# Keep the business moving Use this for ongoing responsibility for a business goal, including a recurring Business Loop or a manual continuation. Each run resumes decisions and work; it does not start a new audit. For a bounded request, keep that scope rather than turning it into management of the whole business. ## Establish what we are trying to achieve Recover the existing product, research, plans, results and founder decisions. Understand the intended customer and value, the nearest desired business outcome and horizon, the founder's view of the obstacle, and material time, budget and operating commitments. Ask only about gaps or conflicts that could change the next choice; spread questions through the conversation rather than requiring a completed questionnaire. Keep the founder's original consequential answers or their source alongside a concise working understanding in existing project notes. Distinguish their goal and constraints from their diagnosis, observed facts and your hypotheses. Briefly reflect what you understood; a different diagnosis needs an explanation, while changing the goal or material constraints needs the founder's agreement. Reuse and update this context instead of repeating the initial interview. ## Choose the current focus Start with active work, promises and evaluation points. Review relevant changes in product use, customer feedback, money, operations and acquisition. Begin with summaries when useful, then inspect original evidence for a consequential claim: actual customer journeys and outcomes, messages, payment/refund records, campaign results or market sources. Missing access limits a conclusion; it is not evidence of no customers or no problem. Use [diagnosis](diagnose.md) for a loss or opportunity, [evaluation](evaluate.md) for interpretable results, and [Market and acquisition](../market-and-acquisition.md) for customer, market or channel questions. Reuse applicable research; follow [market evidence upkeep](../market-and-acquisition/market.md#keep-market-evidence-useful) when a decision depends on its freshness. The next action need not use the same channel or approach as a previous pilot. Recommend one primary focus relative to the goal, comparing the nearest credible alternatives by expected contribution, confidence, time, cost, risk and ability to execute. Avoid invented numerical rankings. The choice may be a product improvement, acquisition test, operating repair, specific missing evidence, continuing agreed work, or waiting for its evaluation point. A primary focus does not exclude useful parallel work or urgent customer obligations. Explain concisely why this focus now, what I can produce using the relevant Egbe capabilities, what the founder needs to do or decide, and when the intended effect can be assessed. Name material assumptions and what would change the choice. If uncertainty prevents a useful commitment, propose the smallest question worth answering and how its answers would change the decision; do not substitute indefinite research for action. ## Turn the decision into work Use [founder collaboration](collaboration.md) for the recommendation, materials and dependent approvals. Carry forward their current scope; distinguish approval, correction, rejection and an unanswered proposal. Do not reopen an unchanged approval or treat silence as one. Keep the concrete result, owner, scope, significant limits, evidence and next checkpoint in the work's existing plan. For experiments, retain the hypothesis and evaluation criteria established before execution. Use the applicable delivery or channel method and technical skills to do agreed agent-owned work, not merely repeat an offer to begin it. For founder-owned work, agree the needed result and a useful follow-up point. Help remove blockers and continue independent authorized work while waiting. Preserve an outstanding decision without repeatedly sending the same request. A recurring review is not authorization for new spending, outreach, publication or changes to the product. ## Observe what can change the next decision Tie monitoring to active work: what needs checking, which evidence or condition could change the action, who responds and when another check becomes useful. Reuse existing observations instead of creating a parallel monitoring plan. Campaign stop conditions, urgent failures or replies may need faster checks than the overall Business Loop; their results feed the same work and decisions. Choose timing from the expected signal, reporting delay, risk and founder's preferences. A daily check is a starting option, not a requirement for every business. If nothing can usefully advance before a known checkpoint, propose checking then rather than waking repeatedly to report no change. Keep necessary safety checks and explicitly requested digests. Use `egbe-os-product-routines` for the agreed schedule, delivery and later changes. ## Resume, evaluate and report At the next run recover the current focus, decisions, owners, active work and checkpoint from the project's records before collecting more data. Check what has actually changed and continue from the last verified result. Reconcile an uncertain execution before repeating it; unfinished work is not a reason to restart the original pilot or discard its criteria. Use [evaluation](evaluate.md) to distinguish completed work from its effect on customers and the business. Inspect the agreed result when it becomes evaluable; otherwise state the missing evidence or next useful point of assessment. A finished task should not disappear from attention before its effect is considered. Retain the focus while its reasons hold. Reconsider it when results, constraints, customer harm or a meaningful market change alter the choice; explain the reason and preserve the earlier hypothesis and interpretation. No new signal may mean continuing or waiting, not another study, campaign or daily demand on the founder. Reassess the observations along with the work: retain useful checks, change their timing or scope when justified, and retire those whose purpose has ended. Keep the reason and next checkpoint in the existing plan. These are work-plan changes, not automatic rewrites of the project's standing method. Report meaningful changes, the current focus and why, progress, observed effects and any decision needed. Use `egbe-os-product-report` to package these findings; the report does not make a second independent plan. Keep it concise and in the first person, with source links for important conclusions. Save enough context in existing `egbe-docs/` plans/results for the next run to continue; no separate loop dossier or copy of the entire conversation is needed. The same method works on request when automation is unavailable or declined; do not promise execution after this chat without a verified runner.