Decision 1 of 7
Automation risk decisions
What To Automate Vs What To Keep Manual
Use this to decide whether an action should be automated, prepared for review, or kept manual.
Step 1
Answer guided questions
Step 2
Get verdict
Step 3
Use result in strategy session
Decision Tool
Turn a workflow idea into a yes, no, or fix-first call.
3-5 minutes7 decisions
Action
Decision examples
Example output from the ADA method.
Lead intake
Automate summary and routing recommendation. Keep disqualification and strategic account handling under review.
Proposal review
Automate checks for missing sections and risky language. Keep price, scope, and send approval manual.
Customer escalation
Automate context summary and risk flags. Keep final response and concessions manual.
Reporting
Automate draft brief and variance detection. Keep interpretation, budget, staffing, and public claims manual.
CRM cleanup
Automate duplicate detection queue. Keep merge, delete, ownership, and lifecycle changes under review.
Vendor evaluation
Automate evidence packet and scorecard draft. Keep criteria, weights, risk acceptance, and selection manual.
The simplest rule
Let AI prepare work when the source evidence exists. That is where the capacity comes back. Keep people in control when the action changes revenue, records, expectations, legal exposure, or customer trust.
Why this improves trust
Buyers do not need to believe AI is perfect. They need to see that the workflow has review points, stop rules, and accountable owners.
How to use these examples
Pick the closest workflow, copy the automate/manual split, then adapt it to the specific data, owner, risk, and output.
Quality Bar
Use these checks before calling the workflow ready.
The boundary is written before implementation starts.
The team can explain why each manual step stays manual.
The first version makes people faster without hiding judgment.
The workflow stops when evidence is missing or risk is unclear.
Expansion depends on correction history, not optimism.
Where This Helps
Use it before build decisions get expensive.
Governance planning
Owner training
Workflow design
Automation risk review
Research Basis
Built against practical AI risk and quality standards.
Related Resources
Read this with the workflows and service pages it supports.
FAQ
What should AI automate first?
Preparation work: summaries, classifications, draft outputs, checks, comparisons, routing suggestions, and evidence packets.
What should stay manual?
Final decisions involving money, legal exposure, customer commitments, sensitive data, record changes, or low-confidence evidence.
Can the manual step be reduced later?
Yes, but only after the workflow has evidence, correction history, owner trust, and clear performance data.
Next Step
Use this result on a real workflow.
Bring the bottleneck, current handoff, and the result you want to improve.