Language: English
Start 14-day free trial

Guides

Guides to sharing AI skills and using AI at work

Start with sharing: how to give your team's AI skills to everyone's Claude, ChatGPT or coding agent. Then use AI at work step by step, from a first task to choosing a tool. AI terms, business books and thinkers are explained in plain words. Every guide answers its question in the first paragraph and lists its sources with the date they were read.

Share AI skills with your team

Use AI at work, step by step

Just starting with AI

  • Start using AI at workStart with one task you already do every week, on a free plan, with nothing confidential in it.
  • ChatGPT vs Claude for workChatGPT and Claude both draft, summarize and answer well, cost the same on the personal paid plan, and both have free plans.
  • What not to paste into AINever paste passwords, other people's personal data, customer files, unreleased numbers, or anything under an NDA into a personal AI account.
  • Weekly report with AIPaste your rough notes, give the AI a fixed shape (done, next, stuck), and ask it to flag what's missing.
  • Prospecting email with ChatGPTGive ChatGPT one specific fact about the prospect, the one problem you solve, and a one-line question, and keep it under 90 words.
  • Sales follow-up email with ChatGPTPaste your call notes and ask ChatGPT for a same-day recap: three bullets in the buyer's words, one next step with a date, one question.
  • Marketing content plan with ChatGPTTell ChatGPT your goals, your real hours, and your fixed dates, and ask for one main piece a week, each cut into two to four smaller posts, every row with a goal and an owner.
  • Job description with ChatGPTGive ChatGPT the role brief, a real pay range and the first six months of work, and ask for an honest title, at most five must-haves and plain language.
  • Candidate outreach with ChatGPTGive ChatGPT one specific thing the candidate made or did, the role in a line with the pay range, and one easy ask, and get a message under 400 characters plus two follow-ups.
  • ChatGPT for HR professionalsThe HR jobs ChatGPT does well (an internal FAQ from the questions people keep asking, job posts, policy summaries, interview questions), what to give it for each, and what stays with you.
  • ChatGPT for accountantsWhere ChatGPT helps accountants and bookkeepers: a month-end close checklist with an owner, a reviewer and a day for every task, variance questions, client emails.
  • Performance review draft with ChatGPTGive ChatGPT the expectations in your company's words and dated examples from the whole period, and ask it to rate each expectation separately, write growth areas as behaviours with a next step, and check for recency, halo and vague praise.
  • Pipeline review with ChatGPTPaste a deal's notes and ask ChatGPT to score each MEDDICC or BANT element 0, 1 or 2 from what the buyer said, rank the gaps, give one dated next action per gap, and suggest a forecast category with the rule behind it.
  • Sales call prep with ChatGPTPaste what you know about the prospect and ask ChatGPT for a one-page call plan: a three-line snapshot with sources, two or three hypotheses, eight ranked questions, what you must learn to qualify, red flags, and the opening and close.
  • Call notes to CRM with AIPaste a call transcript or rough notes and ask ChatGPT for a CRM note in plain text: the buyer's pain and its cost kept apart, budget, authority and timeline only as stated, one dated next step, risks, and a stage the evidence supports.
  • Interview questions with ChatGPTGive ChatGPT the role's competencies and ask for one main question each, three follow-up probes, what a strong and a weak answer sound like, and a check against questions you must not ask.
  • Ad copy variants with ChatGPTAsk ChatGPT for three to five ad variants, one angle each, with the offer, call to action and landing page held the same, written to the platform's character limits, plus a test matrix.
  • Marketing strategy with ChatGPTBefore plans and campaigns, ask ChatGPT to work through positioning with April Dunford's five components: what customers would do without you, what only you have, the value it gives, who cares most, and the market category.
  • Financial analysis with ChatGPTAsk ChatGPT to review the assumptions in your forecast rather than build one: every assumption with its source, results typed in as numbers, one-at-a-time sensitivity, plausibility against your history, and the three that matter most.
  • AI for startup foundersThe most useful first job for a startup founder's AI is a decision memo: the recommendation and the ask first, only the context needed, two to four options with their trade-offs, the risks, and a decider and deadline.
  • Budget variance note with AIHave ChatGPT or Claude write the monthly budget variance note from your budget-vs-actual table: a threshold first, each variance favorable or unfavorable, its driver (timing, volume, price, one-off), the full-year impact, a headline and questions for line owners.
  • A PRD from notes with AIPaste your notes, Slack threads and call notes, and have ChatGPT or Claude turn them into a PRD: the problem with its evidence, the users, goals and non-goals, a metric with a baseline and target, must, should and could requirements, open questions, and every guess marked.
  • Checking invoices with AIHave ChatGPT or Claude check a supplier invoice against the purchase order before anyone pays: the supplier's details, bank details first, the arithmetic redone, what was ordered and received, tax and terms, duplicates, and approve or hold with the reason.
  • Landing page copy review with AIHave ChatGPT or Claude review your landing page copy with six concrete tests (the headline on its own, so-what, the opposite test for adjectives, proof next to each claim, every button, one job per section) and return ranked line edits you can paste.
  • Objection handling with AIUse ChatGPT or Claude to prepare for sales objections before a call: classify the objection, acknowledge it, ask one question to find what's behind it, respond to the real issue, confirm, and keep any discount for last and in exchange for something.
