How to Use AI Chat Like a Pro

By now, most small and midsized business teams have settled on an AI chat tool of some kind, whether that’s Microsoft Copilot, ChatGPT, Claude, or something built directly into the apps they already use. Most have also learned, usually the hard way, what not to do with it: don’t feed it a one-line question and expect magic, and don’t hand over information that has no business leaving the building.

Knowing the tool and knowing the landmines only gets a team so far. What actually separates the people getting real, usable output from the ones who quietly stopped bothering after a few disappointing sessions almost never comes down to which model they’re using. It comes down to how they’re talking to it.

Here’s what changes when a team moves from casually poking at an AI chatbot to actually operating it like a tool: fewer generic answers, less re-typing the same request five different ways, and output that’s usable on the first or second pass instead of the fifth. None of it requires a technical background. It requires a slightly different habit.

Give It Context Upfront, not a One-Line Ask

Most disappointing AI answers trace back to the same root cause: the person asked a question the way they’d ask a coworker who already knows the whole backstory. AI chat tools don’t have that backstory unless you give it to them. A one-line prompt gets a one-size-fits-all answer.

The fix is simple, and it takes about ten extra seconds. Before asking for anything, tell the AI four things: who it should act as, what the actual goal is, who the output is for, and roughly what form you want it in.

Vague: “Write a follow-up email to a client.”
Specific: “You’re a project manager at a construction firm. Write a follow-up email to a client whose permit approval is delayed two weeks. The tone should be reassuring but direct, and it should end with a clear next step. Keep it under 150 words.”

The second version doesn’t just sound better, it’s usable without a rewrite. The AI isn’t guessing at your industry, your relationship with the client, or your tone. You told it.

Treat It Like a Conversation, not a Vending Machine

A lot of people treat AI chat like a search engine: type a question, take whatever comes back, and if it’s wrong, start over from scratch with a totally new prompt. That’s the slow way to do this. Every AI chat tool keeps the thread of the conversation, which means it remembers what you asked for and what it already gave you. Use that.

Instead of restarting, refine. If the draft is too formal, say so. If it missed a key point, add it as a follow-up instead of retyping the whole request. Each round makes the output closer to what you actually needed, and it usually takes fewer total messages than starting over would.

“That’s close, but make it shorter and drop the third paragraph.”
“Good structure. Now rewrite it assuming the reader has never seen our pricing before.”
“Keep everything the same, just make the tone less formal.”

This is the single biggest habit shift between someone who finds AI chat mildly useful and someone who relies on it daily. It stops being a slot machine and starts being an editor sitting next to you.

Be Specific About What the Output Should Look Like

Content and format are two different requests, and most people only ask for the first one. Telling the AI what to write is half the job. Telling it how to structure the answer is the other half, and it’s the part that saves the most editing time afterward.

If you want...	Say something like...
A scannable summary	"Give me this as five bullet points, no more than one sentence each."
A specific tone	"Keep it professional but warm, like you're writing to a long-term client."
A word or length limit	"Cap it at 200 words. Don't pad it to hit the limit."
A structured document	"Use headers for each section: Background, Options, Recommendation."
A table instead of prose	"Put this in a table with columns for Task, Owner, and Due Date."

Skipping this step is why so many people end up with a wall of text they then have to manually reformat. Asking for the shape of the answer up front is faster than fixing it after the fact.

Know What Should Never Go in the Chat Box

Getting good at prompting doesn’t mean anything is fair game to paste in. This is the direct continuation of a problem most SMBs already have without realizing it: employees are using AI chat tools daily, and a lot of what they’re pasting in shouldn’t be leaving the building.

The rule of thumb: if you wouldn’t post it on a public forum, don’t paste it into an AI chat window, especially a personal or free-tier account that isn’t covered by your company’s data protection agreements. That includes:

Client and employee PII — Social Security numbers, account numbers, health information, home addresses
Financial data — unreleased earnings, payroll details, banking information, wire instructions
Credentials and access details — passwords, API keys, network configurations
Confidential business strategy — pricing negotiations, M&A discussions, unreleased product plans
Proprietary source code or trade secrets your business doesn’t want resurfacing elsewhere

This is exactly the kind of gap that shows up long before anyone notices a breach. It doesn’t take a sophisticated attack, just an employee trying to move faster and a chat box that doesn’t ask any questions before it remembers what you typed. It’s the same underlying pattern behind how modern phishing bypasses traditional email filters →: attackers and AI tools alike are only ever as safe as the habits around them.

Prompts That Actually Work for SMB Roles

Good prompting isn’t abstract. It looks different depending on what you’re actually trying to get done. A few real examples, built around the tasks that come up constantly in SMB roles:

Role / Task	Sample prompt
Office manager Drafting an email	"You're an office manager. Write a short email to the team reminding them the holiday schedule is posted on SharePoint. Friendly tone, three sentences max, no corporate jargon."
Team lead Summarizing a meeting	"Here are my raw meeting notes: [paste notes]. Summarize into three sections: Decisions Made, Open Questions, and Action Items with owners. No commentary, just the facts."
Operations First-pass report	"Act as an operations analyst. Using this data: [paste or describe data], write a one-page summary for leadership. Lead with the headline finding, then three supporting points, then a recommendation."
Sales Prospect follow-up	"Write a follow-up email to a prospect who went quiet after a demo two weeks ago. Low-pressure tone, reference the demo generally, and end with a simple yes/no question to reopen the conversation."
HR Policy explainer	"Rewrite this HR policy in plain language a new hire with no legal background could understand in one read. Keep the meaning identical. Use short paragraphs."

The Tools Doing the Heavy Lifting in 2026

Prompting skill matters more when the tools themselves keep getting more capable. Microsoft 365 Business Premium and Copilot for Business already puts Copilot inside Word, Outlook, Teams, and Excel, which means the context-setting habits above apply just as much to summarizing a Teams call as they do to a standalone chat window.

The ecosystem is also no longer single-model. Microsoft has begun integrating Anthropic’s Claude models directly into Copilot for certain tasks, giving teams access to more than one AI “brain” without switching apps. And for more involved, multi-step work, tools like Claude Cowork extend the same conversational approach into full task delegation, letting an AI agent work across documents, spreadsheets, and other apps on your behalf rather than answering one prompt at a time. The prompting fundamentals don’t change: context, iteration, and format specificity still determine whether the output is useful, no matter which tool is doing the work.

Related Read: Claude Inside Microsoft Copilot →

Getting good at AI chat is a skill, and like any skill, it’s easy to plateau at “good enough” without ever hitting “actually efficient.” The difference between the two is almost always a handful of habits: context, iteration, format, and knowing exactly where the line sits on what’s safe to share. None of it requires a technical rollout. It just requires someone showing the team what good looks like, and a security foundation that makes the good habits actually matter.

Want your team using AI the right way, not just the fast way?

SBT's Total IT Management pairs Microsoft 365 Business Premium with the training, governance, and security guardrails that turn scattered AI use into a real productivity advantage.

TALK TO SBT ABOUT AI ENABLEMENT→