Caregiver Onboarding Checklist: Pre-Start Checks and Induction With AI

A caregiver onboarding checklist should cover four stages: pre-start checks (identity, right to work, background check, references, registrations), induction (policies, mandatory training, systems), shadow shifts with sign-off, and a welcome pack. AI can draft the checklist and each starter's dated plan; a named manager approves every release to client visits.

Most agencies already have a checklist somewhere. The gap is the plan around it: who does what on which day, what the starter has been told, and what is still outstanding at the end of week two. That is where AI earns its place, and it is the part most checklist pages skip.

Follow me on Instagram@sagnikteaches

Why a checklist alone leaves gaps

A plain checklist records that something happened. It does not say when it should have happened, who owns it, or what to do when it slips. A starter whose second reference is still missing on day four is not a tick-box problem; it is a scheduling problem, and it needs a person to chase it.

Connect on LinkedInSagnik Bhattacharya

The approach below separates three layers. The master checklist is the fixed list of requirements, written once and reviewed by a manager. The induction plan is a dated version of that list for one starter, which AI drafts from the master checklist plus the starter's role and start date. The sign-off log is the record, kept in your care management system, of who confirmed each item and when. AI touches the first two layers only. It never holds or decides the third.

Subscribe on YouTube@sagnikteaches

Keep one rule in view throughout: your regulator, insurer and local employment law decide what must be checked before someone visits a client. The structure and wording below do not replace those requirements, so check the current ones with the body that regulates your agency and put the version date on the checklist.

Stage 1: Pre-start checks, item by item

Pre-start checks are the items that must be complete before the starter does anything unsupervised, and in many agencies before they step into a client's home even as an observer. Group them so the office can see at a glance what is blocking a start date.

Two columns do more work than they look like they do. "How to verify" stops a tick meaning "I think someone did this". "Owner" stops everything sitting with whoever is in the office that week. If you adapt the table, keep both.

What AI should and should not touch here

Let AI write the chasing messages and the status summary. Do not let it handle the evidence. A prompt that works at this stage uses only the first name, the item and the date:

You are helping the office team at a small home care agency.
Write a friendly, brief text message to a new starter whose first name is [first name].
Their start date is [date]. We still need: second reference (referee not yet replied),
original qualification certificate. Ask them to bring the certificate to the office on
[day] between 10 and 4, and to nudge their referee. Keep it under 70 words, no jargon.
Do not mention any documents other than those two.

A typical output (illustrative) reads: "Hi [first name], we're nearly there for your start on [date]. Two things left: please bring your original qualification certificate to the office on [day] between 10 and 4, and give your second referee a gentle nudge to reply to us. Let us know if either is a problem. Thanks!" What you fix: add the office phone number, and remove anything the AI invents, such as an extra deadline.

Stage 2: The induction schedule AI can draft for each starter

Induction is where agencies differ most, and where a dated plan helps most. Give the AI your master list of induction topics and the starter's details, and ask for a plan that spreads the load across the first two weeks rather than front-loading everything into day one.

Here is a prompt template you can reuse. Put your own master list between the brackets.

Create a dated two-week induction plan for a new caregiver.
Role: care assistant, domiciliary (home visits). Experience: 3 years in a care home.
Start date: Monday 3 March. Working pattern: Monday to Friday, mornings.
Use ONLY the topics in this master list: [paste your induction topics, one per line,
with the hours your training provider says each needs].
Rules: no more than 5 hours of classroom or e-learning a day; no solo visits;
shadow shifts begin on day 3; each day ends with a 10-minute check-in with the
supervisor. Output a table: Day | Morning | Afternoon | Who signs off.
Flag any topic from the master list you could not place.

The last instruction matters. Asking the AI to flag topics it could not place gives you a check against it silently dropping something, which is the most common failure at this stage.

