Sidekick Orchestration

What AI can do for HR and people managers

AI can build a new hire's first week, answer policy questions from current sources and turn an experienced colleague's know-how into training, with a named person handling what the sources leave open.

// In this article

AI can prepare much of the groundwork in people work: a new hire's first week, answers to policy questions and training built from an experienced colleague's know-how. It works when every answer points to a current, approved source and a named person handles anything that source leaves open. Hiring, pay, performance, discipline and accommodation decisions stay with people.

This guide is for people leads and managers inside a small business. HR consultancies looking for help with client proposals and RFPs should see AI for HR consulting firms.

Before day one: build the first week from the role

Made-up example. The business, records and figures are invented to show the method. They are not a client result or output from a real AI run.

A service company is onboarding a client coordinator. The approved role brief sets three first-week outcomes: find the current scope for a client, prepare a handover and know where to raise an unresolved commitment.

The manager gives the AI these sources:

  • Role brief, version 2: the three outcomes.
  • Access checklist, version 3: the project and document access IT grants for this role.
  • Handbook, version 4 (current): working arrangements and the policy owner's contact.
  • Handover instructions, version 2: how to separate agreed work from open questions.
  • Handbook, version 3 (retired): kept for history and excluded from answers.
  • Manager's calendar: orientation on day one, a demonstration on Friday.

No medical, pay or identity documents go in. The request: a week the manager can edit, each step tied to a source and a person, plus the questions the sources leave open.

The first-week plan

WhenActivitySourceWho is responsible
Before day oneConfirm the access needed for practiceRole brief v2; access checklist v3IT grants access; manager confirms it works
MondayDiscuss the three outcomes and find the current scope for a sample clientRole brief v2Manager sets expectations
TuesdayPrepare a practice handoverHandover instructions v2Experienced coordinator checks the draft; starts only once access works
Wednesday and ThursdayReview it and practise raising an unresolved commitmentRole brief v2; handover instructions v2Manager adjusts workload and answers questions
FridayDemonstrate a handover on a new sampleRole brief v2Manager judges readiness and next support
Open question"Who can approve an exception to the remote-work arrangement?"Handbook v4 has no exception rulePolicy owner answers

Two answers from the plan's question guide:

Where do I find the work we agreed with a client? Start with the current scope document named in the project file. Keep open questions separate from agreed work. Source: handover instructions v2, "Agreed work and open questions." If the scope conflicts with what a client has said, ask your manager.

Can I get a remote-work exception? The current handbook (v4) does not cover exceptions. Ask the policy owner through the people-support contact in the handbook. Only the policy owner can approve one.

The first draft got the second answer wrong. It said "yes" and cited the retired handbook, version 3. The citation looked plausible. The manager checked it against the source list, removed the answer and left the question with the policy owner, then moved Tuesday's practice until access was confirmed. The table shows the repaired plan. A reference has to come from a current version and support the sentence beside it.

Judge the plan by preparation, review and correction time, whether access and introductions arrived, and how the new coordinator handles Friday's demonstration. A faster plan helps the manager; a new hire ready sooner is the result you want. How to tell whether AI is actually helping shows how to count the full effort.

Eight people tasks AI can prepare

These are jobs to test. Support varies by tool and plan.

People taskWhat AI can prepareWho is responsible
Role briefsResponsibilities, outcomes and interview questions from approved requirementsManager defines the work; HR checks language and conditions
OnboardingA role-specific first week with sources, owners and dependenciesManager approves the sequence; IT and trainers deliver their parts
Policy answersA plain answer, the exact current source and any open questionPolicy owner keeps the source current; HR handles individual cases
TrainingJob guides, examples and practice cases from an experienced colleagueColleague checks the procedure; manager checks how it is applied
Internal messagesDrafts of an approved change for each audienceThe named sender checks meaning, timing and any promise
Employee survey themesCombined themes and follow-up questions from responses HR is allowed to shareHR protects confidentiality and decides the response
Manager preparationAn agenda and questions for an agreed discussion goalManager holds the conversation and makes the judgments
HR administrationA list of missing forms or checklist items from permitted recordsHR confirms completeness through the usual process

Survey summaries need care: in a small team, a combined theme or an unusual comment can still point to one person. Decide what can be shared before preparing the summary for leaders. For manager preparation, keep the agenda to the agreed goal; motives, attitude and scores stay out of it.

