The IMPACT prompting formula is a simple checklist for writing better AI prompts: Intent, Message, Persona, Audience, Context, and Tone. Instead of a one‑line request, you give the AI a short brief that explains your goal, who you are, who it’s for, and how the answer should look.
Think of AI as a smart intern. If you say, “Write something about our sales process,” you’ll get something generic. If you say, “Help me fix our no‑show rate on first design meetings,” you’re already in IMPACT territory. One remodeler who shifted from one‑liners to structured prompts cut “back‑and‑forth” edits by roughly 60% in a week because the AI finally understood the real job.
IMPACT works across tools—ChatGPT, Claude, Gemini, Perplexity, NotebookLM—because it improves your instructions, not the model. You can paste the same structured prompt into any of them and expect more relevant, concrete output than with a loose, search‑style query.
Intent is the "I" in IMPACT: the outcome you actually want. A strong intent is specific, measurable, and time‑bound. Instead of, “Help with marketing,” write: “Draft a 150‑word follow‑up email that re‑engages a homeowner who missed yesterday’s design review call.”
A quick test: could someone else on your team read your intent and know when the AI’s answer is “done”? If not, tighten it. For example, “Summarize this 8‑page client proposal into five bullet points a salesperson can review in under two minutes” tells AI exactly what success looks like.
In practice, most professionals under‑specify intent. They skip word counts, deadlines, or constraints and then blame the model for wandering. Try rewriting three of your last AI requests as intent statements with a clear goal, format, and success criteria. Many users report that this change alone improves their first‑draft quality by 30–40%.
Persona, Audience, and Context are the targeting system. Persona is who you are (“I’m the owner of a design‑build firm in California”). Audience is who the result is for (“Busy homeowners nervous about budget creep”). Context is who the AI should act as (“You are a senior remodeling sales coach”).
When you include all three, the same task produces very different—and better—answers. A prompt like: “You are a senior design‑build consultant. I’m a remodeler writing a blog for first‑time renovation clients in the Bay Area. Explain change orders in plain language and avoid legal jargon,” will generate far more useful copy than a bare “Explain change orders.”
A practical example: one team fed their training manual into NotebookLM, then added a persona (“branch manager”), audience (“new hires”), and context (“expert sales trainer”) to create tailored practice scenarios. That shift turned a static PDF into an interactive coach that answered questions the way their best trainer would.
Message and Tone control how the AI works with you and how the final answer feels. Message gives it permission to ask questions and challenge you: “Ask clarifying questions before you answer. Challenge my assumptions if my plan seems risky or incomplete.” Tone shapes the output: “Use plain language, short paragraphs, and concrete examples.”
Adding one sentence of message can prevent expensive mistakes. For example, “If you lack data, don’t guess—ask me,” sharply reduces hallucinations when you’re dealing with pricing, timelines, or contracts. Likewise, specifying tone—“professional but conversational, no buzzwords”—can instantly fix content that sounds robotic or overly academic.
Finally, treat prompting as iterative. Run a first draft, then respond with: “Here’s what to keep, change, and add,” and ask the AI to rewrite. Teams that review and refine prompts monthly—especially when new models launch—build a reusable prompt library that consistently saves hours per week across email, content, and client communication.