AI Employee Onboarding: Plan a New Starter's First Two Weeks

Give AI the role, tools, people and handbook, and ask for a day-by-day plan. Days one and two cover access, people and how work flows; days three to five mix shadowing with one small real task; in week two the starter owns a routine piece of work, with short daily check-ins. Then check every item against what your team can support.

AI-drafted onboarding plans fail in two predictable ways. They're too full, with seven sessions a day and a new tool every hour, because a plan that covers everything looks thorough. And they invent: a policy you don't have, a tool you don't use, a "welcome lunch with the leadership team" at a twelve-person firm. The fix is to feed the AI a fact sheet, cap each day at two or three scheduled items, and read the result as a draft that your team has to be able to deliver.

Follow me on Instagram@sagnikteaches

The input pack to gather before you prompt

An hour spent collecting facts saves a plan full of guesses. Here's an illustrative pack for a new account executive at a twelve-person marketing agency:

Connect on LinkedInSagnik Bhattacharya

ROLE: Account Executive. Reports to: Account Director (Hannah).
Buddy: Sam (Account Manager). Start date: Monday 2 November.
Working pattern: hybrid, office Tue-Thu; first week fully in office.

WHAT THEY'LL OWN BY WEEK 4: weekly reporting for 2 clients; day-to-day
emails for 1 client; campaign set-up checklists for all Hannah's clients.

TOOLS (and who sets them up):
- Google Workspace (email, Drive, Meet) - office manager, before day 1
- Asana (tasks) - Sam, day 1
- HubSpot (CRM, view only at first) - Hannah, day 3
- Ad platforms (read access to 2 client accounts) - Hannah, week 2
- Reporting dashboards - Sam, week 2
- Password manager - office manager, before day 1

PEOPLE TO MEET: Hannah, Sam, the two designers, the paid-media lead,
the office manager, the MD (30 min in week 1).

DOCUMENTS: staff handbook, AI use policy, reporting SOP, campaign set-up
checklist, client one-pagers for the 3 clients they'll work on.

FIXED EVENTS: Monday 9:30 team stand-up; Wednesday client call (Client B);
Thursday internal review; Friday 4pm show-and-tell.

LIMITS: Hannah is at a client workshop all Tuesday of week 1.

That last line is the kind of fact that saves a plan. Without it, the AI will schedule the new starter's most important session with the person who isn't there.

Subscribe on YouTube@codingliquids

The planning prompt and what came back

Using ONLY the input pack below, draft a two-week onboarding plan for
this new starter, day by day (Mon-Fri, weeks 1 and 2).
For each day: a focus, at most 3 scheduled items with the person
responsible, and one thing the starter should be able to do by the
end of the day.
Rules:
- Use only people, tools, documents and events listed in the pack.
- If something would help but isn't in the pack, list it separately
  under SUGGESTIONS, not in the plan.
- Leave at least half of each day unscheduled for reading, practice
  and real work.
- The starter should do one small piece of real work by day 4 and own
  a routine task in week 2.
Input pack: {{pack}}

An illustrative extract of what came back:

WEEK 1 - TUESDAY (Day 2)
Focus: how work moves through the agency
- 10:00 Asana walkthrough with Sam: projects, statuses, how tasks arrive
- 11:30 Campaign set-up checklist review with Hannah
- 14:00 Meet the designers: how briefs reach them
By end of day: can find any client's current tasks in Asana and say
who owns them.

SUGGESTIONS (not in the pack):
- A welcome lunch with the team in week 1
- A short guide to the agency's brand voice

What to fix: the 11:30 session is with Hannah on the day the pack says she's at a workshop. The rules didn't stop it; they rarely catch every conflict, so read each day against the "limits" line yourself. Move it to Wednesday. The suggestions are fair: a team lunch costs little, and a brand-voice guide is a real gap worth writing (the new starter could even help draft it in week two). Everything else in this day is realistic: three items, half the day free, and a clear "can do" outcome.

The fortnight, day by day

After edits, the plan should fit on one page. Here's the agency's version, illustrated:

Notice what isn't there: no day with more than three scheduled items, no "read the whole handbook" day, and no tool introduced before there's a task that needs it. HubSpot waits until day three and ad-platform access until week two, both following the principle of giving access when the work needs it.

