Maliwan Savage / Workflow tools

Workflow Decision Map · Version 2.1

A method you can inspect.

Assess one workflow, compare the options and choose the next justified commitment.

Five lenses, with separate conditions

Value, alternatives, delivery, control and evidence organise the assessment. Permitted data use, accountability, delivery access and error control and the system's authority are conditions for proceeding. A large benefit cannot cancel out an unresolved or blocked condition.

The tool may recommend clarifying the process, evaluating rules, testing AI with review, validating evidence, redesigning the proposal or keeping the current approach. These are decision aids, not approval to deploy.

One standard across teams

The six steps are Define, Map, Baseline, Compare, Controls and Reinvest. Sales, Finance, Talent, Web & design, Content & marketing and general profiles change examples and challenge cases. They do not award points or change the decision rules.

Every workflow needs one unit of work, a fair simpler comparator, a quality or service measure, evidence from comparable cases, an explicit action boundary and a named owner. For a capacity objective, the outcome threshold protects quality while effort falls. For a quality or service objective, compare the stated outcome first; use effort to distinguish equal outcomes.

The framework is a structured starting point for cross-team discovery, not a validated universal standard. The published department examples are synthetic stress tests. Additional workflows, organisational policies and operational trials are still needed to validate suitability in a particular team.

How a recommendation is selected

  1. A blocked condition routes the proposed operational change to defer or redesign.
  2. Unsupervised financial, employment or other high-consequence decisions route to redesign. Narrow the proposal to preparation or individual human approval. Autonomous AI actions with lower consequences still require a narrower, reviewed comparison before this tool recommends a pilot.
  3. An unclear process or a waiting-time bottleneck routes to clarification.
  4. A reported failed AI comparison prompts a simpler alternative or redesign.
  5. Missing decision evidence routes to validation, including a review-time range that reverses the preferred option for a capacity objective. Missing outcome measurements, comparison sources or test scope also require validation.
  6. An option missing the agreed quality or service threshold cannot win on capacity alone. If both options miss, redesign; if only one meets, evaluate that option subject to the other conditions.
  7. For the capacity objective, options that release no capacity support keeping the current approach; stable rules or a stronger simpler comparator favour rules.
  8. A confirmed interpretation task with promising reported evidence and resolved conditions can justify a bounded AI comparison.

Reported conditions are not independently verified. For quality or service objectives, time estimates support the discussion but do not establish the outcome. There is no universal readiness percentage or organisation benchmark.

A single average is insufficient to establish accuracy, fair treatment or acceptable risk. The report adds team-specific challenge cases and keeps success and stop criteria separate. A respondent can still misstate evidence or choose an unsuitable metric; the tool cannot independently detect that.

Calculations and evidence

Monthly effort = monthly items × total active minutes per item ÷ 60 + additional recurring upkeep. Expected rework is averaged across all items. Capacity released = baseline effort − option effort. AI advantage = simpler option effort − AI effort.

Setup effort, waiting time and financial costs are separate. Unknown values stay unknown. Negative release means extra effort. Savings from alternative options are not added together. The supplied review-time range changes only review time, holding other inputs fixed.

Evidence labels distinguish reported measurements, estimates and unknowns. Even a reported measurement has not been verified by this tool. Illustrative examples use invented inputs, including those representing measured evidence.

Older version 2 records remain readable. New outcome and authority fields begin as unknown, so a previously suggested pilot may now require validation. Review those fields before relying on a regenerated brief.

Capacity is an opportunity to improve effectiveness

Reduce routine effortRelease capacityReinvest human timeImprove decisions & feedbackImprove the workflow

Reducing routine work can create time for better decisions, useful feedback and improvements to instructions, checks and exception handling. The report asks where that time would go, who protects it and which outcome should improve.

Track three things separately: capacity released, time actually reallocated and better outcomes. Reinvestment consumes released capacity; it is not an additional saving or an automatic improvement.

Your information and sharing

The assessment and calculations run in your browser. Your report is held in browser session storage. Draft persistence on this device is optional and can be switched off in the questionnaire. Downloading a PDF, editable record or public card creates a local file.

Full PDFs and editable records include your answers. Share them only with intended recipients. The public-card preview omits workflow names, figures, systems and free-text answers, and includes only the recommendation and general method. A result-page URL does not share your personal report.

Optional AI suggestions, when available, require an explicit choice to send the workflow name, trigger, finish and incident through OpenRouter to its OpenAI model provider. Keep confidential and personal information out. Requests exclude providers that collect data for training. OpenRouter and the model provider's applicable retention and data policies still apply; this is not a promise of zero retention. Our endpoint does not log or store the submitted description. Pseudonymous request counters expire after an hour; a site-wide daily counter bounds use. Counters do not contain workflow text.

No email is required. The diagnostic adds no tracking of your answers and does not send your report to Maliwan. You choose whether to make contact.

OpenRouter data policies

Sources and limits

This is Maliwan Savage's practical synthesis, not a certified framework or a claim of compliance. It supports discovery and internal discussion. It cannot establish permitted data use, acceptable consequences or investment approval on behalf of an organisation.

Assess my workflow