  • User stories with AIHave ChatGPT or Claude turn a feature into user stories a team can build and test: one role per story, a real "so that", slices a user can try, Given/When/Then acceptance criteria, edge cases, out-of-scope notes and an INVEST check.
  • Ticket triage with AIHave ChatGPT or Claude triage a batch of support tickets in one fixed format: one category, an S1 to S4 severity from a written rubric, an owner, a short first reply that promises no fix, and the table sorted by what to answer first.
  • User interview synthesis with AIHave ChatGPT or Claude synthesize user interview notes into themes the team can act on: observations tagged did, said or wants, themes with participant counts, verbatim quotes, jobs-to-be-done, insights kept apart from assumptions, and recommendations with a confidence level.
  • Unit economics check with AIHave ChatGPT or Claude recompute your unit economics from your own numbers: CAC with what spend was included, gross margin, LTV on margin not revenue, LTV to CAC, CAC payback, and the definitional mistakes that inflate them.
  • OKRs with AIHave ChatGPT or Claude draft quarterly OKRs that hold up: one to three objectives, two to four key results each from a baseline to a target, outputs moved to initiatives, committed or aspirational, one owner per key result, and a counter-metric for anything that can be gamed.
  • Meeting agendas with AIHave ChatGPT or Claude turn a meeting's goal into an agenda that gets results: one outcome sentence, topics rewritten as questions, each tagged decide, discuss or inform with an owner and minutes, pre-reads, optional attendees, and a recommendation to cancel when nothing needs deciding.
  • Rejection emails with AIHave ChatGPT or Claude write a candidate rejection email that fits the stage: short after an application, personal after interviews, a call first after a final round.
  • Bug reports from tickets with AIHave ChatGPT or Claude turn a customer's ticket into a bug report an engineer can reproduce without reading the thread: a title that names the failure, numbered steps, expected and actual kept apart, the environment with unknowns marked, impact, and the questions to ask the customer.
  • SaaS spend review with AIHave ChatGPT or Claude review your software subscriptions: seats paid against seats used, tools that do the same job, the last day to act before each renewal, and a cancel, downgrade or keep for each tool with an owner and the saving.
  • Vendor comparison with AIHave ChatGPT or Claude compare two to five vendors the way a careful buyer does: criteria and weights agreed before scoring, must-haves as pass or fail, scores with their evidence, total cost over the contract term, and claims only the vendor makes marked unverified.
  • Product-market fit check with AIHave ChatGPT or Claude review your path to product-market fit with Dan Olsen's five-layer pyramid: target customer, underserved needs, value proposition, feature set and user experience, each with its evidence graded.
  • Contract summaries with AIHave ChatGPT or Claude turn a client's contract into a one-page brief for a freelancer: what you're paid and when, how it ends, the indemnity and IP terms in plain words, unusual clauses quoted, every deadline as a date with its section number, and questions for a lawyer.
  • SEO briefs with AIHave ChatGPT or Claude write a content brief for one search query: the intent checked against what ranks, the answer for the first paragraph, an outline that covers what's expected and adds what the ranking pages lack, questions, terms, internal links, and a verify mark on every number it couldn't check.
  • Newsletters with AIHave ChatGPT or Claude turn a list of updates into a newsletter issue people read: one lead story about what readers can now do, three to five short items, one call to action, subject lines and a preheader, and the footer US law asks for.
  • Offer emails with AIHave ChatGPT or Claude draft the offer call script and the same-day recap email from the approved offer: role, start date and pay with every number checked, the conditions and when the offer becomes unconditional, a deadline as a date, and a line that the formal letter governs.
  • Release notes with AIHave ChatGPT or Claude turn merged changes into release notes customers read: New, Improved and Fixed, each line starting with what users can now do, fixes described by the symptom users saw, and an internal changelog with ticket IDs for everything else.
  • Internal announcements with AIHave ChatGPT or Claude write a change announcement people act on: the change, the date and the action in the first two sentences, what changes and what doesn't, the reason in plain words, a dated action for each group, and one named person to ask.
  • NDA first pass with AIHave ChatGPT or Claude give a client's NDA a first pass: mutual or one-way, how broad the definition is, which exclusions are missing, how long it lasts, and clauses that don't belong in an NDA, in a green, yellow and red table with each clause quoted and a question for counsel.
  • ChatGPT Work explainedChatGPT Work is the mode for handing ChatGPT a whole task: you describe the outcome, it plans the steps, uses your files, plugins and approved tools, and returns a result you review, such as a deck or a spreadsheet.
  • Payment reminders with AIHave ChatGPT or Claude write invoice reminders that get paid without souring the relationship: the invoice facts first, a tone that firms up only as the delay grows, consequences only if the contract or the law allows them, and disputes answered instead of chased.
  • Support replies with AIHave ChatGPT or Claude draft a support reply that answers in the first sentence, says no plainly when the answer is no, owns a mistake once, uses the names on screen, and ends with one next step.
  • Weekly client reports with AIHave ChatGPT or Claude turn your notes into a weekly client report a client reads in a minute: on track, at risk or off track with the reason, what you need from them by when, what's done as outcomes, next week's commitments and the agreed metrics.

