Practical guide
AI fluency for marketing operations: test routing rules
Test a fictional campaign routing change with boundary cases, precedence rules and a reversible launch checklist.
Marketing operations managers can practice AI fluency by turning campaign rules into testable behavior. This fictional routing exercise covers rule precedence, missing fields, repeated events and rollback. Its output is a checked rule sheet, not a campaign performance report or a recommendation to contact real people.
Write rules that produce one result
Start with the intended routing behavior before asking AI to configure anything.
The fictional campaign has three rules: existing customers go to the account team; new inquiries requesting a demonstration go to the demonstration queue; other inquiries remain in general review. A record can meet several descriptions, so the ordered rules matter. Existing-customer status takes precedence over a demonstration request in this exercise.
Write the input fields, allowed values and resulting queue. Ask AI to express the logic in plain language, then compare it with those three instructions. A plausible extra rule, such as prioritizing a large company, is outside the supplied policy. Keep suggested changes separate from the behavior being tested.
Build a small routing test set
A useful test set includes conflicts and missing information, not just easy matches.
Use fictional records A through E. A is an existing customer requesting a demo and should reach the account team. B is a new demo inquiry and should reach the demonstration queue. C is new without a demo request and stays in general review. D has unknown customer status; E has a missing request type.
Do not guess the expected routes for D and E. Mark the policy gap and ask the owner to define a fallback before launch. AI can generate more cases, but verify that each adds a different decision boundary. Ten copies of B with different names provide little additional coverage.
Check repeated events without sending messages
Routing an event twice should not quietly create two independent follow-up tasks.
For this exercise, assign event B the identifier EVT-B and replay the same event. The intended behavior is one queue item for that event identifier. A later, genuinely new inquiry receives a different identifier and must not disappear merely because it came from the same fictional person.
Record both expected outcomes in the test sheet. This is a specification for a simulated workflow, not a claim that a particular marketing platform handles duplicates this way. Test in an isolated environment without real contacts or outbound actions. Ask AI to explain which field distinguishes a repeated delivery from a new request.
Separate a rule change from permission to activate it
A passing simulation does not authorize a live campaign or establish contact permission.
Keep a before-and-after rule sheet and identify who owns activation. Record the approved fallback, test results and a way to restore the prior configuration. If the new rule routes existing customers incorrectly, the rollback decision should be obvious without reconstructing the whole conversation.
Introduce a change that moves demonstration inquiries ahead of customer status. Re-run A first: its route changes, showing why precedence needs review. Do not describe that change as an improvement unless the owner actually intended it. Preserve the failed test rather than editing the expected answer to match the new output.
Hand over a checked routing contract
The final artifact should let another operator reproduce the important cases.
Deliver the ordered rules, field definitions, five test records, replay case, unresolved policy decisions and activation owner. Distinguish tests executed from tests merely proposed. If the environment cannot run the workflow, describe the result as a paper walkthrough and leave execution verification open.
Review whether the worker caught the conflicting rule and missing-field cases, not how many AI prompts they used. The demand generation guide addresses who participated in a campaign. This guide addresses whether a routing change behaves as specified before those records enter an operational queue.
Sources and scope
NIST: AI risk management. Skills England: workplace AI foundations.
These references provide background, not validation or endorsement of this exercise. The case details, calculations and suggested review questions are original instructional material. Use them to discuss observable work, not to infer customer outcomes, professional credentials or performance in every setting. Before adapting the exercise, confirm the relevant facts, approved tools, data permissions and decision owners. If you change the case, revisit the expected answers and checks as well. These examples describe practice tasks, not a promise that a particular product includes the fictional features.
Sources: [1] [2]
Sources
- 1.AI RMF Core · NIST
- 2.AI foundation skills for work benchmark · Skills England