A filled-in example for one starter

Say a care agency with twelve caregivers is taking on three starters at once. One of them, [first name], has three years' care-home experience but has never done home visits. The AI's draft for week one might look like this (illustrative):

What you would fix in that draft: check that the medication session is the correct level for the role, because AI cannot know your training provider's levels; move the manual-handling practical earlier if your trainer is only free on Monday; and make sure the registered manager, not the AI's default "supervisor", holds the sign-off for anything that releases the starter to visits.

Stage 3: Shadow shifts that actually teach something

A shadow shift is not a ride-along. It is a structured observation with a purpose, and the senior carer needs a short brief for each one. Without it, the shadow shift becomes a walk behind someone while they do their rounds.

Give each shadow shift a goal and an observation sheet with four or five questions. AI is good at turning your competence list into those sheets.

Turn this competence list into a one-page observation sheet for a senior carer
to complete after each shadow shift. Competences: [paste list].
Format: for each competence, a line the senior carer can mark as "seen and
competent", "seen, needs practice" or "not observed", with space for a note.
Add three open questions at the end about how the starter spoke to the client.
Do not include any client names or details.

The output is a form, not an assessment. The senior carer marks it by hand or in your app, and a manager reads it before agreeing the next step. The "not observed" option is essential: it stops people ticking items that did not come up.

Deciding when a starter goes solo

Do not set the number of shadow shifts in the AI's plan and leave it there. Set a minimum in your master checklist, then require a manager decision based on the observation sheets. A simple rule that works in small agencies: the starter needs at least the minimum number of shadow shifts, no "needs practice" mark on a safety competence (moving and handling, medication, safeguarding), and a first solo visit that is short, close to the office and with a client who is comfortable with a new face.

Stage 4: The welcome pack, drafted once and personalised

A welcome pack is the cheapest retention tool a care agency has, and the part AI does best. It gathers what a nervous starter wants to know: who to call, what to wear, how pay and rotas work, what happens on a bad day.

Build it as a short document of about two pages, plus a link to the full handbook. A useful structure:

• Who is who: names and roles of the office team, with the on-call number and when it is staffed.
• Your first fortnight: the dated plan from stage 2, pasted in.
• Practical answers: uniform, ID badge, mileage, how and when pay arrives, how to request leave.
• If something goes wrong: a client does not answer the door, you feel unsafe, you are running late, you make a mistake. One short paragraph each, written to reassure.
• Who to ask: a named senior carer as a buddy for the first month.

The prompt below takes your facts and writes the first draft:

Write a warm, plain-English welcome pack (about 600 words) for a new caregiver
joining a small home care agency. Use short paragraphs and headings. Facts to use
exactly as given, and nothing else: office hours 8-5 Monday to Friday; on-call
number [number] covers evenings and weekends; pay is monthly on the 28th; rota
is published on Thursdays for the following week; buddy is [first name of senior carer].
Where a fact is missing, write "[office to complete]" instead of guessing.
Tone: kind and practical, not corporate.

That "write [office to complete]" instruction is the safeguard. Left alone, AI fills gaps with plausible details, like a pay date or a mileage rate, that may be wrong. Search the draft for any figure you did not supply before you send it.

Manager approval: where the human gate sits

The whole system depends on a gate that AI cannot open. Define it in writing: before a starter's first visit with a client, the registered manager confirms that every pre-start item is complete and evidenced, the induction items required before visits are signed, and any adjustments (health, language, driving) are in the plan. Put the confirmation in your care management system with the manager's name and the date.

One way to make the gate practical is a one-page release summary. Ask the AI to compile it from the checklist status you paste in (items and dates only, no documents):

From this status list, write a release summary for the registered manager.
Group into "complete with date", "outstanding" and "needs a decision".
Do not say the starter is cleared or ready; only report what the list shows.
List: [paste item, status, date, one per line].

The instruction "do not say the starter is cleared" is deliberate. If an AI summary ever says "ready to start", people stop reading the underlying list. Make the manager read the list, with the summary as a map of it.

