WorkflowLensby Kristin Chen

Methodology · Version 1.0

Evidence before
automation.

WorkflowLens is a decision-support tool, not an autonomous transformation strategy. Its rubric makes the reasoning inspectable so a team can challenge assumptions before implementation.

01 / Assessment sequence

Start with work, not tools

  1. 01

    Understand the outcome

    Define the decision, deliverable, people, volume, constraints, and business priority.

  2. 02

    Map the current state

    Extract tasks, owners, systems, handoffs, inputs, outputs, and friction—then require human correction.

  3. 03

    Assess each task

    Recommend human work, AI assistance, automation, redesign, or no AI using five consistent dimensions.

  4. 04

    Design the pilot

    Specify checkpoints, controls, metrics, assumptions, and a staged implementation roadmap.

02 / Scoring rubric

Five dimensions, scored 1–5

Impact

How materially could the intervention improve time, quality, consistency, or decision usefulness?

1 low / 5 high

Feasibility

Can the task be supported with available technology and a bounded, testable implementation?

1 low / 5 high

Data readiness

Are the inputs available, understandable, permitted, and sufficiently reliable?

1 low / 5 high

Risk

What is the consequence of an incorrect, biased, insecure, or unexplained result?

1 low / 5 high

Change effort

How much process redesign, training, ownership, and integration will adoption require?

1 low / 5 high

03 / Recommendation logic

“Do not use AI” is a valid result

Keep Human

Judgment, accountability, empathy, negotiation, or high-consequence approval dominates.

AI Assist

AI can draft, classify, summarize, compare, or triage while a person verifies the result.

Automate

The task is repetitive, rules are stable, exceptions are bounded, and failure is detectable.

Redesign First

The process or data is too inconsistent to automate responsibly.

Do Not Use AI

Risk, legality, data rights, or explainability makes AI inappropriate for the task.

04 / Evaluation & limits

What the product is—and is not

Evaluation approach

  • Synthetic fixtures across marketing, research, finance, operations, and support.
  • Schema, evidence alignment, risk recognition, and prompt-injection tests.
  • Manual review for recommendation fit, actionability, and unsupported claims.

Limits

  • The report cannot replace legal, security, privacy, or domain-expert review.
  • Scores organize a discussion; they are not scientific measurements.
  • ROI scenarios represent potential capacity value, not guaranteed savings.
Run an assessment →See the worked example ↗