Principles, and the capability spine.
Two lenses on what underlies the framework: the six pillars are how we build, and the capability spine (ten domains, D1 to D10) is what we assess.
Three companies, three targets. Each moves to the level its P&L needs.Three portfolio companies, three targets. What ships in one transfers to the next.
- Month-end close, Industrials: moves from Level 1 to Level 2. 14-day close, all by hand. 10-day close, numbers trusted.
- Insurance approvals, Healthcare: moves from Level 1 to Level 3. 45 min per patient, staff on the phone. 20 min, chatbot finds each insurer's rules. 4 min by agent, nurse signs off · est. $310K/yr.
- Renewal quoting, Software: moves from Level 1 to Level 4. 2 days per quote. Same day. 1 hour, agent drafts. Minutes · est. $480K/yr.
- Month-end close, Industrials portco: moves from Level 1 to Level 2. 14-day close at the portco. 10 days; the fund sees numbers a week sooner.
- Insurance approvals, Healthcare portco: moves from Level 1 to Level 3. 45 min per patient, staff on the phone. 20 min, chatbot finds each insurer's rules. 4 min by agent, nurse signs off · est. $310K/yr to EBITDA.
- Renewal quoting, Software portco: moves from Level 1 to Level 4. 2 days per quote. Same day. 1 hour, agent drafts. Minutes · est. $480K/yr · playbook to the next portco.
Three companies, three targets. Each moves to the level its P&L needs.Three portfolio companies, three targets. What ships in one transfers to the next.
- Month-end close, Industrials: moves from Level 1 to Level 2. 14-day close. 10-day close.
- Insurance approvals, Healthcare: moves from Level 1 to Level 3. 45 min each. 20 min each. 4 min · $310K/yr.
- Renewal quoting, Software: moves from Level 1 to Level 4. 2 days. Same day. 1 hour. Minutes · $480K.
- Month-end close, Industrials portco: moves from Level 1 to Level 2. 14-day close. 10-day close.
- Insurance approvals, Healthcare portco: moves from Level 1 to Level 3. 45 min each. 20 min each. 4 min · $310K/yr.
- Renewal quoting, Software portco: moves from Level 1 to Level 4. 2 days. Same day. 1 hour. Minutes · $480K.
Azure
One system, two lenses.
Together they say what a company needs, and how we hold ourselves to it, at every level.
The six pillars are operating principles, level by level, not aspirational.
The capability spine is ten domains, not ten tools. Companies need a subset of ten domains in the right order; orchestration is a progression across levels, not a tool bought once.
The capability spine: ten domains, not ten tools
Lens one · how we build
- 01Reliability-first architecture
- 02Code-first interfaces
- 03Multimodal data first-class
- 04Write-audit-publish safety
- 05Agent throughput economics
- 06Institutional knowledge as code
Lens two · what we assess
- D1Data Foundation
- D2Governance, Catalog & Identity
- D3Foundation Models & Access
- D4Copilots
- D5Retrieval & Knowledge
- D6Orchestration
- D7Agent Runtime & Durable Execution
- D8Evals & Observability
- D9AI Security & Guardrails
- D10Outcome Capture & Governance
The Level 3 (L3) column is the inflection: six domains turn core at once. That is why L3 is where the work, and the cost, compound.
What never changes, at any level.
How the lenses fit
The Pillars are how we build. The Spine is what we assess. Use the maturity ladder to place the company; use the spine to say which domains are core versus absent. Never call a retrieval layer "a level."
The non-negotiables
Two domains are never skippable: a governed, integrated data foundation (D1) and governance and identity (D2). Deploy AI on ungoverned data and you automate the error, faster.
The "no" list
Anything we can't inspect at the step level; no production observability beyond raw logs; hosted-only with no data-residency control; no SOC 2 Type II or a named path to it.
What a company actually spends to move a level includes the data foundation (D1) underneath, plus platform and run cost.
The six pillars
Operating principles, level by level, not aspirational.
Reliability-first architecture
Scale through parallelism. Atomic, composable, idempotent tools. Full trace visibility. Async-first. Source of truth in one place. "The architecture is the asset."
Code-first interfaces
Code is the universal interface of an agentic system: the only output that is simultaneously deterministic, scoreable, testable, versionable, and replayable. Reduce all problems to code. Prompts versioned like code.
Multimodal data first-class
Text, structured data, images, voice, code: all first-class inputs to retrieval and reasoning. The L1 substrate stops being "the warehouse" and becomes the multimodal source of truth the agents draw from.