What the plan costs the team, in hours

A plan is only real if the people in it have the time. Add up each person's hours from the table before you commit. In the agency's plan, the buddy (Sam) carries about 7 hours across the fortnight: the day-one lunch, the Asana walkthrough, the report pairing on day four, dashboard set-up and the hand-over of Client A. The manager (Hannah) carries about 4: the checklist review, the Client B call, two check-ins and the fortnight review. The office manager needs about 2 hours, mostly before day one.

Seven hours is nearly a full day of Sam's fortnight. Block those hours in Sam's calendar when the plan is agreed, and tell Sam's clients nothing changes; if Sam's week is already full, move work off it rather than hoping. AI can produce this tally from the plan in seconds ("list each person's total scheduled hours by week"), which is a useful check to run every time you regenerate a plan.

Where two-week plans usually slip, and the fix for each

Even a good plan meets a real fortnight. These are the slips that come up most, with how each tends to show itself:

• The buddy disappears. On day six a client emergency pulls Sam away for two days, and the new starter spends them rereading documents. Fix: name a back-up for every scheduled item in week two, and give the starter a "if your session is cancelled, do this" list, such as practising the report on last month's data.
• Ownership arrives without review. Keen to show progress, a manager lets the new starter send a client email unchecked on day eight. It quotes last month's figures. Fix: in the first fortnight, every client-facing output is reviewed before sending, and the plan says so.
• Shadowing never ends. The opposite problem: the starter watches for two weeks and does nothing, because nobody wants to hand over. The day-four real task and the day-seven ownership line exist to force the handover; hold the team to them.
• Access lags the plan. The dashboards session on day six happens, but the dashboard licence was never bought. Fix: every tool in the plan has an owner and a "ready by" date in the pre-day-one checklist, even if it's needed in week two.

When a slip happens, update the plan rather than abandoning it. A plan that moves a session to Thursday still works; a plan everyone has stopped looking at doesn't.

Before day one: access, kit and a welcome that answers questions

Most bad first days are access problems. Put these on a checklist owned by one person, done by the Friday before:

• Accounts created and tested: email, file storage, task tool, password manager, and the AI assistant on your business plan. Log in as the new user once to check.
• Multi-factor authentication ready to enrol on day one, with the office manager present, since the authenticator app goes on the starter's phone.
• Password reset: check your plan. On Microsoft 365 Business Basic, users can't reset a forgotten password themselves; that needs Business Standard or above (or an Entra ID P1 or P2 licence). If you're on Basic, make sure an admin is reachable on day one, because new passwords get forgotten.
• Kit: laptop set up and updated, charger, any second screen, and a working desk if they're in the office.
• Calendar: the first week's scheduled items already in their calendar, so they arrive to a plan rather than a blank screen.

Then a welcome email, sent a few days before. AI drafts it from the plan in seconds; the manager personalises the opening. An illustrative version:

Subject: Your first week - see you Monday 9:15

Hi [first name],

Looking forward to Monday. A few practical things so nothing's a surprise:
- Arrive 9:15; [buddy's name] will meet you at reception. Stand-up is at 9:30.
- Your laptop is set up. Bring your phone to set up two-step login.
- This week is mostly in the office; from week 2 it's Tue-Thu.
- By Thursday you'll draft your first client report with [buddy's name]. No
  pressure to be quick; the point is to learn how we do it.
- Lunch on Monday is with [buddy's name], on us.

Your first week's calendar is already in your inbox. Any questions
before then, just reply.

Hannah

Check any date, time and name against the plan before sending. It's the kind of email where one wrong detail ("arrive 9:30" when stand-up starts at 9:30) sets the wrong tone.

An assistant the new starter can ask instead of interrupting

New starters ask dozens of small questions: where's the brand folder, how do we name files, who approves ad spend. Each is quick; together they cost the team hours and make the new person feel like a nuisance. A shared AI assistant built on your own documents takes most of them.

Set it up as a project in your business AI plan: a Claude Project, a ChatGPT Project, a Gemini Gem (becoming a skill from November 2026) or a Gemini Notebook, uploading the handbook, the AI use policy, the SOPs and the client one-pagers. Don't build a new custom GPT for this: OpenAI is retiring custom GPTs, which stop running on 11 December 2026. The tutorial on setting up Claude Projects as a shared team assistant walks through the set-up, and building a company knowledge base AI can answer from covers getting the documents into shape.

