Skip to main content

Framework Authoring

Framework authoring lets your firm build its own assessment frameworks — encoding your proprietary methodology as a structured, repeatable, AI-evaluated assessment. It is an Enterprise-plan capability.

Celeredge also authors and publishes frameworks centrally (you'll see those with a global or shared source badge). Authoring adds your own org frameworks alongside them.

Who can access​

Available on the Enterprise plan. See Billing & Plans. On plans that don't include it, the authoring surface shows a preview with an upgrade prompt.

How it works​

A framework is authored across a set of tabs that take it from an empty structure to a published, runnable assessment:

TabWhat you define
StructureThe tree: Dimensions → Assessment Areas → Questions, with answer types and ordering
Data collectionSurvey, workshop, and interview instruments that gather evidence for the framework
ScoringThe scoring system, mapping, and normalization that turn answers into maturity scores
PresentationThe visualization/section layout for reports, plus version history
WorkshopsWorkshop definitions and sessions
CompileAn AI step that generates the framework's execution profile — see below
Prompts & reviewReview and approve the AI-generated evaluation prompts per question
PublishMake the framework available — to your organization, or a chosen scope

The AI part: Compile​

Compile is an AI feature. When you compile a framework, Celeredge generates its execution profile — the per-question evaluation logic the assessment engine uses at run time, including the prompts that evaluate each question and the retrieval queries that pull the right evidence. This is what makes a hand-authored structure into something the AI can actually run against a client's documents and data, grounded and cited like every other Celeredge output (see Citations & Trust).

The Prompts & review tab puts you in control of that AI output: you review the generated prompts per question and approve them before the framework can be published. The AI drafts; you sign off.

Framework status​

Authoring moves a framework through states: Draft → Compiled → Published.

  • Draft — being authored; structure and settings editable.
  • Compiled — the AI execution profile has been generated (and should be reviewed/approved).
  • Published — available to run in Diagnose.

Compile before you publish. A framework must be compiled (and its prompts approved) before its assessments can execute. Publishing a framework that hasn't been compiled will cause its assessment runs to fail at start — always compile and review first.

Step by step​

  1. Create a framework and build its Structure (dimensions → assessment areas → questions).
  2. Define Data collection, Scoring, and Presentation.
  3. Compile — let the AI generate the execution profile (evaluation prompts + retrieval queries).
  4. Open Prompts & review, check the generated prompts per question, and approve them.
  5. Publish to your organization.
  6. Make it usable on engagements via practice visibility.

Best practices​

  • Write clear, single-intent questions. The AI generates better evaluation prompts from precise, unambiguous questions than from compound ones.
  • Always review compiled prompts. Compile is a strong draft, not a rubber stamp — read the prompts on your highest-stakes questions before approving.
  • Mirror your real methodology. The value of authoring is encoding your expertise; structure the framework the way your consultants actually assess.
  • Version deliberately. Use the Presentation tab's version history; re-compile after structural changes so the execution profile stays in sync.

Common errors​

SymptomLikely causeWhat to do
Authoring shows an upgrade promptNot on the Enterprise planSee Billing & Plans
Published framework's assessments fail to startPublished without compiling / approving promptsCompile, review & approve prompts, then re-publish
Compiled prompts look off for a questionThe source question is ambiguous or compoundRewrite the question, re-compile, re-review