Using AI every day

  • Stop pasting the same promptFind the prompt you paste most, split it into what applies to you, to one project and to one task, and save each part where your AI reads it on its own: custom instructions, a project or a skill.
  • Custom instructions vs Projects vs skillsCustom instructions shape every chat, a project holds one piece of work with its files, and a skill is a method the AI loads only when a task needs it.
  • Claude Projects explainedA Claude project is a workspace with its own files, instructions and chat history, on every plan (Free up to five).
  • Sales prompts, savedFour ChatGPT and Claude prompts sales reps reuse every week (account research, a first email, call prep, call notes to CRM), written to copy, and where to save each so you stop pasting: custom instructions, a project per account, or a skill.
  • Marketing prompts, savedFour ChatGPT and Claude prompts marketers reuse (a content brief, social posts from an article, ad variants, a campaign readout), written to copy, plus where to save each: your brand voice in custom instructions, a campaign in a project, a repeated task as a skill.
  • HR prompts, savedFour ChatGPT and Claude prompts HR people reuse (a job post, structured interview questions, an onboarding plan, a policy FAQ), written to copy without personal data, and where to save each so you stop pasting: custom instructions, a project per role, a skill per task.
  • Finance prompts, savedFour ChatGPT and Claude prompts finance people reuse (a variance note, a month-end checklist, a payment reminder, a board summary), written to copy, plus what never to paste and where to save each: custom instructions, a project per month, a skill per task.
  • Custom instructions examples for workFour custom instructions examples for work (account manager, operations lead, recruiter, founder), each in the same short shape: role and goal, audience and tone, what not to do and what to do instead, output format, one edge case.
  • Meeting minutes prompt, savedA meeting minutes prompt that puts decisions and action items first (one owner, a verb and a date each), then open questions, a short discussion and what wasn't covered.
  • Claude project for operationsSet up one Claude project for operations and admin work: instructions with your role, tools and rules, knowledge files with the SOPs and lists people keep asking about, and a brief that turns each walkthrough into an SOP.
  • ChatGPT prompts for freelancersFour prompts freelancers reuse (a proposal from a client's brief, a scope with what's not included, a polite invoice reminder, a weekly client update), where to save each so you stop pasting them, and a worked example of turning the proposal prompt into a skill.
  • ChatGPT prompts for project managersFour prompts project managers reuse (a one-page kickoff brief, a weekly status update, a risk list with owners, a meeting agenda), where to save each so you stop pasting them, and a worked kickoff brief with one success measure, scope out, dated milestones and one accountable owner.
  • Claude project for foundersSet up one Claude project as a founder: instructions with your company's stage and rules (never invent a number), knowledge with your metrics, last updates and plan, and the monthly investor update as its first job.
  • Reusable prompts in CursorIn Cursor, save a prompt you reuse as a skill (.cursor/skills, run with /name or attach with @name), keep short always-on guidance as a rule (.cursor/rules), and convert old slash commands with /migrate-to-skills.
  • Reusable prompts in CopilotIn GitHub Copilot, repository rules go in .github/copilot-instructions.md, a reusable prompt in VS Code goes in a .prompt.md file run with /name, and a procedure that should work across Copilot's surfaces goes in an agent skill.
  • ChatGPT Projects explainedA ChatGPT project keeps related chats together with shared instructions and files, so every chat in it starts with the same context.
  • Claude project for salesSet up a Claude project for a territory or key account: instructions with your product, buyers and rules (no discounts the AI invents), knowledge with the one-pager, price list, case studies and account notes, and a brief for a 90-day account plan.
  • Claude project for marketingSet up a Claude project for marketing content: instructions with your audience, goals and rules (no invented stats or testimonials), knowledge with the voice guide, product facts and fixed dates, and a brief for a month's content plan within your real hours.
  • Claude project for recruitingSet up one Claude project per open role: instructions with the hiring rule and job-related-only rules, knowledge with the role scorecard and interview scorecard, and a brief that turns interviewers' notes into a candidate summary the debrief can trust.
  • ChatGPT vs Claude ProjectsChatGPT and Claude projects do the same job: instructions, files and chats kept together, so every chat starts with the same context.
  • Claude project instructions examplesFour Claude project instructions examples for work (a support lead, a finance analyst, a consultant and a marketing manager), each in the same short shape: role and goal, audience and tone with an example phrase, can and can't, sources of truth, format, edge cases and hand-off.
  • ChatGPT Projects vs custom GPTsA custom GPT is a separate assistant you open and share by link, and it retires on 11 December 2026.
  • Claude project for financeSet up one Claude project for finance reporting: instructions with your KPI definitions and rules (never invent a number), knowledge with the definitions, targets and last months' reports, and a brief for the monthly KPI summary: changes worked out, status against target, reasons only from evidence, and up to three things to act on.
  • ChatGPT memory for workChatGPT memory carries context from earlier chats into new ones, through saved memories and reference chat history, managed in Settings > Personalization.
  • A glossary for your AIAn AI uses your company's words when it reads a short glossary: one preferred term per concept, a plain definition, the words to avoid, usage notes and an owner, plus a Use / Don't use list to paste into instructions.
  • Custom instructions character limitChatGPT's custom instructions hold 1,500 characters on Free and Go and 5,000 on Plus, Pro, Business, Enterprise and Edu, per OpenAI's help page.
  • Brand voice in custom instructionsMake ChatGPT write in your brand's voice in every chat: a short voice guide in custom instructions (three or four traits, words to use and avoid, one sentence that sounds like you), built from texts that already sound right, with the full guide in a project.
  • Custom instructions for accountantsWhat accountants and bookkeepers put in ChatGPT custom instructions: role, currency and number rules, never invent a figure, flag rather than decide, and plain words for owners.
  • Custom instructions for HRWhat HR people put in ChatGPT custom instructions: role and tone, job-related criteria only, no guesses about protected traits, a check mark on anything about pay, leave or discipline, and policies quoted from a project.
  • Custom instructions for salesWhat sales reps put in ChatGPT custom instructions: what you sell and to whom, facts only from your research, no prices or discounts without approval, and the lines every cold email needs by law.
  • Gemini Gems vs ChatGPT ProjectsA Gemini Gem is a custom version of Gemini for one task you repeat; a ChatGPT project keeps one piece of work's chats, instructions and files together.
  • ChatGPT prompt library extensionsA prompt library Chrome extension adds your saved prompts to chatgpt.com, and to do that it asks to read and change data on the site, your chats included.
  • Claude prompt libraryAnthropic's prompt library page now redirects to its Prompting best practices guide, which teaches the patterns instead of listing prompts.

