Clear guides for the manager choosing training and the employee putting AI to work.
AI training for employees teaches staff to use the AI tools your company has approved on their actual work: giving the tool clear context, keeping sensitive information out, checking what comes back and staying accountable for the result. The training that tends to hold up is short, specific to each role and built around practice tasks reviewed against criteria everyone can see.
Read the guideA useful business prompt is a brief another person could inspect: the deliverable, the reader, the approved facts, the limits and what to do when something is missing. Start with the task, not a magic phrase, and judge the prompt by whether its answer can be checked.
Read the guideAI writing for business works when the tool produces the draft and a person keeps responsibility for the message, the evidence and the final edit. Give it a bounded source, check every claim against that source and edit for the reader before anything is sent.
Read the guideTo fact check AI output, pick out the claims that matter, open the actual source for each one, recompute any important figure independently and record what you changed. Keep the tool's confidence separate from the evidence: a fluent answer is not a source.
Read the guideAn AI training plan for employees sets out who is trained first, which tools and information are allowed, which practice tasks people do each week, who reviews the work and what evidence you look at after a month. Thirty days is enough to establish a small, reviewable routine; it is a plan, not a promise about results.
Read the guideA small-business AI policy names the approved tools, the information that must stay out of them, who checks the output and whom to ask when a case is unclear. Write it from the work your team actually does, then test it on one real workflow before anyone signs it.
Read the guideThis AI policy template gives you ten clauses to complete: purpose, scope, owner, approved tools, information rules, output review, decisions about people, disclosure, reporting and review. Each placeholder is a decision your company has to make, and the finished draft still needs review by the people who own your tools, data and approvals.
Read the guideAn AI skills assessment should ask people to do a role-relevant task with an AI tool and judge the work against criteria they can see in advance. Keep demonstrated work separate from confidence and completion, and say plainly what the result does not show.
Read the guideA team is ready to practise AI when it has an approved tool, clear information boundaries, a task with a visible review standard, and a manager who can make time and handle questions. Check those four things before buying more training.
Read the guideStart with one recurring task you already know how to judge, give the AI tool the material and limits a new colleague would need, and check the result against the source before it leaves your hands. What that first task is, and what checking means, depends on your role, which is what this guide and its role pages cover.
Read the guideAI is most useful in marketing for first drafts, channel variants, research summaries and campaign options built from a brief you supply. It is least safe when it is allowed to invent proof, so every claim, figure and comparison still needs a person to trace it back to a source.
Read the guideIn sales, AI is most useful for account research summaries, call preparation, follow-up drafts, CRM notes and objection practice. It should not invent facts about a prospect, promise outcomes or set prices, so a rep still checks every claim and owns every commitment.
Read the guideHR teams can use AI to draft and simplify policies, prepare onboarding material, summarise survey themes and answer routine questions from approved documents. Decisions about people, such as ratings, pay, discipline and dismissal, should stay with accountable humans, and employee data needs stricter handling than most other work.
Read the guideRecruiters can use AI assistants to draft job adverts, interview questions, outreach and candidate updates from a brief they supply. Screening, ranking and rejecting candidates is a different matter: it falls under specific regulation in several places, and decisions about candidates should stay with people who can explain them.
Read the guideFor people working in finance teams, AI is useful for drafting variance commentary, explaining and building spreadsheet formulas, summarising long documents and preparing questions for a forecast review. Every figure it touches still needs recalculating from the source, and it should never be the reason a number is believed.
Read the guideAccountants and bookkeepers can use AI to draft client emails, summarise guidance, write procedure notes, explain spreadsheet logic and suggest transaction categories for review. Professional judgement, confidentiality and responsibility for the numbers stay with the accountant, and tax or technical answers must be checked against primary sources.
Read the guideIn customer service, AI is most useful as the agent's own drafting and reading assistant: it summarises a long ticket, drafts a reply from the approved policy and adjusts the tone, and the agent checks every fact before anything reaches the customer. Buying a customer-facing chatbot is a separate decision with different risks, and this guide is not about that.
