By Toby Doebrich · Playbook Perspectives
AI can give you a very confident paragraph about an offer you don’t actually sell. That’s a problem, even if the commas are excellent. Before I judge a writing tool by how finished the first draft looks, I want to know how much checking and editing it needs.
Choose a low-stakes assignment
Use one small task, such as a social caption explaining an existing service. Pick material you can safely use in the selected tool; leave out confidential client information, personal records and unpublished commercial details.
Prepare a short reference sheet containing the actual offer, audience, practical facts, words to avoid and two examples of writing that sound right for the business.
Keep an untouched copy. You need a stable reference to tell whether the draft changed something important.
Make the brief specific
For example, an imaginary walking-tour operator could supply: a two-hour tour, a named public meeting point, a maximum group size and an explicit instruction not to promise access to closed venues.
Those are hypothetical details, not a real offer.
A useful prompt might be:
“Write a short social caption for the audience described in this brief. Use only the supplied facts. Preserve all limitations. Do not add testimonials, results, discounts or superlatives. If an important detail is missing, identify it separately instead of inventing it. Match the tone of the two examples without repeating their wording.”
The prompt sets a standard. It does not remove the need to review the result.
Keep the comparison fair
If you compare tools or configurations, give them the same reference sheet and assignment. Record the tool, selected model or mode, date and whether you provided follow-up instructions.
Do not compare one tool’s first attempt with another tool’s heavily coached revision and call it a fair contest. If you test a revision round, apply the same correction request to each candidate.
A small exercise cannot establish which tool will be best for every task. It can reveal which workflow deserves a more careful trial.
Review five things
Accuracy: Is every factual statement supported by the reference sheet?
Completeness: Did it retain the details and limitations the audience needs?
Voice: Could the business plausibly say this, or does it depend on generic enthusiasm?
Judgment: Did the draft make a promise or implication that the business would not approve?
Editing effort: What had to be changed before someone could use it?
Treat an invented factual claim as a stop-and-correct problem, not something to average away with a good tone score. A pleasant voice cannot compensate for an inaccurate offer.
Count the work after generation
A fast first draft can still require substantial checking. Record the time spent preparing the brief, reviewing the output and making corrections.
Compare that total with a draft written using your existing process. The result may support a narrow use: alternate openings, a first outline or variations based on already approved copy. That can be useful without handing over the entire publishing decision.
Use the result as a workflow decision
At the end, write one sentence: “We will use this tool for ___, with ___ reviewed by ___ before publication.”
That is a more useful outcome than declaring that the tool understands your brand.
Good output earns a place in a process. It doesn’t eliminate the need for one.