Bringing your team along

  • Notion contractors: guest or paid memberNotion adds anyone on your company's email domain as a paid member, not a guest, even under the guest limit.
  • Get your team to use AI the same wayAgree on three shared tasks, write down how each is done, keep one copy of the instructions everyone's AI uses, and name an owner.
  • Teach a new hire your AI workflowGive a new hire the team's AI account, the one-page rules, and the shared instructions for their first three tasks before their first morning ends.
  • AI playbook for a sales teamA sales team gets consistent results from ChatGPT or Claude when it agrees four jobs (account research, first emails, call notes, deal reviews), shares one set of instructions for each, and sets three rules on customer data and pricing.
  • AI rules for employees (HR)An employee AI policy written by HR covers approved tools, the people data that never goes into AI, the hiring and performance decisions AI may not make alone, disclosure, and training.
  • AI training for employeesTrain employees on AI with three 45-minute sessions on their own work: the rules and one real task, the team's shared instructions, and checking AI output.
  • AI usage policy templateA one-page AI usage policy template to copy and fill in: approved tools, what never goes into AI, review before anything leaves the company, disclosure, new tools, incidents, training and an owner.
  • AI pilot plan for a teamAn AI pilot that ends in a decision: one or two tasks, three to six people, a baseline measured before you start, success and stop criteria agreed up front, four weeks, and a decision date.
  • AI for onboarding new employeesUse AI to prepare a new hire's onboarding, not to replace the people in it: a 30/60/90-day plan, a first-week schedule, a team glossary and answers to the questions every new hire asks.
  • Client data and AI in financeOnly on a business plan that doesn't train on your data, within your engagement terms and professional rules, and never in a personal account.
  • One reply tone for a support teamA support team's AI replies sound the same when everyone's AI uses one tone guide and three real replies, not each agent's own prompt.
  • AI use disclosure to clientsA one-page note an agency or freelancer sends clients before work starts: which AI tools and accounts you use and for what, what client data never goes in, who reviews the work, when you'll say AI was used, and what the client owns.
  • Rules for AI-written marketing claimsAI writes confident claims fast: "the fastest", "proven", "trusted by thousands".
  • Expense policy checks with AIHave AI check each expense claim against your written policy before an approver sees it: every line marked pass, flag or fail with the rule quoted, missing receipts listed, the out-of-policy amount totalled, and one question per flag.
  • Spec review with AIHave AI review each spec before engineering starts: walk the flow as a first-time user, check each screen's empty, loading, error and no-permission states, flag vague words, check limits, metrics and rollback, and rank findings as blocking, should fix or nit.
  • Contractor AI checklistBefore a freelancer or contractor uses AI on your work, check the agreement: which tools and accounts they may use, what of yours never goes in, who reviews the output, what ownership the contract actually transfers, and the usual scope, payment and exit terms.
  • Team meeting minutes formatGive every meeting's minutes the same shape, whoever runs it: decisions first, actions with one owner, a verb and a date, open questions, a short discussion and what wasn't covered.
  • Brand voice for a team's AIA marketing team's AI writing sounds like one brand when everyone's AI reads the same short voice guide, built from real samples: measured patterns, the four tone dimensions, three or four this-not-that traits each proved with a quoted line, a word list and the mechanics.
  • Operations team AI playbookAn operations team gets consistent results from AI when it agrees, on one page, which tasks AI helps with and under what rules: a scored scan of repeated tasks, the top three with the simplest approach, two-week experiments with a baseline and a stop rule, three rules and one owner per task.
  • One interview scorecard for the teamA hiring team compares candidates fairly when every interviewer rates on one scorecard built before the first interview: four to six competencies, 1 to 4 anchors that describe an answer, evidence fields, ratings made alone right after the interview, and a hire rule frozen in advance.
  • Tracking a team's AI useClaude Team and Enterprise have a usage analytics dashboard (weekly active members, seats, skills and connectors used, spend); OpenAI documents user analytics for ChatGPT Enterprise and Edu.
  • Reviewing AI drafts before sendingA team stops AI mistakes reaching clients when every AI draft that leaves the team gets the same check: does it answer the request, which claims are supported, checkable or suspect, any invented sources, promises and tone, internal consistency, and a ranked fix list with a verdict.
  • Sales onboarding with AIGet a new sales rep up to speed with the team's shared AI setup: battlecards for the main competitors, learned by role-play with the AI playing the buyer, then real call prep with a manager reviewing the drafts.
  • Marketing onboarding with AIGet a new marketer writing in your brand's voice from week one: the team's brand-voice skill, the claims rules and the content project on day one, then a real first piece drafted with the voice skill and checked for claims before an editor reads it.
  • Internal FAQ with AITurn the questions people keep asking in Slack and email into an internal FAQ: AI clusters and counts them, answers only from official sources with a link, flags what needs a leader, and every answer gets an owner and a review date.
  • Share Gemini Gems with a teamYes, you can share a Gemini Gem: click Share, add people as Viewer or Editor, or copy a link.
  • Share Notion outside the workspaceShare a Notion page with a client or contractor three ways: invite them as a guest from Share, send a link anyone can open that expires, or publish it to the web.
  • Team prompt library templateA team prompt library works when every entry has the same fields: a task name, when to use it, the inputs, the prompt, a good example, an owner and the date last tested.
  • Transfer a custom GPTInside a ChatGPT Business or Enterprise workspace, an owner can reassign a custom GPT to another member, and merging a personal workspace brings your own GPTs along.
  • Notion guest permissionsA Notion guest can edit a page with Can edit or Full access, and only Full access lets them share it.
  • Share prompts in CopilotIn Microsoft 365 Copilot you save a prompt after it runs, then share it to a Microsoft Teams team from the prompt gallery, where teammates find it under Team prompts.
  • Share prompts in LibreChatIn LibreChat a prompt's owner shares it with users, groups or roles as Viewer, Editor or Owner, once their role has the share permission, and teammates pick it from the library or with a slash.
  • Claude Team setup guideSetting up Claude Team comes down to seven steps for the owner: seats and invitations, roles, single sign-on, the capabilities skills need, connectors, the first organization skills and the feedback setting.
  • ChatGPT Business setup guideSetting up ChatGPT Business comes down to six steps for the owner: seats and billing, invitations and roles, single sign-on, plugins and connectors, the team's first skills, and a plan for the custom GPTs that retire on 11 December 2026.
  • Admin vs Owner on team plansOn Claude Team, Admins invite and remove members, while Owners also handle billing, seats, integrations, capabilities, skills for the organization and roles, and the one Primary Owner alone requests data exports.
  • Cursor team rulesCursor Team Rules are instructions an admin writes once in the dashboard that apply to every member's Agent chats across all the team's repositories, optionally enforced so nobody can switch them off.
  • Claude for accounting firmsBefore client work goes into Claude, an accounting firm puts it on a business plan, which Anthropic doesn't train on by default, turns off feedback ratings, and checks its confidentiality rules with counsel (in the US, IRC section 7216 for tax return information).
  • Claude for marketing agenciesA marketing agency gets most from Claude with one project per client, holding that client's brand guide and past work, and skills for the methods it uses on every client: the voice rewrite, ad variants, the content calendar, the weekly client report and a claims check.
  • Claude for small law firmsBefore client matters go into Claude, ABA Formal Opinion 512 asks a lawyer to judge, for each AI tool, who could see client information, to read the tool's terms, and to get informed consent before using a tool that learns from what goes in.
  • Can't share a custom GPTOn a personal ChatGPT account (Free, Go, Plus, Pro), greyed-out GPT sharing is expected: these accounts can't create or publish new GPTs.
  • Share a Claude Project with your teamClaude projects are shared on Team and Enterprise: Share, add people from your organization, Can view or Can edit.
  • Share a ChatGPT projectEvery ChatGPT plan, Free included, can share a project: Share, invite people with chat or edit access, or let the owner open it to anyone with the link.

