Plan an instrumented campaign with AI

Translate the request into a measurable objective

Define the action, audience, time window and decision owner before selecting channels.

The fictional brief asks for awareness among AI enablement leaders. That phrase does not yet define an observable result. For practice, the owner sets a six-week planning target of 12 qualified demo requests from 300 named accounts. Both numbers are synthetic assumptions chosen for the exercise, not historical performance or a recommendation for another campaign.

Write the qualification rule before launch: a request counts only when the person belongs to a named account, identifies a current assessment or enablement initiative and accepts a follow-up conversation. Keep delivered messages, visits and registrations as separate funnel events. They can help diagnose the plan, but they are not substitutes for the stated result.

Ask AI to expose missing decisions such as geographic scope, exclusions, consent requirements, sales ownership and the date on which the account list is frozen. The responsible owner confirms those conditions. Do not let the model fill an unknown baseline with an invented industry average.

Map audience questions to one offer

The offer should answer a documented audience question rather than a generic campaign theme.

The fictional packet contains three internal notes: leaders ask how to inspect AI-assisted work, how to define evidence and how to avoid overstating a score. The proposed offer is an evidence-planning workshop. These notes are synthetic planning inputs. They do not prove demand or establish that every account has the same priority.

Create a message table with audience question, proposed answer, supporting material, excluded claim and next action. The workshop can promise a structured discussion of evidence planning. It cannot promise improved business performance or validated assessment outcomes unless separate approved evidence supports those claims.

Use AI to create message variants only after the table is approved. Review each variant against the same exclusions. A shorter headline must not turn a bounded workshop description into a claim that the organization has solved AI measurement.

Allocate channels and budget as hypotheses

A channel plan records why each route might reach the audience and what evidence would change the allocation.

Allocate the fictional 60-unit budget as 25 units to direct email, 20 to a partner newsletter and 15 to a webinar. Units avoid implying a real currency or media price. The allocation is a decision to test, not evidence that one channel has a known return. Record the assumption behind each route and the person authorized to change it.

For direct email, the hypothesis is that named-account relevance will produce qualified visits. For the newsletter, the hypothesis is that the partner reaches the intended role. For the webinar, the hypothesis is that a detailed evidence topic warrants a scheduled session. The packet includes no delivery, audience-quality or conversion history, so all three remain uncertain.

Ask AI to challenge the allocation with failure scenarios: incomplete contact permission, a partner audience mismatch or low attendance. The output should list evidence requests and contingency choices. It should not rank channels using fabricated performance data.

Instrument the funnel before launch

Define each event, source and reconciliation rule before results can influence the story.

The plan distinguishes message delivered, landing-page visit, registration, attendance and qualified request. For each event, record the source system, identifier, timestamp, owner and known duplication risk. A registration can occur twice, a visit may lack identity and an attendee may not meet the qualification rule.

Create a reconciliation rule that deduplicates by approved identifier and preserves source provenance. If direct and partner links both appear in one journey, retain the attribution rule chosen before launch and disclose its limitation. Do not ask AI to select the most flattering source after outcomes appear.

Set a review condition: pause performance interpretation if tracking is incomplete, and review the campaign after the first 100 named accounts have received an eligible touch. Zero qualified requests at that point triggers investigation, not an automatic conclusion that the offer or audience has failed.

Sources: [1]

Revise the plan when a channel slips

A versioned change log keeps a late operational event from being rewritten as the original strategy.

In the fictional exercise, the partner newsletter moves from week two to week five. The owner must decide whether its remaining contribution can still be observed within the six-week window. Do not backfill the original schedule or pretend the delayed route produced exposure before it occurred.

Create version two of the plan. Keep the original allocation, record the date and reason for the change, and move 10 of the newsletter's 20 units to direct email preparation while retaining 10 for the delayed placement. This is an illustrative response, not a generally optimal allocation. The owner must approve any real budget movement.

Update the measurement note because the newsletter will have a shorter observation window. Preserve separate channel results rather than comparing unequal windows without qualification. Ask AI to identify dependent assets and dates, then verify each change against the actual contract and campaign calendar.

Deliver a campaign decision pack

The final artifact connects objective, audience, message, channels, budget, instrumentation and review conditions.

Deliver the objective and qualification rule, audience-question table, offer boundaries, channel hypotheses, 60-unit allocation, event dictionary, attribution rule, risk register, review conditions and version log. Mark every target and allocation as synthetic. Keep approval owners beside decisions that affect budget, data use or public claims.

The live demand-generation parent reconciles people, events and follow-up eligibility after a webinar. This page plans a coordinated multi-channel campaign before launch. The growth-marketer parent predefines a readout for a single message experiment. This page manages several routes, one budget and an operational schedule change without claiming causal proof.

Review the participant on whether the plan remains inspectable when facts change. Good work may recommend delaying interpretation because tracking failed. A full calendar and polished copy do not compensate for an undefined outcome, unsupported audience assumption or missing authority.

Sources and scope

Government Communication Service: OASIS campaign planning. Skills England: responsible workplace AI use.

The Government Communication Service OASIS framework organizes campaign planning around objectives, audience insight, strategy, implementation and scoring or evaluation. Skills England includes checking AI outputs, spotting errors and making informed decisions in its foundation skills for work benchmark. These primary sources support the general planning and verification background only. They do not validate this exercise or its fictional targets.

All accounts, targets, budgets, channels, events and schedule changes in this exercise are original synthetic examples. They are not campaign benchmarks, customer results or a promise about Two Prune. Before adapting the method, confirm current source material, data permissions, channel terms, product claims and budget authority.

Sources: [1] [2]