Train a new hire on AI in their first month, in three steps. On day one, give them your rules and ask what AI habits they bring. In week one, have them watch a colleague do one real job with AI your way. By day thirty, hand them one AI-assisted job of their own, with a named person checking it.
Sources re-read on November 17, 2026. The survey figures below come from the United States and the United Kingdom, and each one carries its date and its sample. This page also gives you a first-week checklist to copy and a 30-60-90 day plan.
A new hire will use AI in your business from their first morning, whether or not anyone shows them how. Most small businesses show a starter where the kettle is, how the invoicing system works and who approves vacation time, and they leave AI to whatever the last employer allowed. The plan below takes one month and a few hours of one colleague's time.
Why does AI belong in onboarding now?
AI belongs in onboarding because new hires now arrive with AI habits, and the first month is the cheapest time a business will ever have to shape them.
The hiring market has changed which habits walk through the door. In a June and July 2026 survey of 526 HR and learning and development leaders at US and UK companies with 50 to 999 employees, carried out by General Assembly and reported by IT Brief UK on 19 August 2026, 47% said they had hired staff from larger firms in the past year. 72% said they now include AI training in onboarding. And 40% said they had passed over an otherwise qualified candidate because the person lacked AI skills.
Read those three numbers together. Smaller employers are hiring people from bigger ones, those people often bring fluent AI habits, and the habits were formed under someone else's rules. Some will be worth spreading. Some, such as pasting a client file into whatever tool the last firm had licensed, would break your rules by Tuesday.
You cannot assume a previous employer trained them well, either. BCG's AI at Work survey, published in June 2026, which covers close to 12,000 employees, managers and leaders in more than a dozen markets, found that "Close to three-quarters (72%) of all respondents say expectations for the skills they need have shifted; however, only 36% feel that they have received adequate upskilling." Small business owners know the gap from their own side. In Goldman Sachs 10,000 Small Businesses Voices research released on 17 March 2026, 73% of the 1,256 US owners surveyed between January 27 and February 4, 2026 said they would benefit from more access to training and implementation resources, and only 14% said AI is fully embedded in their core operations.
That is what the research says. Here is what I take from it. The first month is when a person learns "how we do things here." If AI is missing from that lesson, a new starter does not conclude that AI is off limits. They conclude that there is no house way, and they fall back on their own.
What are the Three Firsts?
The Three Firsts gives a new hire your rules on day one, your way of working in week one and a job of their own by the end of the first month.
The Three Firsts is how I suggest a small business brings a new hire into its use of AI: by the end of the first day they know the rules, by the end of the first week they have watched a colleague do one real job with AI the house way, and by the end of the first month they own one AI-assisted job that a named person checks. Ask what AI habits they bring before you teach them yours, because the habits they arrived with will be used whether you ask or not.
Each first does a different job. The day teaches limits. The week teaches method. The month turns both into a habit that belongs to your business rather than the last one.
The first day: the rules, and one question
On day one, a new hire should leave knowing what they may use, what must never go in and who to ask.
If you already have a one-page AI policy, hand it over with the employment contract and the fire exits. If you do not, the One-Page Five sets out the five questions it must answer, in words a new starter would understand on their first day. Then set up their login to your approved tool before they need it. A starter with no approved account will open a personal one by lunchtime, and you will have created the problem you were trying to avoid. If yours is a regulated practice, the day-one rules need three more lines, on client confidentiality, checking citations and supervision; an AI policy for a small law firm gives the wording. For seasonal starters, the one-page briefing sheet in this guide to AI for retail and hospitality is the quickest version.
Then ask one question: "What did you use AI for in your last job, and what were you allowed to put into it?"
Listen for two things. The first is a good habit worth stealing, such as a checklist they run on every AI draft. The second is a habit that breaks your rules. Treat the answer the way the Daylight Rule treats any disclosure: nothing they tell you that day counts against them. You want the truth about their habits, and you will only hear it if telling it is safe.
The first week: watch one real job done the house way
In week one, a new hire should watch the colleague who does a job best do it with AI, from the first prompt to the final check.
This is how people at small firms say they want to learn. LinkedIn's Work Change Special Report on small businesses, published in December 2025, found that watching and learning from a trusted colleague tied as one of the three ways employees at US small businesses most prefer to learn a new AI skill or tool, each named by 26%. The report does not state how many people were asked. The other two were virtual training and tutorials, and practicing real-life scenarios with an AI tool. A buddy who shows the work is not a soft option. It is the method staff ask for.