Making AI skills

  • Gemini Gems to skills before 17 NovemberOn personal Google accounts, Gemini Gems turn read-only on 17 November 2026 and are removed on 17 December.
  • Make your first Claude skillTake a doc you already follow (a checklist, a how-we-do-it page, a template), give it a name and a 'use when' description, put it in a SKILL.md file, zip it and upload it in Claude's Customize > Skills.
  • Create skills in Claude CodeMake a folder in ~/.claude/skills (just you) or .claude/skills (the project, shared through git), add a SKILL.md with a description that says when to use it, and Claude Code loads it when a request matches or when you type /name.
  • Claude skills for salesThe five Claude skills a sales team gets most from: account research, first emails, discovery call prep, call notes to CRM, and deal reviews.
  • Claude skills for marketingStart with a brand-voice skill, then add content briefs, social posts from articles, ad copy variants and campaign reviews.
  • Claude skills examples for businessExamples of Claude skills a small business uses, by team: sales, marketing, operations, finance, HR and support, plus Anthropic's own example skills.
  • Deliver AI skills to clientsAn agency can hand clients its method as AI skills their own Claude or ChatGPT runs: a reporting format, a brand-voice guide, a content process.
  • Custom GPT vs Claude skillsA custom GPT is a separate chatbot you open; a Claude skill is a method Claude loads by itself when your request needs it.
  • ChatGPT skills vs Claude skillsChatGPT and Claude both read skills in the open Agent Skills format (a folder with SKILL.md).
  • Claude skills for financeFive Claude skills a small finance team gets most from: a budget variance note, a month-end close checklist, an invoice check, a payment reminder and a monthly KPI summary.
  • SKILL.md vs AGENTS.mdAGENTS.md holds a repository's always-on instructions for coding agents (build, test, conventions); a SKILL.md is one task's method, loaded only when a request needs it.
  • Claude skills vs commandsIn Claude Code, custom commands have been merged into skills: a file in .claude/commands and a skill folder both create the same /name, and old command files keep working.
  • Company context for AIThe one page of company context an AI assistant needs: who you serve, what you sell, your words for things, what it must never promise and what to do instead, which document wins for which topic, and the format you want.
  • Claude skill-creatorskill-creator is a skill from Anthropic that drafts a skill with you and tests it on prompts you agree on.
  • Skills in Claude CoworkClaude Cowork loads the skills, plugins and connectors turned on for your claude.ai account, synced when a session starts; it doesn't read Claude Code's ~/.claude folder.
  • Claude skills for operationsFive Claude skills that cover most of an operations team's repeated writing: SOPs from walkthroughs, meeting minutes in one format, an internal FAQ from the questions people ask, a scan for work worth automating, and vendor comparisons.
  • Founder playbook as AI skillsA founder's playbook (how you quote, onboard a customer, handle a complaint) becomes usable when each repeated method is one AI skill: when to use it, the steps, the output's shape and an example.
  • Claude skills for recruitingFive Claude skills cover most of a hiring team's repeated writing: the role intake, the job post, the interview scorecard, the candidate summary and the reference check.
  • Claude skills librariesWhere to find ready-made Claude skills: Anthropic's own skills, partner skills in the Skills Directory, Anthropic's skills repository on GitHub, Claude Code plugin marketplaces, the skills.sh directory, and team libraries.
  • Claude skills vs MCPMCP connects Claude to outside tools and data; a skill teaches Claude how to do a task.
  • Claude skills vs agentsA skill teaches Claude how to do a task inside the chat you're in.
  • The SKILL.md formatA skill is a folder with a SKILL.md file: YAML front matter with a name and a description, then Markdown instructions, plus optional scripts, references and assets.
  • Writing a skill descriptionThe description decides whether an AI ever uses a skill: it's the only part loaded before a task, and Claude picks among many skills by it.
  • A sales email skillTurn your best rep's first emails into one skill every rep's AI uses: one specific observation, one problem, one line of proof if you have it, an interest ask, under 90 words, two subject lines, and the list rules.
  • A freelancer's skill libraryA freelancer gets more from AI with a small skill library: method skills for how you work, kept once and used for every client, and client facts kept apart, in one project or skill per client.
  • Share a landing page skillShare one landing page skill with your team in AI Skill Pod: one invite link, free to start, and everyone's Claude, Cursor or ChatGPT (Pro, Business, Enterprise) builds pages the house way: voice, section order, one call to action, form fields, claims that need a source, alt text, contrast and preview first.
  • AI skills for HRFive AI skills cover most of HR's repeated writing after the offer is signed: a 30/60/90 onboarding plan, a policy FAQ from the questions people ask, 1:1 prep, feedback in Situation-Behavior-Impact form, and a performance review draft with a bias check.
  • Move a custom GPT to ClaudeCustom GPTs stop running on 11 December 2026.
  • GPT knowledge files into skillsA custom GPT holds up to 10 knowledge files, and GPTs stop running on 11 December 2026.
  • GPT Actions after retirementWhen custom GPTs retire on 11 December 2026, their Actions don't transfer.
  • Gems on Google WorkspaceOn Google Workspace accounts, Gems stay until 2027: business, enterprise and non-profit accounts move to skills in March 2027, education in June.
  • Gem examples as skillsFive Gem examples for work, from meeting notes to client replies, each rewritten as a skill before personal Gems turn read-only on 17 November 2026.
  • Custom GPTs vs Gemini GemsCustom GPTs and Gemini Gems both hold instructions and files for a task you repeat, and both are being replaced by skills: Gems on personal accounts turn read-only on 17 November 2026, and GPTs retire on 11 December 2026.
  • Claude organization skillsOn Claude Team and Enterprise, an owner uploads a skill in Organization settings and it turns on for everyone at once; members can switch it off but not delete it.
  • Claude skills not showing upA Claude skill that doesn't appear is usually switched off, missing code execution, blocked by the organization, or uploaded without a valid SKILL.md; one that appears but isn't used usually has a vague description.
  • Claude skills for customer supportThe Claude skills a support team gets most from are five: triage of incoming tickets, replies in your team's tone, bug reports engineers can act on, incident notices, and an FAQ built from repeated questions.
  • Claude skills for customer successThe Claude skills a customer success team gets most from are five: an onboarding plan for each new account, call notes for the CRM, a churn-risk note, a QBR outline and an expansion plan.
  • Claude skills for product managersThe Claude skills a product team gets most from are five: a PRD from messy notes, a synthesis of user interviews, a scored backlog, user stories with acceptance criteria, and a roadmap update for stakeholders.
  • Claude skills for consultantsThe Claude skills a consultant or small consultancy gets most from are five: a check of each proposal before it goes out, a kickoff brief, meeting minutes with owners and dates, a weekly client report, and a decision memo with the recommendation first.
  • Install Claude skillsIn Claude on the web or in the desktop app on Mac or Windows: Customize > Skills, +, + Create skill, Upload a skill, then a ZIP with the skill's folder at the top.
  • Install skills in Claude CodePut the skill's folder in ~/.claude/skills (Mac and Linux) or %USERPROFILE%\.claude\skills (Windows) for all your projects, or in a repository's .claude/skills for everyone who works there.
  • Use skills in CodexCodex reads skills from $HOME/.agents/skills and from .agents/skills in a repository, each a folder with a SKILL.md.
  • Install Claude skills from GitHubA repository that's a plugin marketplace: add it in Claude under Customize > Plugins > Add marketplace, or in Claude Code with /plugin marketplace add owner/repo.

