Why shared context needs clear boundaries.

When engineering work changes hands, people need to know where the information came from and who can rely on it. That’s the starting point for our approach to Collab.

Principles

Six principles behind the product.

01

Context stays within the right permissions

People and tools see only the context their role and task allow. A connected AI tool never receives more than the person using it.

02

People and approved AI tools contribute

Team members, and the AI tools the organisation approves, can add evidence, notes and proposals to the work.

03

Context Approvers decide what is trusted

Only a person in the Context Approver role can turn working context into approved organisational knowledge. AI tools cannot approve context.

04

Source evidence stays connected

Every conclusion keeps a link to the code, ticket, document or session it came from, so anyone can check it.

05

Uncertainty stays visible

Open questions, assumptions and disagreements are recorded as what they are, rather than presented as settled facts.

06

Existing tools stay in place

Repositories, editors, trackers, CI and communication tools remain the systems of record. cRUD Collab connects them rather than replacing them.

A governed memory model

Promote context by trust level, not by repetition.

01

Working evidence

Keep sources, observations, tests and unresolved questions visible while the team investigates.

02

Reviewed conclusion

Record what people accepted, which alternatives they rejected and which uncertainty still remains.

03

Reusable knowledge

Carry approved context forward with its source, scope, access rules and responsible people still attached.

Human authority

Evidence first. Approval before memory.

A source, observation, model response, test result, proposal and approved decision do not carry the same authority. The system must preserve those differences instead of flattening them into one answer.

Approved AI tools can participate inside an authorised workflow, but people review conclusions, resolve conflicts and decide what becomes reusable organisational knowledge.

Human authority stays explicit

Contribution and approval are different responsibilities.

People and AI tools can help the team understand the work without gaining authority they were never given. Roles, approval scope and policy remain visible from the organisation level down to the context being reviewed. Teams defined around a product make access easier to manage, while Project membership, Context Approver assignments and explicit exceptions keep sensitive work bounded.

Role in the workWhat they can contributeWhat they can approveWhat remains visible
ParticipantQuestions, observations, source evidence and session input within the access they have been granted.Nothing by default. Participation does not turn a suggestion into organisational truth.Identity, activity, source references and the relationship between the contribution and the Session.
ContributorEvidence, analysis, working context, proposed knowledge, draft decisions and recommended actions.A contributor cannot approve their own proposal unless an explicit, scoped policy grants that authority.Source, version, authorship, changes, uncertainty and every review state the proposal passes through.
Context ApproverReview comments, requested changes, conflict resolution and a documented approval decision.Only the context categories, Products or Projects assigned to their approval scope.The evidence reviewed, applicable policy, decision, rationale, approver and time of approval.
Product or Project OwnerMembership, context policy, inheritance rules, exception boundaries and accountable ownership.Ownership does not automatically make one person the sole approver. Approval still follows the configured policy.Role changes, policy changes, exceptions, approval assignments and the audit trail behind them.

Questions

Deciding what the team can rely on

Can an AI contribution become trusted team knowledge on its own?
No. An approved AI tool can contribute material for people to review. An authorised human Context Approver decides whether that material becomes Approved Context.
What happens when the evidence is incomplete or conflicting?
Keep the uncertainty visible. The team can work with a provisional interpretation while keeping it distinct from the conclusions it has reviewed and approved.

See the complete product on the cRUD Collab website.

The cRUD Collab website explains how teams work with it, the principles behind it and the thinking that shaped it.