Aisha Amaira has spent her career at the vital intersection of marketing technology and customer data, helping organizations transform raw information into actionable business intelligence. As a seasoned expert in CRM and data platforms, she has seen the industry move from simple automation to the complex, often misunderstood world of generative AI. In this conversation, we explore the common misconception that a “perfect prompt” is the secret to AI success. Aisha breaks down her recent experimentation with major language models, illustrating how business context, rather than just clever wording, determines the value of an AI’s output. We discuss the transition from vague assignments to functional creative briefs and the enduring necessity of human judgment in an era where machines can generate endless possibilities but cannot always weigh the commercial trade-offs.
Many professionals view the prompt as the primary driver of AI quality. How do you distinguish between a well-crafted prompt and the actual business context that informs it?
The reality is that we often give the prompt itself far too much credit for the final result. By the time a marketer sits down to type those instructions into a tool like ChatGPT or Claude, they have already performed the most difficult parts of the job: they’ve defined the objectives, gathered the necessary context, and decided what a “win” actually looks like for their specific brand. A prompt is essentially just the final, visible artifact of a long chain of conversations, assumptions, and editorial revisions that happened behind the scenes. When you see a high-quality output, it isn’t just because someone used the right “magic words”; it’s because they provided the model with a clear understanding of the constraints and the business reality. Without that foundation, you’re just asking a machine to guess your intentions, which is where most AI projects fall flat.
When you tested the same strategic assignment across ChatGPT, Claude, and Gemini, you noticed they “filled in the missing intent.” Could you describe how that process looked and why it’s a risk for marketers?
In my first run, I gave all three models what I thought was a perfectly reasonable assignment regarding SEO strategy in an AI-driven search world. However, because I left room for interpretation, each model projected its own “personality” onto the task. Claude treated it like a discovery project, Gemini focused heavily on technical AI search optimization, and ChatGPT built out a massive formal consulting framework complete with governance and phased implementation. The risk here is that if a marketer isn’t specific, the AI will fill in those blanks with its own logic, which might have nothing to do with the actual business problem. You might be worried about a sudden drop in leads, but if the AI thinks the goal is “brand visibility,” it will give you a technically correct strategy that fails to solve your immediate financial pressure.
You’ve mentioned that moving from a generic assignment to a functional “brief” changed everything. What specific details did you add to the HVAC company scenario to anchor the AI’s recommendations?
To ground the experiment, I transformed the request into a detailed brief for a regional HVAC company with a very specific set of circumstances: a mature website, a limited budget, and a strict preference for optimizing what they already owned rather than spending a fortune on new content. I explicitly told the models that our primary goal was increasing qualified service inquiries for the upcoming season, specifically focusing on maintenance agreements because they represent the highest long-term value for the business. This changed the “vibe” of the recommendations instantly. Instead of high-level theory, the models started talking about repair-versus-replacement decision guides and protecting the existing SEO environment. By providing those commercial constraints—knowing exactly where the business makes its money—the AI was forced to stop hallucinating generic best practices and start acting like a specialized consultant.
Even with a better brief, the models generated more ideas than a small business could realistically handle. How do you apply human strategic judgment to narrow down that list of AI-generated possibilities?
This is where the “human in the loop” becomes non-negotiable because any recommendation can be technically valid but still be a terrible investment for a specific company. All three models gave me a massive list of opportunities, but as a strategist, I had to ask: “Would I actually present this to a client? Is it supported by evidence of real customer demand?” For instance, an AI might suggest a massive sitewide transformation, but for a regional HVAC shop with a tight budget, that’s a recipe for disaster. I had to filter the output to focus on urgent, seasonal decisions and audit existing service pages before even considering new content. Strategic judgment is about the “no” just as much as the “yes,” and currently, AI is much better at expanding the field of possibilities than it is at narrowing them down based on real-world limitations.
From your findings, what are the core priorities that consistently survived the refinement process and proved to be the most “commercially relevant”?
Across all the revisions, a few key actions stood out as the most valuable for a business trying to navigate the shift to AI search. First, establishing a baseline for how the brand currently appears in AI-generated answers is essential—you can’t fix what you haven’t measured. Second, we shifted the focus from vague “visibility” to measuring actual qualified leads and booked work. We also prioritized auditing local business information and conversion paths on existing pages. There was a strong consensus on creating new assets only when there was a validated information gap that existing pages couldn’t fill. Finally, the most relevant strategy focused on high-value outcomes like maintenance agreements rather than just chasing high-volume, low-intent traffic that doesn’t pay the bills.
What is your forecast for how the role of the “Prompt Engineer” will evolve as businesses realize that context is more important than the prompt itself?
I believe the term “Prompt Engineer” will eventually fade away, replaced by a return to traditional strategic roles that simply use AI as a high-powered assistant. We are going to stop obsessing over the perfect string of text and start focusing again on the quality of the discovery process and the depth of the business brief. The real value won’t be in knowing how to talk to the machine, but in knowing the business so well that you can provide the machine with the right constraints, data, and objectives. We’ll see a shift where the “brief” becomes the most important document in the workflow again. The most successful people in this space won’t be the ones with a library of copy-paste prompts, but the ones who can look at a model’s output and say, “This is technically correct, but it doesn’t align with our budget or our customer’s emotional journey.” The future of AI in marketing is less about the “input” and much more about the “intent.”