Turn an experienced colleague's know-how into practice

Made-up example. The rules and customer messages are invented to show the method.

An experienced colleague handles requests about items that arrive damaged. The approved rule sheet says:

RuleInstruction
R1Collect the order number and a photo of the damage
R2Check replacement stock before offering a replacement; confirm any dispatch date
R3A refund needs manager approval
R4Raise safety issues at once through the safety contact

The colleague adds what experience taught them: acknowledge the trouble, ask what the customer wants and explain the next step plainly. The manager wants new starters to practise spotting missing information and the limits of what they can promise.

From those two sources, AI prepares a one-page job guide and a set of practice cases.

The practice customer writes: "The desk accessory arrived cracked. I need my money back today." There is no order number and no photo.

A deliberately wrong practice answer reads: "Your refund is approved and a replacement will leave today." Checked against the rules, it has three faults: it skips the evidence R1 requires, invents stock and a dispatch date under R2 and approves a refund that R3 reserves for the manager.

The corrected reply:

I'm sorry the item arrived damaged. Please send your order number and a photo of the damage so we can review it. I'll pass your refund request to the manager for a decision and let you know the next step once it has been reviewed.

The coaching prompt for the new starter: "Which sentence promises something we have not checked, and which rule tells you what to do?"

The experienced colleague checks that the guide matches how the job is done. The manager decides how to coach and judges how the new starter is doing; AI feedback is teaching material. The complete made-up people example shows a larger training package.

Make the practice repeatable across the team

A job guide becomes team practice once there is a routine around it. A five-line card for the damaged-item job:

  1. Job and acceptable output: a reply to a damaged-item request in which every promise matches a rule from R1 to R4.
  2. Inputs: the current rule sheet, the job guide and the customer's message, in the AI tool the company has approved.
  3. Review: the customer service lead checks each reply against the rules before it counts as acceptable.
  4. Bounded help: the experienced colleague demonstrates two cases in one agreed 30-minute slot. Tool access problems go to IT; rule and reply problems go to the customer service lead.
  5. Next decision: after the first six real requests, including correction time, decide whether to continue, change the method or stop.

Four roles make this work, and one person may hold two: the leader gives time to practise, the person responsible for the job sets the standard, IT owns access and the experienced colleague demonstrates. For the wider rollout, see a leaders-first plan for building AI champions.

Sources, permissions and privacy

A policy answer needs a current approved source, a named owner and a reference that supports the sentence. Keep general guidance apart from private employee files; mixed versions make a polished answer less reliable.

Microsoft says that Microsoft 365 Copilot only surfaces organizational data to which individual users have at least view permissions. Our inference: an HR folder shared too widely becomes easier to find through AI.

Canada's federal, provincial and territorial privacy authorities publish principles for generative AI that cover personal information. They ask for legal authority and appropriate purposes, and for extra care, including human review of outputs, in high-impact settings such as employment. Use only information approved for the tool; AI and sensitive information covers what to keep out. Sensitive individual situations need an HR or manager contact the employee can actually reach.

Microsoft and privacy-authority guidance checked November 2026.

Choose one job to start

A searchable handbook or a feature in your HR software may already answer the question; use it when it is simpler. AI earns its place when the work needs role-specific assembly, teaching material or answers from a maintained set of sources. Pick one recurring job from the table and compare the result with how the team does it today, including review and source upkeep. If nobody owns a policy, or two sources conflict, settle that first.

To see where onboarding, policy answers or training sit among your other AI opportunities, the free AI Opportunity Map takes about 3 minutes, for you or your whole company.

For companies

Get your team proficient with AI in 90 days.

Sidekick runs your AI rollout, and Chase leads it. We coach your leaders first, write your AI plan and staff policy with you, train your teams, and measure results against goals you set. Start with your free AI Opportunity Map.

Have a question first? Ask it here, text Chase at 778-999-0985, email, or connect on LinkedIn.