Opinions expressed by Entrepreneur contributors are their own.
Key Takeaways
- Better AI results come from better prompts, not pricier tools. To extract the greatest value from AI use, you need to provide the model with data about context, audience, examples and a clearly defined task.
- Treat AI use as an iterative process. Refine prompts and responses over multiple rounds rather than expecting a perfect answer from one request.
- AI can accelerate research, writing and routine work, but outputs should be validated by humans — especially when they affect customers or compliance.
Imagine this. Two founders sit down with the same AI subscription. They are competitors. They target the same client base. They have similar offerings. The outcome should be the same, right? But the results are anything but. One founder trims hours from his firm’s weekly workflow. He fashions a successful marketing campaign.
The other one? It doesn’t go as well. His copy is generic and forgettable. The AI actually made his job harder, because now he needs to make some sense of what is correctly termed “AI slop.”
So, what happened? They had different prompts. One put real thought into their prompt and followed through. The other didn’t, and the results showed. For entrepreneurs, that’s good news. The newest or most expensive AI stack is no longer the ticket to success. Process is the key.
As IBM has reported, the content and structure of prompts can affect the quality and relevance of large language model outputs. To extract the greatest value from AI use, one needs to provide the model with data about context, audience, examples and a clearly defined task.
Defining the task
AI users often go wrong right off the bat. They input goals but don’t really tell the model what to do. Maybe it’s a request to improve a marketing campaign, help an entrepreneur scale a business or just write better copy. But there is no measurable outcome in these prompts. They are vague, and so the results that are returned are just as useless.
A better approach is to define the task in a simple sentence and to attack a corresponding measurable outcome. Provide the AI with data. Imagine you are a B2B software company selling to mid-market manufacturers. You want your email open rate to rise to 25% from 18% over the next 90 days. Ask for specific changes to accomplish that, and also request metrics to gauge whether or not it’s working.
The AI now has a market, an audience, a baseline and even a deadline. The task has been clearly defined. And there’s no need to interpret a generic answer because the output will be specific.
Providing context
To realize the benefits of using AI, you need to provide it with useful context. Tell it about your company stage, customer profile, existing processes, constraints and the format you need. IBM calls this context engineering. Large language models don’t have long-term memories. You need to provide them with as much context as possible when designing an effective prompt.
That doesn’t mean you need to copy all of your documents and paste them into a chatbot. Instead, select information that can lead to better decisions. If you are creating a new campaign, you might input information on target customers. If you are crafting a new sales email pitch, input information about what your target customer might react to, a problem they are trying to solve. You don’t want to overload AI with information. You want to feed it the right information.
A balanced approach
One mistake entrepreneurs make is expecting a perfect, actionable answer from a single prompt. But for those who wish to get more out of their AI, the better bet would be to take an iterative approach. You might at first get back generic responses, but by making follow-on requests, you can get feedback closer to what you are looking for.
IBM defines iterative prompting as progressively refining prompts and responses to improve clarity, completeness and usefulness. Over time, by inputting more precise instructions, examples and constraints, you can obtain better results.
That same mindset can also be applied to how you rely on AI within your organization. NIST’s AI Risk Management Framework, for example, emphasizes managing AI risks and incorporating trustworthiness into the design, use and evaluation of AI systems. In this sense, one should continue to check outputs, rather than treating them as automatically correct.
If you have a small business, you might use AI to accelerate your research, or for analysis or even routine execution. But human judgment is necessary when it affects customers or compliance. The same goes for marketing. While AI can be used to inform your team, it cannot replace them. According to Salesforce, 72% of customers surveyed considered it important to know when they were communicating with an AI agent. And 71% said human validation of AI output was important for building trust.
Remember, AI alone is not a solution. It’s how you use AI that counts.
Some parting advice
You don’t need to revise your AI strategy as soon as you finish reading this. Instead, pick a recurring problem, be it customer-support drafts, sales research or meeting summaries. Then, run through the same process for two weeks. Start out by defining the task. Add relevant context. Iterate on the output and then validate your results. In the end, measure what changed.
If you process shaved time off your workflow and you achieved the results you wanted, you might just be well on your way to making the best of your AI.
The entrepreneurs who harness the most from AI are not the ones with the most expensive subscriptions. They are the ones who know what they want and how to get there. AI is powerful, but at the end of the day, it’s real human beings who are in the driver’s seat.
Key Takeaways
- Better AI results come from better prompts, not pricier tools. To extract the greatest value from AI use, you need to provide the model with data about context, audience, examples and a clearly defined task.
- Treat AI use as an iterative process. Refine prompts and responses over multiple rounds rather than expecting a perfect answer from one request.
- AI can accelerate research, writing and routine work, but outputs should be validated by humans — especially when they affect customers or compliance.
Imagine this. Two founders sit down with the same AI subscription. They are competitors. They target the same client base. They have similar offerings. The outcome should be the same, right? But the results are anything but. One founder trims hours from his firm’s weekly workflow. He fashions a successful marketing campaign.
The other one? It doesn’t go as well. His copy is generic and forgettable. The AI actually made his job harder, because now he needs to make some sense of what is correctly termed “AI slop.”
So, what happened? They had different prompts. One put real thought into their prompt and followed through. The other didn’t, and the results showed. For entrepreneurs, that’s good news. The newest or most expensive AI stack is no longer the ticket to success. Process is the key.
