The advice on a client call was good. The article never got written. By Friday, three short drafts are back on your desk, each missing the point you made on Tuesday.
AI can carry that advice into a complete answer, then prepare the versions a campaign needs: the article, the email, the post, the line a salesperson uses. It can also sort customer questions and help read the results. You still decide what the business stands for and what gets published.
This guide covers two checks: does a claim stay accurate when one answer becomes four pieces, and does a small report really support the next decision? For prompts to try on four common marketing jobs, see how our marketing apprentice used AI to promote his book.
Eight marketing jobs AI can prepare
| Marketing job | What goes in | What comes back | What you decide |
|---|---|---|---|
| Customer research | Interview notes, enquiries and feedback you may use | Themes, supporting quotes and open questions | Whether the sample represents the buyers you want |
| Competitor research | Current competitor pages and authoritative sources | A comparison of offers, claims and gaps | What matters, and whether the facts are current |
| Positioning and campaign planning | Buyer problem, offer, evidence and limits | Message options and a campaign brief | Audience, promise and how you reach readers |
| Original writing | Expert notes, approved facts and writing samples | An article, email or script that answers the brief | Substance, accuracy and whether it sounds like you |
| Adaptation | An approved core answer and each channel's purpose | A post, email introduction or sales excerpt | Whether each version does a different job |
| Editorial review | Sources, current facts and voice examples | Unsupported claims, contradictions and suggested edits | Which edits are sound |
| Production planning | Owners, formats and dates | An asset list and a release schedule | Who does what, and when it goes out |
| Reading results | Comparable reports, definitions and notes on what changed | Calculations, possible explanations and a next test | Whether the evidence supports the decision |
What each job needs depends on your tool and setup. Treat these as uses to try.
One expert answer, four checked pieces
Made-up example. The business, records and figures are invented to show the method. They are not a client result or output from a real AI run.
A service adviser often gets asked: "Should I discount an empty afternoon?" The adviser supplies the figures. A CAD 600 job is offered for CAD 450. Materials and other job costs are CAD 180. Four hours of direct labour cost CAD 160. Both costs arise only if the job goes ahead. Overhead and tax sit outside the calculation, and the afternoon could still fill.
The brief asks for a complete web answer, a newsletter introduction for subscribers who opted in, a short post and an excerpt the adviser can use with prospects. It tells the AI to use only the supplied figures, state their limits, and leave out results, price advice and promised enquiries.
The answer
Should you discount an empty afternoon?
Work out what's left after the costs that arise because you take the job. In this made-up example, a CAD 600 job leaves CAD 260 after CAD 180 of materials and CAD 160 of direct labour. At CAD 450, it leaves CAD 110. Both figures are before overhead and tax.
The price falls by CAD 150, or 25%. What's left falls by the same CAD 150, about 58% of the original CAD 260. A modest price cut has a much larger effect on what remains.
Taking it may make sense if the time would otherwise sit empty and it won't crowd out a better booking. Turning it down may make sense if it displaces full-price work or sets a price customers will expect next time. If you pay the labour either way, count only the costs that actually change.
Before saying yes, write down the offer price, the costs that change, the time needed, what else you could do with that time and what the customer will expect next time.
The claim check
| Figure | Calculation | What the editor can say |
|---|---|---|
| Normal contribution | CAD 600 − CAD 180 − CAD 160 = CAD 260 | Amount left after the stated costs, before overhead and tax |
| Offer contribution | CAD 450 − CAD 180 − CAD 160 = CAD 110 | Same basis, at the offer price |
| Price reduction | CAD 150 ÷ CAD 600 = 25% | Reduction in price |
| Contribution reduction | CAD 150 ÷ CAD 260 ≈ 58% | Reduction in what's left, under these assumptions |
The four versions
| Version and reader | Sample text | What the editor checks |
|---|---|---|
| Web answer for owners | The answer above | Costs and reasoning, with the adviser |
| Newsletter introduction for subscribers | "An empty afternoon can make a discount tempting. In our made-up CAD 600 example, a 25% price cut reduces what's left after the stated costs by about 58%. The full answer explains what to compare first." | Audience, consent, link and figures |
| Short post for an owner between jobs | "Made-up example: CAD 600 less CAD 340 of job costs leaves CAD 260 before overhead and tax. Charge CAD 450 and that becomes CAD 110. Before discounting, compare the time you give up and the costs that actually change." | That the limits survive the shorter version |
| Excerpt for the adviser to use with a prospect | "For your offer, the useful question is what changes if you accept: price, costs, time and future expectations. Your own figures decide your case." | That the example stays general and leaves their price to them |
The planted error
One draft calls the CAD 110 "net profit". It isn't: overhead and tax haven't been taken off. The editor replaces it with "contribution after the stated costs, before overhead and tax", then checks the article, email, post, excerpt and search description for the same label. They also strike a line promising the campaign will bring more enquiries.
A label that overstates the number is easy to miss once and easy to copy four times. Check labels as closely as arithmetic.
When a better-looking report points the wrong way
Made-up example. A team compares two periods of equal length. The service description changed in Period B, and one referral's source is unknown.
| Measure | Period A | Period B |
|---|---|---|
| Landing-page visits | 200 | 300 |
| Enquiries | 8 | 9 |
| Enquiries per visit | 4% | 3% |
| Qualified conversations | 3 | 3 |
Visits rose 50%, but the enquiry rate fell from 4% to 3% and qualified conversations stayed at three. Two explanations remain open: the new description drew different visitors, or visitors read the offer differently. A draft recommends "double output because traffic grew 50%", and the marketer rejects it, because nothing in the report shows output is the constraint.
With numbers this small, a few conversations with readers will tell you more than a percentage. How to tell whether AI is actually helping your business covers how to judge whether a change helped.
What keeps the work distinctive
Start with something your expert knows and your buyer needs. A fluent draft can hide an unclear offer, and that needs a decision about the offer.
Give the AI current facts and writing samples the business likes, and save good corrections in its instructions. Imagined buyer reactions can raise questions; interviews with real customers answer them.
For search, Google says that using AI tools to generate many pages without adding value for users may violate its spam policy on scaled content abuse. It also advises you to check all AI-generated content for accuracy, including titles and descriptions (Google's guidance on generative AI content). Checked November 2026.
Search itself is changing too: a growing share of buyers ask ChatGPT or Google's AI for a recommendation and never click a link. See why AI search matters for your business.
Start with one answer
Choose one question buyers already ask. Prepare the full answer and one version for another channel. Count the effort from notes to approved piece, corrections included, then watch whether readers use the answer.
For help with voice samples and review, see the human and AI writing guide. To see a full package, look at the content example.
If your best marketing ideas stay in client calls and never become an article, book a free 30-minute call. We'll look at one question your buyers ask and what it would take to answer it well.


