How to Use AI at Work
How to Use AI at Work
A Guide For Small and Medium-Sized Organizations
Everyone says “use AI”; we’re here to show you how. This is the no-jargon, genuinely useful guide – what AI is, 20+ ways to use it, how to write a prompt that works, a 90-day plan, and how to stay safe. Built for any business, by a team that runs AI on its own every day.
What “AI” Actually Means
When people say “AI” at work today, they almost always mean generative AI – software that produces text, summaries, images, or code from a plain-language request. The most common form is the large language model, the engine behind Copilot, ChatGPT, and Claude.
The mental model that matters: an LLM is like an extremely well-read, very fast assistant who will confidently attempt anything you ask, types instantly, and occasionally makes things up with a straight face. That’s why it’s important that a human always stays in the loop.
What you type in; the instruction. Better, more specific instructions get better results. That’s the entire skill of prompting.
The specific AI engine doing the work (Copilot, ChatGPT, Claude, Gemini). New versions ship constantly; the concept doesn’t change.
When the AI states something false as if it were fact. The single most important reason to verify before you trust.
The background you give the AI so it understands your situation. The more relevant context, the better the output.
An AI that can take steps on its own – look something up, fill a form, draft and route a message – not just answer. More powerful meaning it needs more oversight.
How much text the AI can hold in mind at once. You don’t need the math, just know there’s a limit.
A saved prompt plus reference material you reuse, so you don’t retype instructions. The biggest time-saver once a workflow clicks.
Your Work Sorted
Before you touch a tool, spend ten minutes on the most useful exercise in this whole guide: sort your real tasks into three buckets. Here’s where common work tends to land. Make your own list and you’ll see exactly where to start.
AI can do most of it
Routine, repeatable, low-judgment work.
Draft a standard email Summarize a long report Format messy data into a table Take and clean up meeting notes First-pass categorize requestsYou do it WITH AI
Your judgment, sped up by a smart assistant.
Analyze and compare a vendor list Plan a project or campaign Write a proposal draft in your voice Prep for a client call Pressure-test a decisionOnly you can do it
Trust, ethics, judgment, and people.
Coach a struggling teammate Make the ethical call Own a tough decision Build a client relationship Read the room and adjustLook at the shape of your own work. Mostly bucket one? AI can clear real space on your plate - start with the most tedious task this week.
20+ Ways to Use AI at Work, Easy to Advanced
Start at the top and climb. Each level adds a skill to your toolbox. Tap any card for examples and a copy-ready prompt.
Tap any card to see examples and a ready-to-use prompt.
Write a Prompt That Works: the CRIT Method
A bad first result almost always traces back to a vague prompt. CRIT fixes that in four parts, and the third one is a part beginners often skip. Fill in the fields and copy your prompt.
Your CRIT prompt
Telling the AI to ask you questions first turns a one-size-fits-nobody answer into one built for you – because it gathers what it’s missing instead of guessing. Learn more about prompting in our AI Starter Series.
Five Steps to Get Started This Week
You don’t need a strategy deck to begin. You need one good first problem and a safe place to try it. Tap each step.
Pro tip: give it a week, not a day. The awkward “this is slower than doing it myself” phase is real and short.
Not your most important workflow – your most tedious one, straight out of Bucket 1. The weekly report. The meeting notes. The email you send 30 times a month. Low stakes, high frequency, easy to check.
For anything touching customer data, employee data, or financials, use a business/enterprise tier (Microsoft Copilot, ChatGPT Enterprise, Claude for Work) where your data stays yours and isn’t used for training.
Be specific about who it’s for, what format you want, and what good looks like. Tell it to ask you questions first. The difference between “write a follow-up email” and a full CRIT prompt is the difference between something you delete and something you send.
Treat the first draft like work from a sharp but brand-new intern: usually good, occasionally confidently wrong. Verify any fact, number, name, or quote before it leaves your hands.
When a prompt produces a great result, save it as a reusable assistant and share it. The compounding starts the moment you stop solving the same problem twice.
A 90-Day Plan to Go From Curious to Capable
A 90-day rhythm turns a single experiment into an organization-wide habit. Run it for yourself or roll it out across a team.
How CIT Uses AI Inside Our Own Business
We don’t sell AI we don’t use. This is operational, not theoretical, and it’s the same model we recommend to every client: augment people, protect data, keep a human in the loop.
