How business context changes AI recommendations

Each AI success story appears to finish the identical manner. Somebody shares a outstanding output, and virtually instantly somebody asks, “Will you share the immediate?”

It’s an affordable request. We regularly give prompts an excessive amount of credit score. By the point somebody writes a immediate, they’ve already outlined aims, gathered context, weighed tradeoffs, and determined what success appears like.

Prompts seize the results of conversations, assumptions, revisions, and editorial judgments that got here lengthy earlier than anybody began typing.

To see how a lot that course of issues, I gave the identical strategic project to ChatGPT, Claude, and Gemini, altering solely the data that got here earlier than the immediate.

The experiment

The project was easy. An organization had invested in search engine optimisation for years, and its web site was effectively established. Management knew search habits was altering as AI-generated solutions turned extra widespread, however nobody knew what, if something, ought to change concerning the advertising and marketing technique.

I gave the identical project to ChatGPT, Claude, and Gemini. The aim was to create a strategic roadmap figuring out:

• What info needs to be gathered.
• Which alternatives deserved consideration.
• Which assumptions required validation.
• What ought to occur earlier than content material creation started.

I wasn’t making an attempt to determine the “profitable” mannequin. I wished to see how the responses modified because the project advanced.

Be the brand customers find first.

Track, grow, and measure your visibility across Google, AI search, social, local, and every channel that influences buying decisions.

Start your free trial

First run: The fashions crammed within the lacking intent

At first, I believed I’d crafted a superbly cheap immediate. Then I spotted what I’d truly written was a superbly cheap project. Somewhat than reproduce it line by line, right here’s what it included:

• A request for strategic steering somewhat than advertising and marketing copy.
• An evidence that it was for a longtime enterprise involved about its visibility in AI-generated search.
• Directions to determine lacking info to stop hallucinations and invented info.
• An expectation that the response would prioritize planning over execution.

The whole opening project is proven in Determine 1.

It outlined the issue, established the kind of assist I wanted, and discouraged unsupported assumptions. It additionally left room for interpretation.

Then I ran the project by means of all three fashions. The responses got here again rapidly. They have been considerate, well-organized, and full. 

Every would have appeared cheap by itself. Aspect by facet, nevertheless, they revealed one thing necessary: They weren’t actually answering the identical query.

• Claude handled the project as the start of a discovery mission. 
• Gemini interpreted it primarily as an AI search optimization mission. 
• ChatGPT expanded it into a proper consulting framework with governance, measurement, and phased implementation.

Not one of the interpretations have been unreasonable. The difficulty was that the fashions hadn’t been instructed which enterprise drawback mattered most.

Was the issue declining site visitors? Model visibility? AI citations? Lead technology? Aggressive strain? Future readiness? Every requires a special technique. 

With out sufficient info to know the underlying intent, every mannequin provided its personal interpretation. When an project leaves room for interpretation, fashions don’t merely fill in lacking info. They fill within the lacking intent.

I finished serious about how you can enhance a immediate and began serious about how you can enhance the temporary.

Second run: Higher context, higher suggestions

Skilled entrepreneurs start with conversations. They be taught what success appears like, which prospects matter most, the place the enterprise makes its cash, what constraints exist, and which compromises are acceptable. Then they flip that info right into a useful temporary.

That’s what my course of deliberately skipped for the experiment.

For the second run, I saved the unique project largely intact and added the data that will usually emerge throughout these conversations.

The revised temporary established the shopper as a regional HVAC firm with a mature web site, a restricted finances, and a choice for enhancing current property earlier than creating vital new content material. It additionally clarified the target of accelerating certified service inquiries in the course of the upcoming season, with upkeep agreements representing the highest-value long-term end result.

The revised temporary is proven in Determine 2.

I didn’t add an elaborate prompting components or ask the fashions to purpose in a specific manner. I provided the data a strategist would ordinarily collect earlier than recommending a plan of action.

The distinction was rapid. All three fashions retained their unique approaches, however their suggestions have been grounded in the identical enterprise.

• Claude remained discovery-oriented, specializing in seasonal demand, upkeep agreements, current property, and finances prioritization.
• Gemini remained extra technical, tying its suggestions extra intently to native service questions and business constraints.
• ChatGPT retained its framework-driven method, prioritizing repair-versus-replacement choices, maintenance-plan growth, and defending the corporate’s current search engine optimisation atmosphere.

Extra enterprise context remodeled the project into a quick. The responses turned extra detailed, extra commercially related, and extra intently aligned with the identical enterprise problem.

The standard of the response depended much less on the sophistication of the immediate than on the standard of the enterprise context behind it.

---

Ultimate run: The place human judgment nonetheless issues

A greater temporary introduced the fashions nearer collectively. It didn’t robotically make their suggestions helpful.

All three generated extra potentialities than a regional enterprise with a restricted finances may fairly pursue earlier than the upcoming season. They recognized alternatives, organized them, and defined why every would possibly matter. Selecting amongst them nonetheless required strategic judgment.

Selecting among the many suggestions required strategic judgment. I evaluated them by a special commonplace:

• Would I current this to a shopper? 
• May they act on it? 
• Was it supported by out there proof? 
• Did it align with the aims, or was it merely an affordable search engine optimisation exercise?

Any suggestion may very well be technically legitimate and nonetheless be a poor funding. A content material alternative would possibly seem helpful however lack proof of buyer curiosity or demand. A measurement framework may very well be thorough however unrealistic, given out there staffing or methods.

Just a few priorities persistently survived throughout mannequin revisions:

• Establishing a baseline for the way the corporate seems in commercially vital AI-generated solutions.
• Measuring certified leads, booked work, and upkeep settlement progress alongside citations and AI referrals.
• Defining what counts as a professional inquiry earlier than treating visibility as successful.
• Auditing current service pages, native enterprise info, and conversion paths earlier than commissioning substantial new content material.
• Creating new property solely when current pages couldn’t tackle a validated info hole.
• Emphasizing pressing seasonal choices and upkeep settlement alternatives somewhat than a sitewide transformation.

The fashions expanded the sector of potentialities, however strategic judgment nonetheless wanted to slender it.

Own the conversation before your competitors.

See where your brand appears, where it doesn’t, and exactly how to win more visibility across search, AI, local, social, and every channel that matters.

Start your free trial

Prompts are proof, not explanations

A immediate screenshot sometimes conceals how the project framed the duty, how the temporary added enterprise context, and the way human judgment established priorities.

Actual prompts are helpful as a result of they present how a job was framed and what info the mannequin acquired. They don’t reveal what the consumer already knew, what the primary run uncovered, what assumptions have been rejected, or why the subsequent revision modified.

That’s why a immediate that produced a powerful end in one scenario delivers one thing merely ample in one other. The phrases are the identical, however the considering behind them isn’t.

A immediate makes discovery, analysis, enterprise judgment, and editorial overview seen to the mannequin. It doesn’t exchange them.

What the immediate leaves out

The experiment began with an project. It advanced into a quick. Human judgment established the priorities.

A immediate is the primary seen artifact of that workflow. We ask others for prompts as a result of they’re straightforward to share. They match right into a screenshot, a social put up, or a slide.

The conversations, altering assumptions, false begins, and judgment calls behind them are tougher to bundle. They’re additionally much more instructive.

The subsequent time somebody shares a powerful AI output, asking for the immediate should still be worthwhile. Simply don’t mistake it for the total clarification. The extra helpful query is how they obtained there.