Read the guideAI helps project managers most with the writing and reading around a project: turning meeting notes into an action log, drafting status reports, stress-testing a plan for missing dependencies and summarising change requests. Commitments on scope, dates, budget and people stay with the project manager.
Read the guideProduct managers get the most from AI when they use it to work through evidence they already have, such as interview notes, support tickets and feedback, and to draft and challenge specs. Deciding what to build, for whom and why stays with the product manager, and every AI summary of customer evidence needs checking against the source.
Read the guideFor a people manager, AI does two jobs: it is a drafting and thinking aid for your own work, such as preparing one-to-ones, structuring feedback and summarising updates, and it is a capability your team needs to use safely. The second job matters more: set clear rules, protect time to practise and judge the work rather than the tool usage.
Read the guideIn business operations, AI earns its place on the documents and checks around a process: drafting and updating SOPs, comparing supplier quotes, cleaning up messy lists and finding gaps in a handoff. Automating a process end to end is a separate project, and it only works once the process is written down and someone owns it.
Read the guideThe best AI tools for work are usually the general assistants your team already has access to, approved for named tasks and learned through practice on real work. Start from the work and the accounts you pay for, not from a ranked list.
Read the guideTo use ChatGPT for work, sign in to the account your employer has approved, give it the context and source material for a specific task, ask for a defined format and check every fact before the output leaves your hands.
Read the guideGood Copilot training for employees starts with which Copilot each person actually has, tidies file permissions first, then uses Microsoft's free learning alongside practice on each role's real tasks in Outlook, Teams, Word and Excel.
Read the guideClaude for business works best when the team uses a company plan, keeps recurring work in shared projects with standing instructions and reference files, and checks every output against its source before it is used.
Read the guideGemini for Workspace puts Google's AI assistant inside Gmail, Docs, Sheets, Meet and Drive, plus the Gemini app. A team gets the most from it by signing in with work accounts on a qualifying edition, setting the admin controls and practising on real tasks.
Read the guidePrompt engineering is the practice of writing and refining the instructions you give an AI tool so the answer is useful, checkable and fit for its purpose. For most staff it means writing a clear brief: the task, the context, the source material and the shape of the answer, then reviewing what comes back.
Read the guideA prompt template is a reusable prompt with blanks for the parts that change: the task, the reader, the source material and the format. Good templates save retyping and make answers easier to review; they only work when the person filling them supplies real facts.
Read the guideThe best ChatGPT prompts for work are not clever one-liners; they are short briefs that name the task, the reader, the facts to use and the format you want. Below are ready-to-adapt prompts for common office tasks, each with the facts to supply and the check to run before you use the answer.
Read the guideA good Microsoft Copilot prompt states a goal, adds context, names the source Copilot should use and sets expectations for the answer. The difference from other assistants is the source: Copilot can work from your files, emails and meetings, so the best prompts tell it exactly which ones.
Read the guideFor most small and mid-size companies, AI for business means using general assistants such as ChatGPT, Claude, Gemini or Copilot on everyday language work, with clear rules about data and a named person checking the output. The hard part is not the tool. It is choosing the right tasks, setting boundaries and teaching people to review what comes back.
Read the guideA small business gets the most from AI by using one general assistant, such as ChatGPT, Claude, Gemini or Copilot, on a handful of repeated writing and summarising jobs, with the owner or a named colleague checking anything that reaches a customer. Start small, keep sensitive information out and expand only what proves useful.
Read the guideThe generative AI use cases that work best in business are drafting, summarising, rewriting, extracting and explaining text, where a person supplies the source material and checks the result. The use cases to avoid, or approach slowly, are those where nobody can check the output or where the output decides something about a person.
Read the guideAn AI workflow is a repeated task written down as steps: what goes in, what the AI does, what a person checks and what comes out. A good one can be followed by someone who did not write it and still produce work the reviewer would accept.