Write-audit-publish safety
Gating at the edge, not human-in-the-loop. Implement alongside, not on top. Shadow runs, migrate on evidence. "Evals are not QA. They are the heartbeat."
Agent throughput economics
Don't over-scale infrastructure. Parallelize the agents. "Add an agent, not a server." Per-agent budgets. Cost attribution at L4. Cost-per-task replaces cost-per-seat as the metric.
Institutional knowledge as code
Compacted retrieval over raw injection. Reduce institutional knowledge to code (runbooks, decisions, policies) and let the agents read it. The L1 source of truth becomes the L5 colleague's long-term memory.
The ten domains
Companies need a subset of ten domains in the right order; orchestration is a progression across levels, not a tool bought once.
One trusted version of the truth?
Can we see, secure, and control who touches data and agents?
Which models, accessed how, swappable how easily?
Where do humans get help before anything runs autonomously?
How does the system ground answers in enterprise data?
Who orchestrates the work, and how autonomously?
What does long-running, failure-tolerant agent work run on?
Can we score it, trace it, and gate deploys on it?
Will it survive a security review and adversarial input?
Can BOD prove the outcome and own the loop?
How hard each domain is “on” at each level
Read across a row to see a domain’s life cycle; read down a column to see what a level demands at once. The L3 column is the inflection.
The L3 column is the inflection: six domains turn core at once. That is why L3 is where the work, and the cost, compound.
Multi-domain tools earn anchor status
A tool that covers many domains reduces integration surface, which is why the hyperscaler platforms anchor L3+. Single-domain specialists are best-of-breed and slot in alongside.
| Tool / platform | Domains covered | Anchor levels |
|---|---|---|
| Databricks (Mosaic AI, Unity Catalog, Agent Bricks, Workflows, Vector Search, Gateway) | D1, D2, D3, D5, D6, D7, D8, D9 | L3 to L5 · broadest single-platform coverage |
| Snowflake (Cortex, Horizon, Cortex Agents, Tasks) | D1, D2, D3, D5, D6, D7, D9 | L1 to L4 (analytics-first) |
| AWS (Bedrock, AgentCore, Step Functions, Guardrails, SageMaker) | D3, D6, D7, D9 (+D1 via Redshift) | L3 to L4 (AWS-anchored) |
| Claude (Anthropic) | D3, D4 (Claude Code), D6 (Agent SDK) | L2 to L5 |
| LangGraph + LangSmith | D6, D7, D8 | L4 |
| LlamaIndex | D5, D6 | L3 |
| Braintrust · Langfuse | D8 | L2 to L5 (eval loop · OSS traces) |
| Temporal | D6, D7 (durable exec) | L3 to L5 |
| Lakera | D9 | L3+ |
Common questions.
What are the six pillars of the Blue Orange framework?
The six pillars are how we build: reliability-first architecture, code-first interfaces, multimodal data first-class, write-audit-publish safety, agent throughput economics, and institutional knowledge as code. They are operating principles, level by level, not aspirational.
What is the capability spine?
The capability spine is what we assess: ten domains, D1 to D10, not ten tools, from Data Foundation (D1) and Governance, Catalog & Identity (D2) through Orchestration (D6), Evals & Observability (D8) and Outcome Capture & Governance (D10). Companies need a subset of ten domains in the right order; orchestration is a progression across levels, not a tool bought once.
Which AI capabilities can a company never skip?
Two domains are never skippable: a governed, integrated data foundation (D1) and governance and identity (D2). Deploy AI on ungoverned data and you automate the error, faster.
Why is Level 3 the inflection point in the framework?
The Level 3 (L3) column of the activation matrix is the inflection: six domains turn core at once. That is why L3 is where the work, and the cost, compound. Read across a row to see a domain's life cycle; read down a column to see what a level demands at once.
What tools will Blue Orange not build on?
Anything we can't inspect at the step level; anything with no production observability beyond raw logs; hosted-only tools with no data-residency control; and anything without SOC 2 Type II or a named path to it.
Does Blue Orange build on Databricks and Claude?
Yes. We build on Claude and Databricks. In our tool coverage, Databricks (Mosaic AI, Unity Catalog, Agent Bricks, Workflows, Vector Search, Gateway) covers eight of the ten domains, the broadest single-platform coverage, anchoring Level 3 (L3, where software runs a whole workflow and a person reviews the exceptions) to L5. Claude covers models and access, copilots (Claude Code) and orchestration (Agent SDK), L2 to L5.