The instructions matter more than the documents:

You help new team members find answers in the attached documents.
Answer only from these documents and say which one you used.
If the documents don't cover the question, say so and suggest asking
[buddy's name] (process questions) or [manager's name] (client questions). Never guess at
policy, pay, holidays or client information not in the files.

An illustrative exchange, including the kind of error you should expect in testing:

Q: How do I name a file for a client report?
A: Per the reporting SOP: CLIENTCODE_Report_YYYY-MM-DD, saved in the
   client's Reports folder. (Source: Reporting SOP, section 2)

Q: How many days' holiday do I get in my first year?
A: The staff handbook says 25 days plus public holidays.
   (Source: Staff handbook, section 6)

The second answer was wrong in testing: the handbook had been updated to a different figure, but an older copy had been uploaded. That's the realistic failure. The assistant faithfully answers from whatever you gave it, so keep one current version of each document in the project, and tell new starters to confirm anything about pay, holidays or contracts with their manager. The same assistant helps everyone after onboarding too, which is what makes it worth the hour of set-up.

A reading path instead of "read the handbook"

Handing someone a 40-page handbook and six SOPs on day one guarantees they'll skim all of it and remember none. Ask AI to turn your documents into a reading path matched to the plan: which sections to read on which day, in the order the work needs them, each with two or three questions to check understanding.

From the attached documents and the two-week plan, create a reading
path: for each day, list the specific sections to read (not whole
documents), about 30-45 minutes in total, chosen to prepare for that
day's or the next day's work. After each section, add 2 short
questions that check the key point. Leave out sections not needed in
the first two weeks and list them at the end as "later".

The agency's day-3 entry, illustrated: "Campaign set-up checklist, sections 1-3 (15 minutes): Q1, what has to be confirmed with the client before any budget is spent? Q2, where do tracking links get recorded? Client B one-pager (10 minutes): Q, what are Client B's two goals this quarter?" Those questions then give the day-3 check-in something concrete to talk about. The "later" list is just as useful: it tells the new starter, explicitly, what they're allowed to ignore for now.

If your SOPs don't exist yet, or live only in people's heads, this is a good moment to write them; writing SOPs with AI shows how to go from rough notes to usable procedures.

Check-ins at day three, day five and day ten

Short, scheduled conversations catch problems while they're small. Keep each to 15-30 minutes and ask the same kind of questions, so the new starter can prepare:

• Day 3 (15 minutes, buddy): What's been most confusing? Is anything blocking you, like access, equipment or not knowing who to ask? What did the reading path get wrong?
• Day 5 (30 minutes, manager): What's clear now that wasn't on Monday? How did the first real task go? What do you need next week?
• Day 10 (30 minutes, manager): Which parts of the role can you now do on your own? What goals make sense for the next month? What would you change about these two weeks for the next person?

AI's role here is small and useful: it can draft the agenda from the plan, and afterwards turn the manager's rough notes into a short written summary with agreed actions, which the new starter receives the same day. Keep the conversation itself human. If you want the fortnight's goals to carry into formal reviews later, use the same competencies the role was hired on; the tutorial on interview questions and scorecards explains how those are built.

Adapting the plan for remote, part-time and senior starters

The same prompt works with different inputs. Three illustrations:

• A remote starter at a PR consultancy. Add to the pack: which tools replace the desk chat (a video call at 9:00 and 16:30 each day in week one), when kit ships, and a buddy with a standing daily slot. Ask the AI to put a short video call before every task that would, in an office, start with "come and look at this". Remote plans that copy office plans leave people isolated by day three.
• A part-time bookkeeper at a financial planning firm, three days a week. Tell the AI the working days, and ask for the plan by working day rather than calendar day, so "day 6" falls in week two, not the wrong Monday. Fewer people to meet, more time on the systems they'll use alone.
• A senior hire at a video production company. Less shadowing, more listening: ask for one-to-one conversations with each team member and key client in the first fortnight, with the question "what should I change and what should I leave alone?", and a first piece of ownership chosen with the MD rather than handed down. Senior starters who are over-scheduled with training feel patronised.