Pick one job the starter will do in their first month. The colleague does it in front of them: the brief they write, the material they hand over, what they check in the draft and what they change. If your business keeps house briefs, the starting prompts for jobs that come around every week, this is when the starter gets them. Then the roles swap, and the starter does the same job while the colleague watches.
One distinction matters here. The New Hire Test uses an imaginary new hire to check whether a prompt is clear enough. The Three Firsts is about the real one, sitting at the next desk, who needs to see a good brief written, not hear that briefs matter.
The first month: own one job, with a named checker
By day thirty, a new hire should own one recurring AI-assisted job, and a named person should check that work until the checking finds nothing worth changing.
Choose a job that comes around every week, has a standard everyone recognizes and already gets checked before it leaves the building. A weekly client update, an inventory reorder note or the first draft of a quote all qualify. Give the checker a name. "Someone will look at it" means nobody will.
For a junior hire, the checking is the training. The Second Job Rule holds that every junior job does two jobs: the work, and the making of your next experienced person. The named checker is where the second job happens, because the junior learns judgment from what gets changed and why.
At the end of the month, the starter's manager asks the three questions of the Confidence Test, which I use to find out whether AI confidence exists in an organization: "Do you know what you are allowed to use it for? Have you used it on your own work this week? If it got something wrong, would you tell someone?" Three yeses mean the Three Firsts worked. A no tells you exactly which first to go back to.
What should a new hire's first week of AI training include?
A new hire's first week should include the rules, a login, one honest conversation about old habits, one real job watched and the same job done under a colleague's eye. Copy the checklist, put names where the brackets are and add it to the onboarding sheet you already use. It is my suggested list, not a legal requirement, so cut any line that does not fit your business.
Before they arrive (manager)
- Login to the approved AI tool created with their other accounts
- One-page AI rules printed or sent with the contract
- Buddy named: [the colleague who does the week-one job best]
- Week-one job chosen: [one job they will do in their first month]
Day 1
- Reads the AI rules with [manager], ten minutes, questions welcome
- Logs in to the approved tool and knows that personal accounts are not for work
- Is asked: "What did you use AI for in your last job, and what were you allowed to put into it?" Answers written down, nothing held against them
- Knows who to ask before putting anything new into AI: [name]
Days 2 and 3
- Watches [buddy] do the week-one job with AI from the first prompt to the final check
- Is given the house brief for that job, if one exists
- Writes down what [buddy] checks before the work goes out
Day 4
- Does the same job while [buddy] watches
- [Buddy] says what they would change, and why
Day 5
- Fifteen minutes with [manager]: one thing learned, one question, one habit from the last job worth keeping
- Thirty-day job and its checker confirmed in writing: [job], checked by [name]
What does a 30-60-90 day AI plan for a new hire look like?
A 30-60-90 day plan takes a new hire from following your way of using AI, to improving it, to passing it on. The Three Firsts cover the first thirty days. The table carries the same idea through a typical three-month probation. It is a suggested plan to adapt, not a researched standard, so change the days to match your own review dates.
| Stage | What the new hire does | Who checks | The evidence it is done |
|---|---|---|---|
| Days 1 to 30 | Learns the rules, watches one job, then owns one recurring AI-assisted job | The named checker reads every piece before it goes out | Three yeses to the Confidence Test at day thirty, and the checker is changing less each week |
| Days 31 to 60 | Takes on a second AI-assisted job and writes, or improves, the house brief for the first | The checker moves to spot checks, say one piece in three; the manager reads the brief | A written house brief a colleague could follow, and a short note of each spot check |
| Days 61 to 90 | Shows the next starter, or a colleague who has not used AI on that job, how it is done, and suggests one change to the rules or the approved tool list | The manager, at the 90-day or probation review | The colleague can do the job, the suggestion has been answered, and AI use is a line in the review |
The middle column is the one small firms skip. If nobody is named, nobody checks, and the new hire's habits become the house way by default. When the spot checks show a gap, training staff to check AI output covers what the checker should be looking for.
What I See When I Train Business Teams
Across the business teams I train, the newest people in the room are often the most fluent with AI and the least sure of the rules.
They know their way around the tools, sometimes better than their managers. What they do not know is what this business lets them put in, and they rarely ask in front of colleagues. The question tends to come afterwards, quietly, at the side of the room: "Am I actually allowed to use this on client work here?" When that question arrives in week six, the person has usually been answering it for themselves since week one.
The longer-serving staff show the opposite pattern. They know the rules, the clients and the standards, but many have never used AI on their own work. That makes them the right people to show a starter the house way, provided they have learned it first. You cannot hand on a way of working nobody has agreed.
