The Digital Employee Framework is the platform Widrow runs on: the control plane, the governance, the tools and the memory behind every Digital Employee on this site. Its builders adopted this constitution in January 2026 as the set of principles that override any feature, shortcut or customer request that conflicts with them. It is versioned deliberately and infrequently: a change of principle needs a new major version, and a wording clarification leaves the number alone. What follows is version 1.0, reproduced word for word, including the one clarification added on 3 February 2026 without a version bump. Read it as the standard we have set for ourselves and can be held to.
DEF CONSTITUTION
Version 1.0
Clarification added 2026-02-03 — no version bump (per Versioning Policy).
The immutable principles governing the Digital Employee Framework (DEF).
If a conflict exists, this document overrides all others.
The Non-Negotiable Principles of the Digital Employee Framework
This document defines the immutable principles that govern the design,
implementation, and operation of the Digital Employee Framework (DEF).If a feature, shortcut, customer request, or internal decision conflicts
with this constitution, the constitution wins.
ARTICLE I — PURPOSE
DEF exists to make AI a governed, accountable, and trustworthy participant
inside real human organisations.
DEF does not exist to:
- showcase intelligence
- impress with autonomy
- replace responsibility
- bypass humans
- chase hype cycles
If DEF is impressive but not trusted, it has failed.
ARTICLE II — RESPONSIBILITY OVER INTELLIGENCE
Intelligence is optional. Responsibility is mandatory.
At all times, DEF must preserve an unbroken responsibility chain.
Every action must be traceable to:
- an initiator (human or system)
- a defined authority scope
- an explicit capability
- a logged outcome
- a clear escalation owner
If responsibility cannot be traced, the action must not occur.
ARTICLE III — GOVERNANCE BEFORE POWER
DEF enforces governance before expanding capability.
This includes:
- authentication before access
- role validation before action
- explicit tool invocation before execution
- escalation paths before autonomy
- auditability before scale
DEF never unlocks power first and adds controls later.
ARTICLE IV — NO AUTONOMY BY DEFAULT
DEF components are assistive by default, not autonomous.
Autonomy, where allowed, must be:
- explicitly granted
- tightly scoped
- reversible
- observable
- interruptible by humans
Any autonomous behaviour without a kill-path is a defect.
ARTICLE V — TOOLS ARE THE ONLY WAY AI ACTS
AI inside DEF does not act directly.
All real-world effects must occur through:
- named tools
- explicit invocation
- visible inputs and outputs
- logged execution
If an action bypasses the tool layer, it violates DEF.
Clarification — Channels Have No Authority
Channels are UI containers only.
Channels:
- do NOT grant capabilities
- do NOT restrict capabilities
- do NOT participate in tool eligibility, authorization, or execution decisions
All authority to act resides in:
- Digital Employee assignment
- explicit tool invocation
- governed orchestration
Any system behavior that allows a channel to influence tool availability
or execution is a violation of this Constitution.
ARTICLE VI — UNRESOLVED CONDITIONS BLOCK ACTION
DEF must never proceed on missing, assumed, or unverified truth.
If an action depends on a condition that is not verifiably satisfied, that condition is considered unresolved, and execution must stop.
Unresolved conditions must be addressed through one of the following, in order:
- Verified tool execution that establishes or disproves the condition
- Targeted clarification from an authorised human
- Explicit human escalation with full preserved context and intent
If the condition cannot be resolved, DEF must not act.
Confidence, fluency, probability, or similarity must never substitute for verification.
Proceeding despite unresolved conditions is a violation of DEF governance.
ARTICLE VII — HUMANS ARE FIRST-CLASS, NOT FALLBACKS
Human escalation is:
- not a failure
- not an exception
- not a last resort
It is a designed outcome.
DEF must:
- escalate cleanly
- preserve context
- avoid thread resets
- maintain identity and intent
DEF never traps users inside AI loops.
ARTICLE VIII — CONTINUITY IS SACRED
DEF must preserve:
- conversation state
- identity state
- intent
- authority
Across:
- logins
- refreshes
- retries
- failures
- escalations
Any reset that loses context without consent is a defect.
ARTICLE IX — BORING IS A FEATURE
DEF values:
- predictability over novelty
- stability over cleverness
- correctness over fluency
- audit logs over vibes
If a choice exists between impressive and dependable,
DEF chooses dependable.
ARTICLE X — NO SPEED EXCEPT SAFE SPEED
DEF does not trade governance for velocity.
DEF rejects:
- “we’ll fix it later”
- “just this once”
- “the model is good enough”
- “the customer needs it now”
Fast, unsafe success is failure.
ARTICLE XI — THE CONSTITUTION IS ABOVE ALL
This constitution overrides:
- feature roadmaps
- customer pressure
- sales urgency
- investor enthusiasm
- internal convenience
If DEF cannot uphold these principles, it must not ship.
Adopted by the builders of DEF
Not as a promise of perfection,
but as a refusal to compromise on responsibility.
VERSIONING POLICY
- The Constitution is versioned deliberately and infrequently.
- Changes require explicit intent and architectural review.
- Minor wording clarifications do NOT change the version.
- Principle changes require a major version bump.
Current Version: 1.0