Choosing how to share skills

  • Claude Team vs Pro for a small teamClaude Team costs $25 a seat billed monthly ($20 yearly) for 2 to 150 people; Pro is $20 a person billed monthly.
  • Claude Team vs EnterpriseClaude Team is for 2 to 150 people at $25 a seat billed monthly; Enterprise is $20 a seat billed yearly with a 20-seat minimum plus usage at API rates.
  • ChatGPT Team (Business) vs PlusChatGPT Team was renamed ChatGPT Business in August 2025.
  • Share Claude skills on GitHubCommit skills to a project's .claude/skills folder, publish them as a Claude Code plugin marketplace on GitHub, or keep a plain repo people download from.
  • Serve skills over MCP to Claude DesktopYes: run an MCP server that serves your skills and add it in Claude as a custom connector (every plan; Free allows one), or as a local server in Claude Desktop's config.
  • Free prompt management toolsFree plans of PromptHub, PromptLayer and Langfuse, a shared doc, and a free skill library compared: what each free plan allows (private prompts, users, limits), who it's built for, and which fits a team that uses ChatGPT and Claude in chat.
  • Stop skill version driftSkills shared as ZIP files drift: each person keeps the version they uploaded.
  • Share a ChatGPT plugin outside your workspaceChatGPT Business shares plugins only inside the workspace, and Pro and personal accounts can't share Site-hosted plugins at all.
  • AI automation tools for opsZapier, n8n and Power Automate for an operations team: free plans, prices and who each suits, read from each vendor's pricing page, plus where a chat AI with shared skills fits instead.
  • AI tools for small-business financeQuickBooks' Intuit Intelligence, Xero's JAX, Copilot in Excel and a general AI assistant compared on what each does for a small finance team, from each vendor's own pages.
  • ChatGPT Business vs Claude TeamChatGPT Business and Claude Team compared from each vendor's own pages: who each is for, training on your data, admin and SSO, shared projects, shared skills, custom connectors, and what happens to custom GPTs.
  • ChatGPT Business vs EnterpriseChatGPT Business and Enterprise compared from OpenAI's own pages: what both include, what Enterprise adds (SCIM, role-based access, data residency, the Compliance API), how custom connectors and custom GPTs differ, and how to decide.
  • Claude Max vs TeamClaude Max is one person's plan with five times Pro's usage; Claude Team is for 2 to 150 people, with shared projects, skills deployed to the organization and no training on your content by default.
  • Copilot agents vs custom GPTsAgents built in Microsoft 365 Copilot's Agent Builder and ChatGPT's custom GPTs compared from each vendor's docs: what you configure, what knowledge each can use and which licence that needs, how sharing works, and what happens to custom GPTs after 11 December 2026.
  • ChatGPT plans for small businessChatGPT has four plans for one person (Free, Go, Plus, Pro), Business for teams from two users, and Enterprise through sales.
  • Gemini Gems vs Claude skillsGoogle is replacing Gemini Gems with skills: on personal accounts, Gems become read-only on 17 November 2026 and are removed on 17 December; work accounts follow in 2027.
  • skills.sh vs a team podskills.sh is Vercel's public directory and open-source CLI for installing agent skills from GitHub into coding agents.
  • Dust vs a team skill podDust is an agent platform: people build and run agents inside Dust, over company data from Slack, Notion, Drive and other connectors, with a choice of models.
  • Sharing a scorecard with interviewersEvery interviewer should rate on the same scorecard, but interviewers sit on different AI plans: Claude Team, ChatGPT Business, personal accounts, outside panelists.
  • Prompt sharing sites vs a team libraryPublic prompt sites (marketplaces like PromptBase, galleries like PromptHero, vendor packs like OpenAI Academy's) are good for ideas; a private team library holds the prompts your company's work depends on.
  • TypingMind for teams vs a skill podTypingMind Teams is a separate chat app for a team on the team's own API keys, with shared prompts, agents and plugins set up by an admin.
  • Share sales skills with the teamEvery rep should prep calls from the same battlecards and skills, but sales teams sit on different AI plans: Claude Team, ChatGPT Business, a reseller's workspace, contractors on personal accounts.
  • Marketing AI tools comparedJasper, WRITER and Copy.ai are marketing AI platforms with their own apps, brand settings and workflows; a general AI plan with a shared brand-voice skill keeps writers in the ChatGPT or Claude they already use.
  • AI budget per seatBudget AI per person by the work each person does: which seat type they need on Claude Team or ChatGPT Business, the shared tools on top, and a line for usage beyond the seat.
  • Notion guest limitNotion's Free plan allows 10 guests; paid plans show no number.
  • Claude Code on the Team planClaude Code comes with every paid Claude plan, Team included, and shares the plan's usage, so on Team the choice is the seat: standard, or premium with five times the usage.
  • Claude Team and ChatGPT invoicesNo: Claude Team and self-serve ChatGPT Business take card payment only.
  • Add or remove AI seatsOn Claude Team an owner adds seats from Organization settings and pays the prorated amount right away, but a lower seat count only starts at the next renewal, with no refund for removing someone.
  • Do team plans train on your data?No, by default: Anthropic doesn't train on Claude Team or Enterprise chats, and OpenAI says ChatGPT Business data is never used for training.
  • Claude Enterprise vs ChatGPT EnterpriseBoth add SCIM, audit and compliance tools, retention controls and finer roles to the team plans, and neither trains on your data by default.
  • Microsoft 365 Copilot vs ChatGPT BusinessMicrosoft 365 Copilot is an add-on to a Microsoft 365 plan that works inside Word, Excel, Outlook and Teams with your company's mail and files; ChatGPT Business is a separate workspace with projects, skills and connectors.
  • Cursor Teams planCursor Teams is priced per active seat, with Premium, Super and Ultra seats that each carry a weekly allowance for Cursor's own models, a free seat, and a shared prepaid balance for other models.
  • Copilot Business vs Cursor TeamsGitHub Copilot Business adds AI to the editors and GitHub your team already uses, priced per seat with a shared pool of AI credits; Cursor Teams is its own editor with per-seat weekly allowances and a shared balance.
  • Copilot Business vs EnterpriseCopilot Enterprise includes everything in Copilot Business, about twice the AI credits per seat and priority access to new models and features, but it requires GitHub Enterprise Cloud.
  • Gemini Enterprise vs ChatGPT BusinessGemini Enterprise is Google Cloud's AI workspace that searches your company's data where it lives, across Google Workspace, Microsoft 365 and other tools, with a Business edition for up to 500 users; ChatGPT Business is OpenAI's team workspace with projects, skills and plugins.
  • Claude Pro or ChatGPT Plus for workClaude Pro runs on Anthropic's Consumer Terms and ChatGPT Plus on OpenAI's Terms of Use.

