At Trenegy, we’ve used AI in countless ways, from application development to data conversion to ERP implementations. One of the biggest determinants of AI success is the prompt. Aside from an AI strategy and preparing data, prompting will make or break the output.
Prompting seems simple enough. If I explain what I want, AI will deliver. Even if it’s a little unclear, AI is smart enough to figure out what I’m trying to say, right?
There’s a bit more to it than that. AI can make inferences and make sense of our inputs, but that does not always guarantee the right output. Here are a few recommendations for prompting AI to generate outputs that are more accurate and useful.
What exactly are you looking for? AI will not magically deliver the desired results if it’s lacking context. Identify the end goal of what you’re looking to achieve and the steps needed to get there. Craft the prompt based on that information so AI knows what to deliver. This might include storytelling, providing examples and additional context, walking it through your thought process, etc.
Feeding AI too much at once increases the likelihood it will miss certain details or even entire tasks altogether. Suppose you’re using AI to parse through data. Instead of providing 10 instructions at once, break it down into smaller tasks and give instructions one at a time. Once it’s accomplished the first task accurately, move on to the next. Verification of outputs also becomes more manageable with this approach. For example:
An overloaded prompt: “Analyze last month’s daily oil output by well, calculate each well’s average daily production, and then identify the five lowest-producing wells. Craft an 80-word email to the field engineers recommending maintenance priorities that includes a bar chart of the daily averages.”
Why this overloads AI: Multiple analysis steps, crossing into deliverable creation, divergent formats (bar chart, email).
A more effective way to start: “Calculate the average daily oil production for each well from last month’s data.”
After this is verified, move on to the next step.
AI can steer you in the wrong direction without specificity. It needs pertinent details to deliver the desired output. For example:
Vague: “Analyze our rig utilization in early 2025.”
Specific: “For our Gulf Coast fleet of 12 onshore rigs, calculate monthly utilization rate (days drilling ÷ available rig-days) for Q1 of 2025.”
At the same time, providing irrelevant information won’t deliver the desired output either. For example, we created a tool at Trenegy that leverages AI to create standard deliverables faster. We quickly learned: Too much information and the deliverable will be overloaded. Too little information and the deliverable won’t be strong enough.
Put simply, the goal is to give AI the information needed to get the output you want. Good inputs = good outputs. Poor inputs = poor outputs.
And remember, always verify what AI creates and says.
At Trenegy, we help organizations understand and leverage AI realistically to drive value and see results. To chat about how we can help your organization, email us at info@trenegy.com.
Trenegy is a non-traditional consulting firm based in Houston, Texas.