Onboarding onto your AI tools themselves (what's allowed, which accounts to use, what not to paste in) deserves its own session in week one; onboarding new hires onto your AI tools and rules covers that part.

How you'll know the fortnight worked

Judge the plan by outcomes at day ten, not by how much was covered. Five checks:

1. Access complete for everything the next month's work needs, with nothing granted that it doesn't.
2. At least one piece of real work delivered and used, such as a client report sent or a checklist completed on a live campaign.
3. One routine task owned, with the reviewer checking rather than redoing it.
4. The new starter can name who to ask for process, client, technical and admin questions without looking it up.
5. Their feedback is specific. Ask: what was the most useful day, what was wasted, and what did you need that nobody gave you? Put the answers into the plan before the next hire.

Save the final plan, the prompt, the input pack and the reading path together. The next starter's plan then takes about an hour: update the pack, rerun the prompt, fix the clashes, and send the welcome email. That repeatability, more than any single plan, is what AI brings to onboarding.

New-starter planning questions

How much of the first two weeks should be training versus real work?

Aim for the new starter to do a small piece of real work by day three or four, and to own something routine by the second week. Pure training for two weeks leaves people anxious and under-used. A useful rough split is mostly learning and shadowing in week one, then mostly supervised real work in week two.

Should the new starter get access to everything on day one?

No. Give them what the first week's plan needs, and add the rest as the plan reaches it. That keeps day one manageable and follows the principle of least access. Client systems, finance tools and admin rights can wait until there's a task that needs them, with the manager approving each one.

Can AI answer a new starter's questions instead of their manager?

It can answer 'where is' and 'how do we' questions from your own documents, which removes a lot of small interruptions. It shouldn't replace the manager's check-ins, feedback or judgement calls. Set it up to answer only from your files and to send people to a named colleague when the documents don't cover something.

Further reads

• Gemini Notebook (Formerly NotebookLM) for Small Business Teams — Another way to give new starters a searchable document assistant.
• How to Use ChatGPT Projects to Keep Client Work Separate — Set up ChatGPT Projects so client material stays separated.
• How to Create SOPs From Screen Recordings With AI — Turn screen recordings into the SOPs a new starter reads.
• Custom GPT vs Claude Project vs Gemini Gem: Which Should You Use? — Choose the right shared-assistant format now custom GPTs are retiring.
• How to Set Up an AI-Assisted Hiring Process for a Small Team — The hiring stages that come before this plan.
• How to Write Fairer Performance Reviews With AI — Carry the onboarding goals into the first review.
• How to Write Job Descriptions With AI That Attract Good Hires — AI writes fluent job adverts; good candidates respond to specifics. The inputs, prompt and eight edits that turn a generic draft into one worth applying to.
• How to Write an Employee Handbook With AI and What to Check — A staged method for drafting a small-firm handbook with AI, plus the checks that catch imported legal figures, accidental promises and contradictions.
• How to Run Staff Surveys and Exit Interviews With AI Analysis — A 12-question survey, an exit interview script, anonymity settings, a theme-coding prompt with checks, and an online clothing shop's first round of results.
• How to Build a Staff Training Matrix With AI — A skills-by-people grid with level definitions, expiry formulas and a gap score, built with AI prompts and filled in for an eight-person jewellery workshop.
• Opening and Closing Checklists: Build Them With AI in an Hour — Turn a walk-through of your premises into opening and closing checks that staff can follow, record and hand over between shifts.
• Employment Contracts With AI: Drafting Steps and Legal Checks — Draft a clear employment contract with AI from a job facts sheet, then send the right clauses to a lawyer instead of paying for the whole thing.
• How to Write a Health and Safety Policy With AI and What to Check — Draft a three-part health and safety policy with AI from your real hazards, then check it on a walk-round before anyone signs it.
• How to Answer Staff HR Questions With an AI Helpdesk — Build an internal AI helpdesk that answers everyday HR questions from your own policies and hands personal or sensitive matters to a named person.
• AI for HR in Small Businesses: 12 Tasks You Can Hand Over Safely — Twelve HR jobs AI can draft or sort for a small team, each with an example and a watch-out, plus the people decisions that stay human.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: project fact sheet on Microsoft 365 Business Basic password reset, OpenAI's custom GPT retirement and ChatGPT and Claude Projects, checked against vendor pages September 2026.