Skip to content

The Loop · People and agents

People set direction.
Agents help the work.

Give every contribution a purpose, an allowed scope and someone who can judge the result. People and agents can work from the same evidence while keeping responsibility for decisions clear.

A useful request has a clear handoff

  1. Agree on the result

    Name the outcome, its checks and the person responsible.

  2. Set the agent’s limits

    Give the task, accessible sources and allowed actions.

  3. Keep the evidence

    Record what changed, what was checked and what remains open.

  4. Make the decision

    The responsible person checks the result and chooses the next step.

Illustrative method · each request and decision has its own purpose.

Agree who is responsible

Useful roles.
Clear decisions.

The Focus owner takes responsibility for the outcome within their area of responsibility. Business, product, design, analysis and engineering can contribute different knowledge. An agent’s output adds to that work; it does not settle a decision by itself.

The outcome owner.

Make the purpose and boundaries clear. Bring in people who can check the value and decide whether to continue, change direction or close the Focus.

In a product team, the Product Owner brings reviews together, is responsible for the retrospective and decides when to close the Focus. Contributors still own decisions within their responsibilities.

The contributors and reviewers.

Define their work, share useful decisions and raise blockers early. Review the result with the people closest to the outcome, such as an analyst, designer or delivery specialist.

The team holds one another accountable for following through. A task is complete when its agreed checks pass.

The permitted agent.

Read the task and relevant sources, ask about gaps and help within the authorized request. Keep proposed changes and evidence available for review.

A method role does not grant workspace permissions. Agent actions follow the current member, connection, feature and source access.

Responsibilities across The Loop

Keep the question clear at each practice.

Focus, Execute, Observe, Review, Learn and Adapt are connected practices. Return to an earlier question when the evidence changes. The examples below describe useful requests, not a promise that every client or tool supports every action.

Method guidance · people’s decisions and bounded agent help
PracticePeople decide and checkUseful agent help
FocusAgree on the outcome, its owner, the scope and how to check success.Read the accessible brief. Identify missing context, assumptions and questions before suggesting a plan.
ExecuteChoose the work to do, resolve scope questions and check contributions.Help with the agreed task through configured tools. Keep decisions, output and verification beside the work.
ObserveChoose relevant evidence, check its source and notice what is still unknown.Summarize the available evidence. Separate recorded facts from interpretations and missing information.
ReviewJudge the result against the intended outcome and acceptance criteria.Prepare a comparison of the expected result and saved evidence. Name gaps instead of treating a summary as approval.
LearnDecide which lesson is useful, why it matters and where it applies.Draft a lesson from reviewed sources, including its limits and the situations where it may not fit.
AdaptApprove the next change and give follow-up work a clear home.Propose a better brief, check or preparation step. Apply supported changes only within the request’s authorization.

Give an agent a useful request

A clear task.
A boundary it can follow.

Start with a work-item identifier, the intended result, acceptance criteria and useful source links. Say whether you want reading, a proposal, a draft or an authorized change. Name the decisions you want brought back for review.

In Tellenze, use Definition for the saved scope and Conversation for questions and progress. Check the output against the source, including any uncertainty the agent reports.

Prepare an agent work request
Synthetic example · a request to inspect a handover task
“Read PORT-18 and its linked handover guidance. Check whether the steps and acceptance criteria are clear. List missing facts and propose edits for review. Keep private source material within its current access.”

The person responsible reviews the proposed edits. The request asks for a proposal; it does not authorize unrelated changes or external messages.

Use current context and permissions

A useful connection.
A current source.

Check the connection and the sources needed for this request. A link does not grant access, and a summary can be outdated. Resolve missing or changed context before relying on a proposed action.

Required Context stays separate.

Where it applies, supported agent changes need the current pinned reading and a valid receipt. An ordinary document link does not make reading required. A receipt records delivery, not understanding or correctness.

Check Required Context

Choose the right agent path.

Looper can answer from accessible workspace context and propose supported changes for confirmation. Ask an agent targets a configured companion on Project work for research, clarification or follow-up coordination.

Agent reactions have limited workspace context and no filesystem access. Repository changes use your normal development workflow.

Read the reaction boundaries

Review actions that reach others.

Check recipients and copy before explicitly authorizing an email. Review publication and reader access are explicit author actions. Deployment or another external change needs authorization for that action.

Preparing a draft or plan is a separate result from sending, publishing or applying it.

Read connected-agent email controls

An illustrative shared contribution

Make a handover usable by the next person.

Synthetic example: a team wants a new teammate to follow a support handover without guessing. These are proposed responsibilities and checks, not measured customer results.

  1. 01

    Agree on the outcome

    The responsible owner names the handover’s purpose and the person who can check it. The task says which steps and decisions the new teammate needs.

  2. 02

    Ask for a bounded review

    An authorized agent reads the accessible task and guidance, flags missing steps and proposes clearer criteria. It names any source it could not read.

  3. 03

    Try the handover with a person

    The teammate opens the guidance with their own account and follows the steps. The team records unclear instructions and updates the agreed work.

  4. 04

    Keep the result and next decision

    The reviewer checks the finished guide and saves the evidence. Separate enhancements get explicit follow-up work. The owner decides whether the Focus’s main value has been delivered.

Closing a Focus does not complete unfinished tasks. Keep their true state and history, and give remaining work an explicit follow-up Focus. A completed agent request also does not mean the original task is complete.

A few useful details

Keep responsibility easy to understand.

Do we need agents to use The Loop?

No. People can use the six practices with their existing board and shared documents. Add an agent where a clear request and usable context make its help worthwhile.

Does being the Focus owner make someone an administrator?

Focus ownership describes responsibility for the outcome. Member and administrator permissions are separate. Work, sources and available actions still follow the workspace’s current access rules.

Does The Loop enforce a fixed sprint length or Focus limit?

Focus timing fits the outcome and may be timeboxed. The recommendation of at most three active Focuses per team helps keep attention manageable; it is not a product restriction or a per-person limit. Read Focuses for the full guidance.

Can an agent publish the review it helped prepare?

In Tellenze, Looper and MCP can help prepare a review draft. Preview publication, named reader access and Publish edition remain explicit actions by its author in the interface. Check the actual report for its intended audience.

Make judgment possible

A shared purpose.
Evidence people can check.

Choose what would show that the outcome happened, then keep both the result and its limits clear.

Choose outcomes and evidence

Keep the next step useful

Continue with The Loop.

COME TAKE A LOOK

A little less busy.
A lot more together.

Curious? Join the waitlist and see how we welcome your team.

Tell us a little about your team. We’ll explain the next steps and invite you when we have capacity for your workspace.

Join waitlist Start with your email. No payment needed to join.