The starter kit Jay gives every leader and team he works with. Five sections of ten tips each: read it in 20 minutes, ship something this afternoon.
Before tools and prompts, mindset. Most AI failures aren't technical: they're how you frame the question.
An intern with the world's reading list, who still needs direction, examples, and review. The job is delegation, not magic.
Try: When you're stuck, ask "How would I brief an intern on this?" Then paste that into your AI.
The leaders who win with AI aren't coders: they're people who can articulate what "good" looks like. Specificity is the skill that matters.
Try: Before each prompt, write the one-sentence outcome you want. Paste it in.
Stop using it to type faster. Start using it to think clearer. The best uses involve back-and-forth, not one-shot generation.
Try: Reply to its first answer with "what would you push back on if you were my sharpest colleague?"
Don't reinvent. Keep a personal library. Most of your "AI productivity" comes from reusing the same 20 prompts.
Try: Turn on Memory in your AI tool so it stops making you repeat yourself. Then keep one short doc for the five prompts you reuse most.
The jump this year is that AI can reach your actual work. Connectors let it read your mail, your calendar, your drive, and your notes, so you stop being the courier between your tools and the chat window.
Try: Turn on one connector today. Start with your calendar, it is the lowest risk.
Ask whether the task is worth an hour of your own rate. If it is not, run it through AI. Stop hand-typing what a prompt does in 30 seconds.
Try: Pick the dumbest, most repetitive thing on today's list. Prompt it instead.
The fastest way to a clear strategy is to draft 10 bad versions in 20 minutes with AI. Then pick.
Try: Ask AI for 10 different angles on your next decision. Pick the best three.
Anyone can prompt. What separates good output from great is editing, and editing is taste, which is yours alone.
Try: Spend most of your AI time editing, not prompting. Jay runs about two thirds editing.
Every quarter you wait is a quarter your competitors spend learning. The learning is the moat, not the tool. Imperfect adoption beats perfect strategy.
Try: Pick one workflow this week. Apply AI imperfectly. Iterate.
You'll discover the real limits by doing, not by reading. Push every workflow. Map the walls.
Try: Pick one core workflow. Use AI for every step this week. Note what breaks.
A few prompt patterns do most of the heavy lifting. Master these and you'll never write a vague prompt again.
"Write for a busy CFO who needs the bottom line in 30 seconds" produces wildly different output than "summarize this." Audience is half the brief.
Try: Add "Audience: [specific role + context]" to every prompt for a week.
"Goal: Y. Must include A and B. Must not do C. Length: 200 words." Skip the "you are a world-class expert" opener. Current models do not need it, and it just costs you typing. Goal and constraints remove most of the back-and-forth.
Try: Write your next 5 prompts as a goal plus a list of constraints.
Paste 2 examples of what good looks like. Examples carry the things you cannot put into words: tone, rhythm, and how much detail is too much.
Try: Find an example of the output you want. Paste it. Ask for "more like this."
"Ask me up to five questions, then write." This beats guessing what it needs, and it surfaces the parts of the brief you had not thought through.
Try: Add "ask me questions before you start" to your next big prompt.
"Don't use the word 'leverage.' Don't open with a question." Negative constraints sharpen output faster than positive ones.
Try: Add a "Don't do" list to your next 3 prompts.
"Give me three drafts, each with a different angle." One great, one decent, one terrible: the terrible one teaches you what to avoid.
Try: Default to "give me 3 versions" on creative work.
"Now act as a skeptical senior exec and list the 5 weakest points in what you just wrote." Self-critique catches problems you'd miss.
Try: End every important draft with "what's wrong with this?"
"Write me a prompt that will produce X" works embarrassingly well. AI is great at prompting itself.
Try: Next time you're stuck, ask AI to write a prompt that would solve your problem.
"Here's my first pass, make it sharper" almost always beats "write something for me."
Try: Always write 100 words of your own first. Then improve with AI.
Put your standing instructions in a Project. ChatGPT and Claude both have them. Load your role, your audience, and your rules once, and every chat inside that project starts there. This is where the time savings come from.
Try: Create one Project for your most repeated task. Paste in your context.
These are the highest-ROI plug-ins for an average knowledge worker's week. Pick three and watch your output change in two weeks.
Connect your inbox instead of pasting into it. Turn on the Gmail or Outlook connector in ChatGPT or Claude, then ask: "Read my unread mail from the last 24 hours. Which three need a reply today, which three can wait, which can I archive?" If you use Gemini or Copilot, this is already inside your mail app.
Try: Tomorrow morning, triage before opening your inbox.
Granola, Fathom, or the recorder built into Zoom, Teams, and Meet will hand you decisions and action items without a prompt. Your job is the follow-up: "What did we decide that we did not write down?"
Try: Turn on a recorder for your next meeting. Ask it that question afterwards.
Most "polish" is starting from a blank page. Have AI draft. Spend your real energy on the edit.
Try: Next blog/post/email: AI drafts in 5, you edit in 25.
"Here's a CSV. What are the 3 surprising things in this data?" AI is a fast junior analyst that never gets bored.
Try: Upload the actual file, do not paste the rows. Then ask for a chart of the trend as well as the summary.
