Learn With Nathan
Sign in
Full sample chapter · AI Business Automation
What Should You Automate First?
Choose the first automation using frequency, stability, value, data readiness, exceptions, consequence, ownership, and reversibility.
Chapter 01
What Should You Automate First?
Choose the first automation using frequency, stability, value, data readiness, exceptions, consequence, ownership, and reversibility.
The best first automation is rarely the most impressive process; it is the smallest stable workflow where value and control can be proved.
AI can organize process evidence and compare candidates using an agreed rubric. The accountable person must understand the real work, expose exceptions, judge consequence, and approve the candidate.
This chapter keeps automation opportunity selection practical for a business reader: the job, evidence, controls, finished artifact, and success measure are all visible. A low-to-medium technical learner can later implement the same method in approved software.
The business case for automation opportunity selection
Start with a measurable result, not an AI feature
This method makes automation opportunity selection observable as business work: trigger, inputs, rules, exceptions, owners, systems, controls, outcome, and operating cost.
Do not prioritize automation of sensitive, consequential, unstable, low-volume, or poorly owned work merely because it consumes attention.
National Institute of Standards and Technology provides official guidance relevant to the workflow, obligation, or control used here.
Design the work before asking AI
Good inputs and acceptance rules reduce rework
Map the current work and baseline first. Then define trigger, authorized data, decisions, actions, systems of record, human approvals, failure routes, audit evidence, and success criteria.
The two concepts to remember are and . Automation candidate means a bounded workflow evaluated for frequency, stability, value, data, exceptions, and consequence. Exception rate means the share of cases that cannot follow the standard path and require different handling.
| Weak setup | Business-ready setup | Why it matters |
|---|---|---|
| Tell us what we should automate with AI. | Score these eight operations workflows on monthly volume, minutes, delay, error cost, rule stability, exception rate, data quality, reversibility, permissions, owner, and pilot effort. Recommend one draft-only pilot and show why the others wait. | It compares actual workflows, includes risk and readiness, and limits the recommendation to a test. |
Confirm consent, confidentiality, account settings, permissions, and applicable policy before giving any tool the information used in automation opportunity selection.
VALUE-RISK selection: the practical playbook
A repeatable sequence a busy professional can follow
- 1. Inventory friction. List repeated work, volume, delay, errors, handoffs, and affected people.
- 2. Test stability. Map rules, inputs, exceptions, policy change, and source quality.
- 3. Score consequence. Assess external action, sensitivity, reversibility, compliance, and failure cost.
- 4. Estimate value. Measure baseline labour, wait time, rework, system cost, and capacity released.
- 5. Choose pilot. Select smallest useful scope, owner, shadow mode, cases, metric, and stop rule.
Context to provide: Current process, owners, systems, fields, rules, volumes, exceptions, permissions, service expectations, and baseline.
AI job: Design or run the ranked automation candidate scorecard within the supplied rules; stop and route cases that do not meet them.
Return: Ranked automation candidate scorecard, test evidence, exception queue, run log, owner, and operating metric.
Quality rules: Do not invent data, ownership, approval, consent, identifiers, status, or business outcomes; do not perform an external action outside explicit authority.
Worked example: Choosing among eight office workflows
Follow the evidence from messy input to an approved result
VALUE-RISK selection: the working loop
Five controls move a business workflow from mapped work to tested and monitored operation
Audiobook description
The figure shows a five-step path from left to right. Step 1, Inventory friction, List repeated work, volume, delay, errors, handoffs, and affected people. Step 2, Test stability, Map rules, inputs, exceptions, policy change, and source quality. Step 3, Score consequence, Assess external action, sensitivity, reversibility, compliance, and failure cost. Step 4, Estimate value, Measure baseline labour, wait time, rework, system cost, and capacity released. Step 5, Choose pilot, Select smallest useful scope, owner, shadow mode, cases, metric, and stop rule. A reminder below the path says to begin with a bounded use case, verify the result, and improve the workflow.
| Starting material | AI-assisted result | Human review |
|---|---|---|
| Process interviews, monthly volumes, timing sample, error logs, system map, exception estimates, data classes, owners, and consequence ratings. | A scorecard ranking invoice-intake extraction for a draft-only pilot while postponing expense approval and external lead messaging. | The operations team increases the exception estimate after sampling real cases and rejects a candidate with no clear system owner. |
Apply it safely and measure the gain
A faster draft is useful only when the finished work is better
Start with a small volume, reversible actions, representative cases, and an owner who currently performs the work. Run in shadow or draft mode before enabling writes.
Measure: Track baseline completeness, exceptions discovered, pilot value, correction time, incidents avoided, and candidates stopped. Measure the complete workflow, including preparation, review, correction, exceptions, and recovery—not generation speed alone.
Keep the ranked automation candidate scorecard, approval record, exception notes, and metric result as the evidence for whether this automation should be repeated, changed, or stopped.
Pause and apply
Reflection questions
- Where does automation opportunity selection create the most avoidable delay, inconsistency, or risk today?
- Which input, judgment, or external action must remain under explicit human control?
- What evidence would justify expanding this automation after the first test?
Sources and further reading
Chapter 1 endnotes
- National Institute of Standards and Technology. Artificial Intelligence Risk Management Framework 1.0, 2023. NIST AI RMF.
- Office of the Privacy Commissioner of Canada. AI, Privacy, and Your Business, 2025. Official Canadian privacy guidance.
- Agent Worker Academy. AI Business Automation Production Brief, 2026. Academy blueprint.