AI terms, explained

  • What is RAG?RAG lets an AI look up your documents before it answers.
  • What is GraphRAG?GraphRAG builds a knowledge graph from your documents so an AI can answer questions about the whole set.
  • What is a knowledge graph?A knowledge graph stores facts as things and the relationships between them.
  • What is an ontology?An ontology is the agreed list of the kinds of things in your business and how they relate.
  • What is context engineering?Context engineering is choosing everything an AI sees when it answers: instructions, examples, documents, tools and history.
  • What is MCP?MCP is the open standard that connects AI apps like Claude and ChatGPT to outside tools and data.
  • What is a forward deployed engineer?A forward deployed engineer works inside a customer's team to make software fit their real work.
  • What is an agent harness?An agent harness is everything around the model that makes it an agent: instructions, tools, memory files, the loop and the stop rules.
  • What are AI evals?An eval is a test for an AI system: an input, and a way to grade the output.
  • What are AI skills?An AI skill is a folder with a SKILL.md file: a name, a description and instructions an AI loads when a task needs them.
  • What are AI agents?An AI agent is a model that decides its own steps and uses tools until a job is done.
  • Prompt vs context engineeringPrompt engineering is writing the instructions; context engineering is choosing everything the AI sees.
  • What is a vector database?A vector database stores embeddings so an AI can find text by meaning, not exact words.
  • Fine-tuning vs RAGRAG gives an AI your current facts at answer time; fine-tuning changes the model with training examples.
  • AI governance policyAn AI governance policy says which AI tools your team may use, what data never goes in, who checks output, and who owns it.
  • LLM evaluation for non-engineersHow to evaluate what an LLM gives you without code: check claims, sources, tone and gaps before you use a draft, and when to build a test set.