Before a big call: "Argue against my plan as a hostile board member would." Then "Now defend my plan."
Try: Steelman your next big decision before the meeting.
The big models are close enough now that the tie-breaker is your existing stack. Gemini if you are in Google Workspace. Copilot if you are in Microsoft 365. Claude or ChatGPT for the strongest standalone chat. When you need sources, use the deep research mode inside whichever you picked.
Try: Check what your company already pays for before buying anything new.
Hour-long talk → tweet thread, LinkedIn post, blog article, newsletter, 3 short clips, 1 quote graphic. One piece of work, week of content.
Try: Take your last recorded talk and extract 7 pieces this week.
"How might this email be misread by [recipient]? Where are the ambiguities?"
Try: Pre-mortem your next sensitive email before hitting send.
"Explain [thing] like I'm a smart 12-year-old. Then like I'm an expert. Then point out where those two answers disagree."
Try: Pick a topic you've been avoiding. Apply this. Time it.
Wispr Flow (free tier, or about $15 a month for Pro) or your phone's voice typing, then paste into AI and say "polish this for [audience]." Most people talk about three times faster than they type. The draft is rougher, so you trade typing time for editing time.
Try: Voice-dictate your next 5 messages. Polish with AI.
The bottleneck isn't tooling. It's getting humans to actually use it. These tips are about culture, not code.
Don't pilot "AI for marketing." Pilot "AI for our weekly content brief." Specific, owned by one person, measurable.
Try: Name the ONE workflow you'll change first. Write it down today.
Every team has one person who's already messing around with AI on their own time. Find them. Make them the visible lead.
Try: Name them this week. Tell them out loud that experiments on work time are fine.
End every standup with: "What prompt saved you time this week?" Shared learning compounds. Hidden learning evaporates.
Try: Add this question to your next team meeting agenda.
Get the team in a room. List the 10 most common tasks. Build a prompt template for each, together.
Try: Schedule a 60-min "prompt sprint" with your team this month.
"AI is going to make us all redundant" is the silent killer of adoption. Do not promise nobody's job changes. Say what is true: the tasks change first, and the people who learn the tools have the most say in what happens next.
Try: Put "job impact" on your next team agenda. Don't dodge it.
Adoption fails when "use AI" is added on top of existing workload. Carve out 2 hours/week of "explore AI" time.
Try: Block 2 hours this Friday for your team to play with AI.
People learn AI by watching peers struggle and succeed. A 90-min team session beats five 30-min individual ones.
Try: Run a 90-min team AI session instead of solo onboarding.
When someone saves 4 hours with AI, broadcast it. Slack channel, all-hands, leader email.
Try: Create #ai-wins in your team's chat tool this week.
Tool sprawl kills adoption. Check what you already pay for first, because Google Workspace and Microsoft 365 now include Gemini and Copilot. Standardize on that plus at most one other.
Try: Pick your team's "official 3" AI tools this week.
Pick a number. When you hit it, do something visible. Adoption is a story, not a memo.
Try: Set your team's first AI milestone today. Pick the celebration.
Once AI is in your workflow, two questions matter: is it costing us anything risky, and is it actually paying off?
Use a business plan (ChatGPT Business, Claude Team, Microsoft 365 Copilot) or a local model for PII, contracts, and financials. Business plans do not train on your data by default. Free consumer accounts may.
Try: Audit what your team has pasted this month. Adjust.
You don't need 30 pages. You need: "Use these tools. Never these. Always disclose AI use in [contexts]."
Try: Draft it with the AI Use Policy Starter.
Customer-facing content, regulated industries, legal docs: disclose. Internal emails: don't bother.
Try: Make your disclosure rule explicit this week.
"We're more productive with AI" is unprovable. "We saved 12 hours this week" is. Multiply by hourly rate. Communicate the number.
Try: Use the ROI Calculator to turn hours into a number your CFO recognises.
$20/month for ChatGPT Plus is the best ROI in your tech stack. Don't cheap out at the start.
Try: Audit your team's AI subscriptions. Upgrade where it matters.
One doc, updated weekly: "Workflow → time saved → who did it." After 8 weeks you have a story.
Try: Start with the Prompt Performance Tracker instead of a blank doc.
Three questions before you ship AI output: 1) Did I verify the facts? 2) Does it sound like me/us? 3) Could it hurt anyone if wrong?
Try: Paste those three questions at the top of your prompt doc.
Your "what AI fails at for us" list is more valuable than the success list. It tells you where humans still earn their salary.
Try: Start an #ai-fails channel and post the first one yourself.
AI tools are improving faster than any other category. The tool that lost 6 months ago may be winning now.
Try: Block one 60-min "tool audit" on your calendar every 90 days.
AI as "a project" sunsets. AI as "the way we deliver more output with the same team" lives in the P&L forever.
Try: In your next leadership update, tie your top AI win to a P&L line.
You don't need to act on all 50. Pick three. Ship one this week. The leaders who win with AI are not the ones who read the most: they're the ones who shipped one bad version first.
Take the 90-second AI Opportunity Quiz, or book a 20-minute call.