I came to business training from education: I began as a secondary school teacher, became an assistant headteacher and went on to lead digital strategy in further education. Onboarding is the most educational moment a business has. I put it this way on my business page: "The hard part of AI at work is not the technology. It is helping people use it well, safely and every day. That is what an educator does." The first month is where "every day" starts.
What does this look like in a 30-person firm?
Take a hypothetical 30-person insurance broker that hires an account handler from a large national broker.
Day one. The office manager, who keeps the firm's list of approved AI tools, hands over the one-page policy and asks the question. The new handler explains that her old employer licensed an enterprise tool and let staff put client policy details into it. Here, client names and policy numbers never go into AI. She would have done it on her second day, in good faith. She also describes a checklist she ran on every AI-drafted renewal summary, which the firm does not have. The office manager writes it down.
Week one. The senior account handler drafts three renewal reminder letters with AI in front of her, using the firm's house brief, and talks through what he checks: the renewal date, the premium and anything the client was promised last year. Then she drafts the next three while he watches.
Day thirty. She owns the weekly renewal reminder drafts, and her team leader checks each batch before it goes. Her manager asks the three Confidence Test questions and gets three yeses. Her checklist has become a house brief that the rest of the team now uses.
Notice which way the learning ran. The firm taught her its rules, and she taught the firm a better habit. Onboarding done this way is a two-way door.
Before you plan the sessions, check whether each person's gap is skill, permission, time or judgment; the Gap Test sorts that on one page.
What if a new hire knows more about AI than you do?
Treat a fluent new hire's knowledge as an asset you bring inside your rules, not a reason to skip onboarding. In a family firm, where the fluent newcomer may also be a shareholder, the Two Hats Rule shows how to give them a defined decision without handing over the owner's.
The day-one question matters most for exactly this person, because their habits are the strongest and the least likely to match yours. Ask, write down what is worth keeping, and then give them a role in the week-one session for the next person who joins. A starter who arrives knowing more than the team can become the colleague who shows the next starter, once they have learned the house way themselves.
What should you do this week?
You can put the Three Firsts in place before your next starter arrives, without a training budget or an HR department.
- Find your one-page rules. If you do not have them, writing them is your first job, because day one depends on them.
- Copy the first-week checklist into your onboarding sheet. At the least, set up the approved AI login with the other accounts and ask the day-one question in the first conversation.
- For each role you hire into, name one week-one job and the colleague who does it best with AI.
- Name the thirty-day job and its checker before the start date, not after.
- Put the three Confidence Test questions in the manager's diary for day thirty, and the two later rows of the 30-60-90 table in the diary for days sixty and ninety.
- If you are interviewing this quarter, ask candidates what they used AI for and what they were allowed to put into it. The answer tells you as much about their judgment as about their skill.
Where to go next
If your current team has never had its own week-one session, start there, because it is hard to show a new hire a house way that nobody has agreed. That is exactly what my AI training for teams is built for. If you want to see where your business stands first, the free Business AI Readiness Scorecard takes about five minutes.
Sources and further reading
Sources re-read on November 17, 2026.
- General Assembly survey of HR and learning and development leaders, reported in "Smaller firms use AI training to recruit & keep staff", IT Brief UK, 19 August 2026. https://itbrief.co.uk/story/smaller-firms-use-ai-training-to-recruit-keep-staff
- "AI at Work: Why Strategy Matters More Than Tools", Boston Consulting Group, June 2026 (close to 12,000 employees, managers and leaders in more than a dozen markets). https://www.bcg.com/publications/2026/ai-at-work-why-strategy-matters-more-than-tools
- "Survey: Small Businesses Embrace AI, But Need Training and Support to Fully Harness It", Goldman Sachs 10,000 Small Businesses Voices, 17 March 2026 (1,256 US small business owners, surveyed January 27 to February 4, 2026 by Babson College and David Binder Research). https://www.goldmansachs.com/pressroom/press-releases/2026/small-businesses-embrace-ai-but-need-training-and-support-to-fully-harness-it
- "Work Change Special Report: How Small Businesses Can Win in 2026", LinkedIn Economic Graph, United States, December 2025 (sample size not stated). https://economicgraph.linkedin.com/content/dam/me/economicgraph/en-us/PDF/Work-Change-Small-Business-Report-Dec-2025-US.pdf
Dan Fitzpatrick is a keynote speaker, Forbes contributor and five-times bestselling author who has trained more than 150,000 people across 30+ countries. About Dan



