AI Workflow Opportunity Assessment — Practical Super Intelligence

Stage 1: Task & Baseline Volume

Step 1: What specific operational workflow do you want to evaluate?

Focus on a single recurring, bounded task rather than a broad department function.

Examples:

Operational Workflow Category

Role Currently Executing This Task

Step 2: Frequency, Volume & Time Investment

Accurate volume estimates ensure our economic calculator computes realistic labor savings.

Monthly Units / Items Processed

Current Manual Minutes Per Unit

Baseline Labor Footprint: ~62.5 human hours per month spent on this workflow.

Step 3: Primary Operational Bottleneck

What is the primary operational friction or cost in the current process?

Processing Backlog & Delays Slow turnaround time delays downstream operations or customer fulfillment.

Error & Rework Costs Clerical typos, missed discrepancies, or misclassifications require costly rework.

Specialist Labor Waste High-value staff are bogged down in mundane mechanical copy-pasting.

Stage 2: Process Consistency & Data Quality

Step 4: Process Standardization & Rules

How standardized is the workflow, and how predictable are the inputs?

Standardized & Documented (+10 pts) Written SOP exists; inputs follow predictable structures/templates; clear rules govern decisions.

Moderately Variable (+6 pts) Mostly consistent, but 15% to 25% of cases require discretionary human judgment or unusual layouts.

Unstructured / Ad-Hoc (+1 pt — Warning Trigger) Highly variable; outcomes depend heavily on unwritten tribal knowledge or subjective negotiation.

Step 5: Data Cleanliness & Accessibility

What format is the required information in, and how accessible is it?

Clean & Digitized (+10 pts) Searchable digital text (PDFs, spreadsheets, cloud database, email body) with historical examples available.

Mixed / Semi-Structured (+6 pts) Scanned documents, low-resolution images, or messy tables requiring initial OCR or cleaning.

Siloed / Physical (0 pts — Warning Trigger) Trapped on paper, handwritten notes, legacy closed terminals without copy-paste, or unrecorded phone calls.

Stage 3: Operational Risk & Human Control (Safety Gates)

Step 6: Data Sensitivity & Confidentiality

What sensitivity level applies to the data processed in this workflow?

Public / Low Sensitivity (+10 pts) Marketing copy, open catalog data, non-confidential operational product descriptions.

Internal Operations (+7 pts) Vendor names, standard B2B invoices, general operational correspondence, dispatch logs.

Confidential Commercial (+4 pts) Margin calculations, proprietary supplier pricing agreements, customer lists.

Regulated / Highly Sensitive (Safety Gate Trigger) Personally Identifiable Information (PII), Social Security numbers, banking credentials, medical records.

Step 7: Downstream Error Consequence

If an undetected error slips through, what is the worst-case operational impact?

Minor / Internal Inconvenience (+10 pts) Internal typo easily caught and fixed with zero external or financial impact.

Moderate / Operational Friction (+6 pts) Minor delay, internal rework, or small administrative adjustment (under $200 impact).

Significant Financial / Client Impact (+2 pts — Safety Gate Warning) Misbilled invoice ($500+), incorrect material shipment, contract terms error, or customer churn.

Catastrophic / Legal / Safety Hazard (Safety Gate Override ➔ Red Light) Regulatory fine, legal breach, OSHA safety violation, or severe data compromise.

Step 8: Human Verifiability & Audit Speed

Can a qualified person verify output accuracy before real-world execution?

Rapid Verification Under 30 Seconds (+10 pts) A human reviewer can visually spot-check or confirm accuracy in seconds against the source document.

Moderate Verification (1–3 Minutes) (+6 pts) Reviewer must check a couple of cross-references, but validation is straightforward.

Laborious / Complex Verification (+2 pts) Checking the output takes almost as long as doing the task from scratch.

Unverifiable / Blind Execution (Safety Gate Override ➔ Red Light) Outputs cannot be verified prior to real-world execution, or autonomous execution is expected.

Step 9: Exception Handling Architecture

What happens when the model encounters an ambiguous case or low confidence?

Dedicated Specialist Triage Queue (+10 pts) Uncertain or low-confidence cases are automatically flagged and routed to a named specialist.

Ad-Hoc Review (+5 pts) Team reviews flagged cases as time permits, but no formal queue exists.

No Defined Exception Path (0 pts — Warning Trigger) No structured exception path planned; ambiguous cases risk being processed blindly.

Stage 4: Pilot Target & Exit Gate

Step 10: Primary Pilot Success Metric

How will you determine whether your 30-day pilot was successful?

Required Minimum Accuracy

Target Pilot Duration