The gap between “we should do something with AI” and “AI is part of how we operate” is crossed by doing, safely and with a trusted partner. We cross it ourselves before we asked any client to – through our Managed IT and Intelligence Services teams.
✓ Our company runs on private, secure AI. We are security and compliance experts, ensuring our business and yours are following appropriate and safe AI governance guidelines.
✓ Our teams are trained on AI usage. We run internal cohorts to train our employees on AI safety and efficiency in the workplace, allowing us to provide industry-leading services to organizations.
✓ It augments our teams. AI does the data heavy-lifting so our employees can focus on judgment, problem-solving, and people.
✓ We teach what we practice. We offer training resources like our AI Starter Series.
From Trying It to Running On It
CIT’s Intelligence Services ensure that your complete AI effort covers four areas. It can be easy for an organization to be strong in one but weak in the rest. We work with organizations to make sure you’re solid in all four.
Using AI Safely
You can get almost all of the upside and avoid almost all of the downside by getting a few things right. This is the difference between AI as an asset and AI as your next incident report.
The biggest avoidable mistake: pasting client records, employee PII, financials, or patient and student data into a free public tool that may train on it. It’s a data-handling problem before it’s an AI problem – and it’s entirely preventable. A free Cybersecurity Gap Analysis confirms your data is ready before you turn AI on.
✓ Mind where your data goes. Is it used for training, and where is it stored?
✓ Keep a human in the loop. AI drafts, a person decides.
✓ Watch for shadow AI. No sanctioned tool means people use an unsanctioned one. Visibility beats prohibition.
✓ Write a one-page policy. Define your terms first, and review it regularly – it’s a living document, not a one-time PDF.
✓ Check it against your frameworks. HIPAA, FERPA, PCI, CMMC/NIST 800-171, cyber insurance, and the NIST AI RMF.
Most Asked Questions
Fancy a little Q&A? Dive into the questions we hear most.
They’re all large language models. The practical difference is where they live and how your data is handled. Copilot is built into Microsoft 365 and works inside your existing Microsoft data with enterprise protections. ChatGPT and Claude are standalone tools with free and paid tiers. For work, the tier matters more than the brand: a business tier keeps your data private and out of training; a free tier often does not.
It can be, with the right tool. The risk isn’t “AI” in the abstract — it’s pasting sensitive data into a free consumer tool that may train on it. Business tiers generally keep your data private. For regulated data, you may need a private deployment inside your own secure cloud. Rule of thumb: if you wouldn’t post it publicly, don’t put it in a free public tool.
No. If you can write a clear instruction in plain English, you can use AI. The skill that matters is being specific — which is exactly what the CRIT method gives you. That’s a communication skill, not a coding one.
Use CRIT: Context, Role, Interview, Task. The Interview step — telling the AI to ask you questions before answering — is the one most people skip and the one that most improves results. When a prompt works well, save it and reuse it.
Do the three-bucket task triage, pick one tedious weekly task, and try it with a business-grade tool for a week. Don’t try to transform the company in month one — that stalls. Prove one workflow, then expand. A Standalone AI Readiness Assessment shows where you stand and what’s realistic.
For most businesses, no — it changes what their time is spent on. It’s how CIT uses it internally: the AI aggregates data and context so our people do the higher-value work. Teams that feel threatened hide what’s working; teams that feel supported surface it.
An agent is an AI that can take actions on its own rather than only answering when asked. It’s more powerful and needs more oversight. Most organizations should get comfortable with basic AI assistance first; agents are a phase-two capability once tooling and governance are in place.
Keep it to one page: approved tools, what data can and can’t go in, when a human must review, and who to ask. Define your terms first, and review it twice a year. CIT’s AI Starter Series walks teams through building exactly this.
Individuals usually feel a difference within their first week on one workflow. A team running the 90-day plan typically has several reliable workflows and written guardrails by day 60, and demonstrable time savings by day 90. Speed depends far more on adoption and good first use cases than on which tool you pick.
CIT meets you where you are. Just exploring? An AI Readiness Assessment shows where you stand. Deploying? Managed Intelligence Services cover training, strategy, tooling, and governance as one program — including Microsoft Copilot rollout and the private, secure AI that regulated organizations need. Since 1992 we’ve helped Minnesota, Wisconsin, and Iowa organizations adopt new technology safely.
Know AI matters but not sure where to start?
A free Cybersecurity Gap Analysis is the fastest way to confirm your data is ready for AI before you turn it on.