Three starters in one month: the time sums

Take a twelve-caregiver agency that hires three starters in a month. The office coordinator used to spend about 45 minutes per starter writing individual plans, a further 30 chasing missing items by phone and about 30 adapting the welcome pack: 105 minutes a starter, or 5 hours 15 minutes across three. These figures are illustrative; time your own.

With the three-layer approach, the coordinator spends about 15 minutes pasting the master induction list and the starters' details, then reviews three drafts at about 15 minutes each (45 minutes). Chasing messages are drafted in bulk, about 5 minutes per starter (15 minutes). The welcome pack is edited once, not adapted three times: 30 minutes. That is roughly 1 hour 45 minutes in all, and the checking, which is the point, still takes a human. The saving of about three and a half hours matters less than the consistency: all three starters get the same structure and the same safety competences, and nobody's induction depends on who was in the office that week.

What to measure to check it worked: days from offer to first supervised visit; the number of items outstanding on day one (aim for none that block a start); the starter's own rating of the first fortnight, asked at the week-two review with a plain question such as "what did you wish someone had told you?"; and whether the same competence gaps show up across starters, which points to a weakness in the induction rather than in the people.

Three mistakes this system catches, and one it cannot

First, the silent skip. An AI-drafted plan leaves out a topic from your master list. The flag-unplaced-topics instruction and a manager comparing the plan to the master list catch this.

Second, the stale checklist. A requirement changed and the master checklist was not updated, so every starter is checked against the old version. A version date on the document and a yearly review slot in the manager's diary prevent it.

Third, the over-shared file. Someone pastes a passport scan or a background-check result into a chat tool to "summarise" it. Write a one-line rule in your office procedures: documents stay in the care system; AI sees first names, dates and item names only. Link this rule to your wider approach to privacy policy drafting for a home care agency.

The one it cannot catch is a person who looks right on paper and is wrong in practice. That is what shadow shifts, reference calls and an honest week-two conversation are for, and why recruitment screening and onboarding belong in the same loop as how you recruit caregivers and how you check in with them later.

Adapting the checklist for different starters

One master list does not fit everyone. Here are four common variations and what changes in the plan.

An experienced carer from a care home. They know personal care but may be new to lone working, travel time and home environments. Shorten the practical skills content; lengthen the lone-working, boundaries and recording sections. Shadow shifts focus on how visits are organised, not on basic tasks.

A first-time carer with no experience. Plan more shadow shifts and a slower build-up, and add a short session on dignity and communication with people living with memory loss. In the AI prompt, state "no previous care experience" so the plan does not assume it.

A starter whose first language is not the one your care plans are written in. Ask the AI to rewrite the welcome pack in plainer wording, never to translate legally important policies without a human who speaks the language checking them. Pair the starter with a buddy and give them time to read care plans before shadow shifts.

A part-time or bank worker. Compress the plan but never the safety competences. Their shadow shifts may land on different days, so the sign-off log matters even more.

For wider onboarding patterns across other kinds of business, compare a general two-week plan for a new starter, and for the policy side of a care team, a drafted employee handbook gives the welcome pack something solid to point to.

One more filled-in example: the week-two review

The week-two review is the point where the plan meets reality, so give it its own short template. Ask the starter and the buddy the same five questions, then compare the answers. A prompt that tidies the notes without inventing anything:

Here are my handwritten notes from a week-two review with a new caregiver.
Rewrite them as a half-page summary with three headings: "Going well",
"Needs support" and "Agreed actions with dates". Use only what is in the notes.
If a heading has nothing under it, write "Nothing recorded". Notes: [paste notes].

Illustrative notes might say the starter felt rushed on the first teatime call, was unsure how to log a refused meal, and wants another shadow shift on a double-handed call. The summary turns that into three dated actions: one more shadow shift on Tuesday, a ten-minute walk-through of the recording app on Wednesday, and a call from the buddy on Friday. The manager reads it, changes anything that looks wrong, and files it. Nothing in it should surprise the starter, because it is their own words in a tidier layout.