Business and strategy books

  • Choosing a strategy frameworkFive forces, SWOT, VRIO, Wardley maps, the strategy kernel and more: which strategy framework fits which question, and how to choose in four questions.
  • Porter's five forcesHow to run Porter's five forces: define the industry, rate each force with evidence, find what limits profit, and pick where to compete.
  • Porter's generic strategiesCost leadership, differentiation, and focus: how to choose one of Porter's generic strategies on evidence, avoid the middle, and name its trade-offs.
  • Porter's value chainHow to run a value chain analysis: list primary and support activities, give each a cost share and driver, find where buyers see value, and pick the changes.
  • VRIO frameworkVRIO in plain words: test each resource for Value, Rarity, Imitability, and Organization, read the result from parity to sustained advantage, and act on it.
  • Helmer's 7 PowersHelmer's 7 Powers in plain words: an advantage lasts only with a benefit and a barrier.
  • SWOT and TOWSA SWOT analysis that leads to action: evidence for every item, external factors first, then a TOWS matrix that pairs them into SO, WO, ST, and WT strategies.
  • BCG's five strategy stylesYour Strategy Needs a Strategy, summarized: rate your market on predictability, malleability and harshness, pick one of five styles, and change how you plan.
  • Rumelt's strategy kernelRumelt's strategy kernel from Good Strategy Bad Strategy: write a diagnosis, a guiding policy and coherent actions, then test for the signs of bad strategy.
  • Strategy choice cascadeLafley and Martin's strategy choice cascade: five linked choices on one page, where to play and how to win as a pair, and a what-would-have-to-be-true check.
  • ERRC grid and strategy canvasHow to draw a strategy canvas and fill an ERRC grid (Eliminate, Reduce, Raise, Create), from Kim and Mauborgne's Blue Ocean Strategy, with a worked example.
  • Ansoff matrixThe Ansoff matrix sorts growth options into market penetration, market development, product development and diversification.
  • Scenario planningTest a decision against four plausible futures.
  • Theory of ConstraintsThe Theory of Constraints from The Goal, in plain words: find the one step that limits output, exploit it, subordinate the rest, add capacity last, repeat.
  • Build-measure-learnThe build-measure-learn loop in plain words: test your riskiest assumption with the smallest MVP, set the metric and threshold first, then pivot or persevere.
  • Crossing the chasmCrossing the Chasm in plain words: why sales stall after early adopters, and how to pick one beachhead segment, complete the whole product, and grow next door.
  • Jobs-to-be-done switch interviewHow to run a jobs-to-be-done switch interview: who to recruit, the timeline from first thought to first use, the four forces, and the job statement.
  • THE MODEL sales funnelTHE MODEL, the Japanese B2B sales classic by Yasutaka Fukuda, splits revenue work into marketing, inside sales, field sales and customer success.
  • Wardley mappingA Wardley map places what your business is made of by visibility to the user and by evolution, from genesis to commodity.
  • Sun Tzu: know the enemyWhat Sun Tzu's "know the enemy and know yourself" really says, its two weaker cases, and how to run it as a two-sided competitor assessment.
  • Musashi's Book of Five RingsWhat Musashi's Book of Five Rings says about the two ways of seeing, kan and ken, and how to use it to step back from an urgent number before you decide.
  • Measure What Matters (OKRs)Measure What Matters in plain words: what OKRs are, where they came from, how to write objectives and key results, and the mistakes that turn them into a task list.
  • Good to GreatGood to Great in plain words: Level 5 leaders, first who then what, the Hedgehog Concept's three circles, the flywheel, and how to test your own focus.
  • The Innovator's DilemmaThe Innovator's Dilemma in plain words: why well-run leaders lose to cheaper, simpler entrants, sustaining versus disruptive innovation, and how to check a rival.
  • High Output ManagementHigh Output Management in plain words: a manager's output, leverage, task-relevant maturity, one-on-ones and OKRs at Intel, with a way to audit your own week.
  • Thinking, Fast and SlowThinking, Fast and Slow in plain words: System 1 and System 2, anchoring, the planning fallacy, loss aversion, the outside view and the pre-mortem, applied to decisions.
  • Influence (Cialdini)Cialdini's Influence in plain words: reciprocity, commitment, social proof, liking, authority, scarcity and unity, and how to use them honestly on a page or offer.
  • Positioning (Ries and Trout)Positioning: The Battle for Your Mind in plain words: the brand ladder, being first, owning one word, positioning against the leader, and how to find your position.
  • Hooked (the Hooked Model)Nir Eyal's Hooked in plain words: trigger, action, variable reward and investment, how to find the weak phase in your product, and where habit design turns manipulative.
  • The Mom TestThe Mom Test in plain words: the three rules for customer conversations that even your mom can't lie to you about, good and bad questions, and how to turn talks into commitments.
  • Never Split the DifferenceNever Split the Difference in plain words: tactical empathy, mirroring, labeling, accusation audits, calibrated questions and safe no's, with a way to prepare a negotiation.
  • The Hard Thing About Hard ThingsBen Horowitz's The Hard Thing About Hard Things in plain words: the struggle, peacetime and wartime CEOs, layoffs and firing executives, and telling the team the truth.
  • Zero to OneZero to One in plain words: vertical versus horizontal progress, why competition destroys profit, the four traits of a lasting monopoly, starting small, and secrets.
  • The Lean Product PlaybookThe Lean Product Playbook in plain words: Dan Olsen's product-market fit pyramid, underserved needs, the value proposition, MVP scope and the six-step lean product process.
  • Obviously AwesomeObviously Awesome in plain words: April Dunford's five components of positioning, why to start from competitive alternatives, and the steps to position a product.
  • Traction (the Bullseye)Traction by Gabriel Weinberg and Justin Mares in plain words: the 19 traction channels and the Bullseye framework for testing them and focusing on the one that works.
  • The E-Myth RevisitedThe E-Myth Revisited in plain words: the technician, manager and entrepreneur, the fatal assumption, the franchise prototype, and how to build a business that runs without you.
  • Start with WhyStart with Why in plain words: Simon Sinek's Golden Circle of why, how and what, why purpose-first messages work, the limits of the idea, and how to write your own.
  • Radical CandorKim Scott's Radical Candor in plain words: the two axes, the four quadrants, praise and criticism, asking for feedback first, and how to check feedback before you give it.
  • Getting Things DoneGetting Things Done in plain words: David Allen's capture, clarify, organize, reflect and engage, next actions, projects, the two-minute rule, and the weekly review.
  • Shape UpBasecamp's Shape Up in plain words: shaping work, fixed time and variable scope, the five ingredients of a pitch, the betting table, and six-week cycles with cool-down.

Business thinkers' principles

  • Kazuo Inamori: Amoeba Management and philosophyKazuo Inamori's Amoeba Management (small units with their own profit, hourly efficiency, management by all) and the Kyocera Philosophy's root question, what is right as a human being, with sources.
  • Konosuke Matsushita: dam management and collective wisdomKonosuke Matsushita's dam management (keep a planned reserve in cash, equipment, people, stock, technology and products) and management by everyone with collective wisdom, explained from Panasonic and PHP sources.
  • Taiichi Ohno: 5 whys and the seven wastesTaiichi Ohno's 5 whys (keep asking why until the root cause is removed) and his seven wastes, from overproduction to correction, explained from Toyota and Lean Enterprise Institute sources, with office examples.
  • Warren Buffett: moat, circle of competence, margin of safetyWarren Buffett's moat, circle of competence and margin of safety, explained letter by letter from the Berkshire Hathaway shareholder letters, with steps to apply them to your own business.
  • Charlie Munger: inversion and mental modelsCharlie Munger's inversion (invert, always invert), his latticework of mental models and his advice to buy wonderful businesses at fair prices, from Poor Charlie's Almanack and the Berkshire letters, with a pre-mortem you can run.
  • Jeff Bezos: working backwards and one-way doorsJeff Bezos's working backwards from the customer, one-way and two-way door decisions and Day 1, from his shareholder letters, plus the PR/FAQ written before building, as Amazon describes it.
  • Elon Musk: first principles and the 5-step algorithmElon Musk's first-principles reasoning (price a product from its raw materials) and his five-step order for processes: question requirements, delete, simplify, speed up, automate last.
  • Peter Thiel: the contrarian question and monopolyPeter Thiel's contrarian question from Zero to One, and the monopoly ideas from his lecture Competition is for Losers: start small, the four characteristics, be the last mover.
  • Andy Grove: OKRs and managerial leverageAndy Grove's OKRs as he ran them at Intel, his one-on-ones, task-relevant maturity and the manager's output equation from High Output Management, explained from accessible sources.
  • Henry Ford: flow and wasteHenry Ford's principles of flow (bring the work to the worker, the moving line, an even flow of materials) and waste of materials and human effort, from his 1922 book My Life and Work, with office examples.
  • Ray Dalio: principles and believabilityRay Dalio's decision principles: treat each case as "another one of those", write your criteria down, weigh views by believability, and use transparency, from principles.com and his 2017 TED talk.
  • Paul Graham: do things that don't scalePaul Graham's essay Do Things that Don't Scale (July 2013): recruit users by hand, start with a narrow market, delight early users, do things manually first, skip the big launch.
  • Sam Altman: the Startup PlaybookSam Altman's Startup Playbook for Y Combinator founders and his essay How To Be Successful: users who love the product, focus and intensity, momentum, the CEO's jobs.
  • Jensen Huang: first principles and new marketsJensen Huang's ideas from his 2024 Stanford GSB conversation: redo things from first principles, create technology and markets together, show people how you reason.