 1: Validation =====
start{"Is AI the
right solution?"}
start -->|"Yes"| audience
start -->|"No"| stop_not_ai([STOP: AI Not
Recommended])
start -->|"Maybe"| stop_discuss([REVIEW: Discuss
with Governance])
%% ===== Stage 2: Audience =====
audience{"Who needs access?
Filters platform options"}
audience -->|"Everyone 500+"| security
audience -->|"Team 10-100"| security
audience -->|"Individual 1-5"| security
%% ===== Stage 3: Compliance =====
security{"Data sensitivity?"}
security -->|"Low"| region_check
security -->|"Medium"| region_check
security -->|"High PII"| stop_legal([REVIEW: Legal
Review Required])
region_check{"Regional restrictions?"}
region_check -->|"None"| data_type
region_check -->|"Approved tool"| data_type
region_check -->|"Unapproved tool"| stop_works([REVIEW: Regional
Approval Required])
%% ===== Stage 4: Data =====
data_type{"What data does
the solution need?"}
data_type -->|"Search company
knowledge"| glean_check
data_type -->|"Create / edit in
Google Workspace"| google_tools
data_type -->|"External Systems"| external_check
data_type -->|"Large Files"| large_context
data_type -->|"Multiple Sources"| multiple_sources
glean_check{"Sources in Glean?
15 connectors enabled"}
glean_check -->|"Yes"| context_check
glean_check -->|"Build connector"| stop_connector([REVIEW: Evaluate
Glean Connector])
glean_check -->|"No API"| stop_no_api([STOP: No
Integration Path])
context_check{"Fits 128K tokens?"}
context_check -->|"Yes"| complexity
context_check -->|"No"| large_context
google_tools{"Google AI tool?
Write / edit / act"}
google_tools -->|"Gemini"| complexity
google_tools -->|"Gems"| complexity
google_tools -->|"NotebookLM"| complexity
google_tools -->|"None fit"| external_check
external_check{"External API?"}
external_check -->|"Yes + governance"| mulesoft_path
external_check -->|"Yes, simple"| complexity
external_check -->|"Glean Action"| complexity
external_check -->|"No API"| stop_no_api
mulesoft_path{"MuleSoft
Agent Fabric"}
mulesoft_path -->|"Proceed"| complexity
mulesoft_path -->|"Custom + logging"| complexity
multiple_sources{"All in Glean?"}
multiple_sources -->|"Yes"| complexity
multiple_sources -->|"Glean + MCP"| complexity
multiple_sources -->|"Custom orchestration"| complexity
large_context{"Large context needs
Claude filtered by
audience + German"}
large_context -->|"Gemini"| complexity
large_context -->|"NotebookLM"| complexity
large_context -->|"Claude"| complexity
large_context -->|"Custom chunking"| complexity
%% ===== Stage 5: Complexity =====
complexity{"Task complexity?"}
complexity -->|"Simple Q and A"| governance_eval
complexity -->|"Guided Workflow"| governance_eval
complexity -->|"Multi-step"| governance_eval
complexity -->|"Integration"| governance_eval
%% ===== Stage 6: Governance =====
governance_eval{"Governance level?"}
governance_eval -->|"High"| operations
governance_eval -->|"Standard"| operations
governance_eval -->|"Minimal"| operations
operations{"What if it breaks?"}
operations -->|"Builder maintains"| builder_check
operations -->|"Platform managed"| recommendation
operations -->|"Mission critical"| stop_governance([REVIEW: Governance
Review])
builder_check{"Builder available?"}
builder_check -->|"Confirmed"| recommendation
builder_check -->|"Backup needed"| stop_backup([REVIEW: Identify
Backup Owner])
builder_check -->|"Cannot meet reqs"| stop_find_owner([STOP: Find
Different Owner])
%% ===== Stage 7: Result =====
recommendation([Platform
Recommendation])
%% ===== Styles =====
classDef s1 fill:#EEF1FF,stroke:#4F46E5,stroke-width:2px
classDef s2 fill:#F5EEFD,stroke:#7C3AED,stroke-width:2px
classDef s3 fill:#FDEEEE,stroke:#DC2626,stroke-width:2px
classDef s4 fill:#E6F4FA,stroke:#0E7490,stroke-width:2px
classDef s5 fill:#EBF1FE,stroke:#2563EB,stroke-width:2px
classDef s6 fill:#FBF3E2,stroke:#D97706,stroke-width:2px
classDef stopNode fill:#FEE2E2,stroke:#DC2626,stroke-width:2px
classDef reviewNode fill:#FEF3C7,stroke:#D97706,stroke-width:2px
classDef resultNode fill:#D1FAE5,stroke:#059669,stroke-width:2.5px
class start s1
class audience s2
class security,region_check s3
class data_type,glean_check,context_check,google_tools,external_check,mulesoft_path,multiple_sources,large_context s4
class complexity s5
class governance_eval,operations,builder_check s6
class stop_not_ai,stop_no_api,stop_find_owner stopNode
class stop_discuss,stop_connector,stop_governance,stop_backup,stop_legal,stop_works reviewNode
class recommendation resultNode
02
Terminal outcomes
Stop
AI Not Recommended
Consider rules-based automation or human process instead. Low volume, high stakes, or requires strict accountability.
Stage 1 - Validation
Review
Discuss with Governance
Mixed signals suggest this needs team discussion. Document the use case and schedule a review with your governance team.
Stage 1 - Validation
Review
Legal Review Required
High-sensitivity data (PII, financial, legal) requires Legal and Governance approval before proceeding.
Stage 3 - Compliance
Review
Regional Approval Required
Tool requires regional regulatory or legal approval before use. Check applicable compliance frameworks for your jurisdiction.
Stage 3 - Compliance
Review
Evaluate Glean Connector
Required data source not in Glean but API exists. Evaluate custom connector feasibility before proceeding.
Stage 4 - Data
Stop
No Integration Path
External system has no API. Consider manual data export, vendor API request, or alternative data source.
Stage 4 - Data
Review
Additional Governance Review
Mission-critical solutions require evaluation against all five criteria: Reliability, Governance, Traceability, Auditability, Fault Tolerance.
Stage 6 - Governance
Review
Identify Backup Owner
Primary builder needs a backup owner identified. Document handoff procedures and schedule knowledge transfer.
Stage 6 - Governance
Stop
Find Different Owner
Current builder cannot meet maintenance requirements. Find different owner, reduce scope, or switch to platform-managed solution.
Stage 6 - Governance
Approved
Platform Recommendation
All checks passed. Dynamic recommendation based on audience, compliance, data type, complexity, and governance needs.
Stage 7 - Result
How to read this tree. Start at “Is AI the right solution?” on the left. Follow branches right through 7 evaluation stages. Diamond nodes are decision points. Rounded nodes are terminal outcomes -- either STOP (red), REVIEW (amber), or APPROVED (green). Use zoom controls or Ctrl/Cmd+scroll to inspect details.
Disqualifiers first. This tree front-loads audience scope (Stage 2) and compliance checks (Stage 3) before asking about data or platform selection. This ensures users never invest time exploring a platform path only to discover it is blocked by scale requirements or regional regulatory restrictions.
Disqualifiers first. This tree front-loads audience scope (Stage 2) and compliance checks (Stage 3) before asking about data or platform selection. This ensures users never invest time exploring a platform path only to discover it is blocked by scale requirements or regional regulatory restrictions.