Building the master checklist and prompt bank in one afternoon

You do not need special software. A shared document tool, your care management system and an AI assistant will do. A sensible order of work:

1. Write the master checklist from your own policies and your regulator's current requirements. Give each item an owner and a "how to verify". Allow 90 minutes.
2. Save the induction prompt, the observation-sheet prompt, the welcome-pack prompt and the release-summary prompt in a single shared note, with a line at the top reminding staff what must never be pasted in.
3. Run the whole process on a past starter, using their first name only, and compare the AI plan with what actually happened. The differences show you where the master list is thin.
4. Agree who approves, who signs, and where the record lives. Print the one-page gate and keep it with the checklist.
5. Run it live for the next starter, time it, and write down what you changed.

If you would rather map the steps with someone, that is what my AI implementation consultation covers: finding the onboarding jobs worth automating and handing over something your team can run.

Onboarding questions care managers ask

Can AI decide whether a new caregiver is safe to start?

No. AI can build the checklist, track what is outstanding and draft messages, but the pre-start decision belongs to a named manager who has seen the original documents. Treat any AI suggestion that someone is cleared as a prompt to look at the file, never as the clearance itself.

Should I paste a starter's identity documents into ChatGPT?

Avoid it. The checklist and the induction plan need only the starter's first name, role and start date. Keep identity documents, background-check results and health declarations in your care management system or secure folder, where your data-protection policy already governs them.

How many shadow shifts should a new caregiver do?

There is no single right number; it depends on your regulator's expectations, the client's needs and the starter's experience. Set a minimum in your master checklist (the illustrative plan here uses three), then let the observation sheets decide whether a starter needs more before a short, supervised first solo visit. Record who signed off each shadow shift.

How often should the onboarding checklist be updated?

Review it whenever your regulator, insurer or training provider changes a requirement, and at least once a year. Keep a version date on the page and have the manager who approves starters sign the new version, so nobody works from an old copy.

Further reads

• Home Care Agency AI: A 90-Day Plan for the Owner — Place onboarding inside a wider AI plan for the agency.
• AI for Care Homes: A Registered Manager's 90-Day Implementation Plan — A registered manager's plan, useful if you also run a care home.
• How to Run Staff Surveys and Exit Interviews With AI Analysis — Ask starters how onboarding went and learn from it.
• AI Answering Service for Home Care Agencies: After-Hours Call-Outs — Show new carers how out-of-hours calls are handled.
• Volunteer Onboarding Checklist for Charities: Build One With AI — A staged onboarding checklist with an AI step for each stage, copy-ready messages, and safeguarding items drafted by AI but approved by your safeguarding lead.
• AI for Home Health Agencies: Faster Replies to Hospital Referrals — A referral intake checklist, ready-to-send missing-information requests and a response-time log, with AI under a signed business associate agreement.
• Domiciliary Care Rota Template: When Software or AI Fills the Gaps — A domiciliary care rota template with travel time and continuity rules, plus how AI can suggest gap cover while a coordinator makes the call.
• Domiciliary Care Spot Check Template: AI Summaries and Action Lists — A spot check form for home care, a fair scoring approach, and AI prompts that turn completed forms into summaries and action lists.
• Home Care Agency Policies and Procedures: A Review-First AI Method — Start from the regulator's standard, draft with AI, and put a named person through a ten-point review before any policy reaches a carer.
• Non-Medical Home Care Leads: An Enquiry-Reply Routine Using AI — A practical routine for answering home care enquiries within minutes: reply templates, after-hours cover and a response-time test.
• Nurse Licence Verification: A Tracker for Small Care Providers — A simple tracker for nurse registration checks and renewal dates, with alerts, an audit trail and clear limits on what AI should do.
• AI for Nanny Agencies: Family Briefs, Candidate Shortlists and Check-Ins — A nanny agency's guide to AI for family briefs, shortlist notes and check-in messages, with bias and data-protection safeguards.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: general home-care onboarding practice; your own regulator's current requirements for staff checks and training; vendor documentation for the AI tools you use.