Read the guideIn most workplaces, AI means staff using general assistants and built-in AI features to draft, summarise and analyse, under rules set by the employer. It helps most where people check its output and know what they may share, and it goes wrong where it is used without consultation, training or review.
Read the guideAI adoption is the point where people use an AI tool on their real work, week after week, and the output holds up to review. It usually stalls not because the tool is poor but because people cannot see which of their tasks it helps with, are unsure what they may share, or never practised checking the output.
Read the guideAn AI hallucination is when a generative AI tool states something false, invented or unsupported as if it were true, such as a citation that does not exist or a figure that is not in your data. It happens because these tools predict likely text rather than look facts up, so the practical defence is to supply sources, ask for evidence and check anything that matters before you use it.
Read the guideAI email writing works best when you give the tool the recipient, the purpose, the facts and the one thing you want the reader to do, then check every promise, date and name before you send. Used that way, ChatGPT, Copilot in Outlook or Gemini in Gmail can produce a solid first draft; used without the facts, they invent them.
Read the guideAI is useful for workplace research when it helps you orient quickly, plan what to look for and summarise documents you supply, and risky when you treat its answer as the source. The reliable pattern is to let the tool find and organise, then open the sources yourself and confirm the claims that will go into a decision.
Read the guideAI for Excel means using an assistant such as Copilot in Excel, Gemini in Google Sheets or ChatGPT to write and explain formulas, clean data and suggest analysis. It saves effort on the parts of spreadsheet work that are fiddly, but every formula it writes needs testing on rows you can check by hand before you trust the total.
Read the guideAI can speed up data analysis by suggesting approaches, writing formulas or code, producing charts and drafting a first interpretation, but it cannot tell you whether the data answers your question or whether a pattern has a cause. Use it for the mechanics, keep the question and the conclusions with a person, and check every number it reports against an independent calculation.
Read the guideFor most employees, AI skills are the practical abilities to brief an AI tool well, use it on a real task, check what it produces and know when not to use it. They are built by doing role-relevant work with feedback, not by learning how models are made.
Read the guideAI literacy is the knowledge and judgement to use AI tools in an informed way: understanding roughly how a tool produces its answers, where it goes wrong, and what you remain responsible for. At work it shows up in small decisions, such as what to paste into a tool and what to check before sending.
Read the guideAI upskilling means helping existing employees do their current jobs better with AI, through practice on their own kinds of tasks rather than general awareness sessions. For a small team, a good programme starts with one approved tool, one task per role and a clear standard for reviewing the work.
Read the guideAn AI acceptable use policy is the day-to-day rulebook for employees: which AI tools they may use, for what, with which information, and what they must check before using the result. The clearest versions sort uses into allowed, needs approval and not allowed, with an example of each from the team's own work.
Read the guideA generative AI policy sets rules for tools that produce new content, such as ChatGPT, Claude, Gemini, Copilot and image generators: which accounts staff may use, what goes into a prompt, how generated content is checked, when it is disclosed and who owns the final version. It covers the risks particular to generated content, especially confident errors, copyright and over-reliance.
Read the guideAn AI governance framework is the set of roles, records and routines that decide which AI uses a company allows and keeps those decisions current. For a small business it can be one named owner, a one-page register of approved tools and workflows, an approval step for new uses and a review on a fixed date, loosely mapped to the NIST AI Risk Management Framework.
Read the guideBelow are two published public-sector examples, three worked policies for fictional small companies and a set of example clauses you can adapt. The examples show how the same few decisions, approved tools, information limits, output checks and a named owner, get written differently depending on the work a team does.
Read the guideThe useful question is not whether a tool can produce an answer. It is whether your team can recognise a good answer, challenge an unsupported claim and decide when human judgement has to take over. These guides build from that starting point.
Our first collection covers choosing employee training, writing clearer prompts, reviewing drafts, checking facts and planning a rollout. The examples use fictional business details. They are designed for practice, not to demonstrate a measured productivity improvement.
Explore the planned role-specific lessons, practical tasks and manager view.
