AI SOP Template: Employee Onboarding
ZarifEmployee onboarding is the most expensive process most companies still run by hand. PwC estimated that by 2026, 67 percent of large organizations would be using AI-assisted onboarding to compress time-to-productivity and remove the manual coordination tax. The wins are not subtle — typical SMB deployments save around $18,000 per year and cut time-to-productivity by roughly 40 percent — but the wins only land when there is a real SOP behind the AI, not a vibes-based "ChatGPT helps with HR" rollout.
This is that SOP. It is a complete, copy-and-adapt template covering the 7 days before start, day one, the first 30/60/90 days, role assignments, AI tools per step, and the prompts that make it work. Pull what you need into your handbook.
An AI-powered employee onboarding SOP is a written, role-assigned sequence of steps that takes a new hire from offer accepted to fully productive, where each step uses a specific AI tool or prompt to remove manual coordination, generate personalized content, and surface blockers in real time.
TL;DR
- The complete template covers 7 phases: pre-boarding (T-7 days), day one, week one, days 30, 60, 90, and the post-90 conversion review.
- Owners are split across HR, IT, hiring manager, and the buddy — never assign the whole flow to one role; that is how onboarding fails.
- Recommended AI stack: Notion AI or Claude Projects for content, Workday or Rippling for IT and access, Lattice or 15Five for check-ins, plus an internal RAG chatbot trained on your handbook.
- Run a weekly automated audit (every Friday) to flag any new hire who is more than 2 days behind on milestones.
- Companies that follow a structured 90-day onboarding SOP retain new hires at roughly 82 percent versus 50 to 60 percent for ad hoc onboarding (Brandon Hall Group benchmark).
Roles and ownership
Before the SOP runs, fix the org chart for it. Every step has exactly one owner and a backup. If everyone owns it, no one owns it.
People Ops or HR coordinator: owns the master timeline, sends communications, runs the AI-generated welcome content, owns the 30/60/90 review cadence.
IT lead: owns access provisioning, hardware shipping, account creation, security training assignment.
Hiring manager: owns role-specific goal-setting, project kickoffs, weekly 1:1s, the 90-day performance review.
Onboarding buddy (peer, not the manager): owns informal Q&A, culture transfer, the "stupid questions" channel, lunch in week one.
AI assistant (your internal chatbot or Claude Project): owns answering policy and process questions on demand so the human owners do not get pinged 40 times a day.
Phase 1: Pre-boarding (T-7 days to T-1 day)
Goal: the new hire feels welcomed and prepared before they walk in.
Owner: People Ops with IT support.
Day T-7: Send the welcome email. Use Claude or ChatGPT to generate a personalized note based on the hire's resume, the team they are joining, and the company's voice. Sample prompt: "Write a 200-word welcome email to [name], who is joining as [title] on [start date]. Reference their background in [2 specific items from resume] and three things they will be doing in their first month: [items]. Tone: warm, specific, not corporate."
Day T-5: IT triggers the provisioning workflow in Rippling, Workday, or Okta Workflows. Auto-create accounts in the standard tool stack (Slack, email, Google Workspace, GitHub, project tools), assign group memberships based on department, and queue laptop shipment. The AI layer here flags any access requests that fall outside the standard pattern for review.
Day T-3: Send the day-one schedule. Generate it with the AI assistant from a template — meetings, who the hire will meet, what they will do — populated from the hiring manager's calendar.
Day T-1: Send a final "see you tomorrow" message. Confirm laptop arrival, remote login if applicable, and the morning meeting link. Include a link to your internal RAG chatbot so the hire can ask any pre-day-one question without bothering a human.
Send the laptop with a printed handwritten card from the team. AI handles the digital coordination; a physical artifact handles the emotional one. The combo lands far better than either alone.
Phase 2: Day one
Goal: the hire ends day one knowing the people, the mission, the tools, and where to ask questions.
Owner: People Ops in the morning, hiring manager and buddy in the afternoon.
Morning block: 30-minute welcome session with HR (benefits, payroll, policy acknowledgments). Use a tool like Lattice or BambooHR to auto-collect e-signatures on policy docs in the background. Walk through the AI assistant — show the hire how to ask it about PTO, expense policy, and tool documentation.
Late morning: hiring manager session. Goals for the role, what success looks like at 30/60/90, who they will work with most. Use the AI assistant to generate a draft of this conversation in advance based on the role description and team OKRs; the manager edits and personalizes.
Lunch: in person if possible, with the buddy. No agenda, no work talk required.
Afternoon: tool walkthroughs. Set up Slack, email, the project management tool, and the codebase or relevant repos. Run a "first task" — usually a small, low-stakes contribution that ships in week one.
End of day: 15-minute debrief with the hiring manager. What was clear, what was confusing, what they need. Log it.
Phase 3: Week one
Goal: the hire makes one real contribution and feels embedded in the team rhythm.
Owner: hiring manager with buddy support.
Schedule the daily 15-minute manager check-in for the entire first week. The AI assistant pre-populates a one-paragraph summary of what the hire did the prior day so the manager walks in informed and the meeting stays focused.
Assign the first concrete deliverable. Engineering: ship a small bugfix or doc improvement. Sales: complete the first 5 prospect research briefs. Marketing: draft a single piece of content or analyze a campaign. The point is a closed loop — start, finish, ship, get feedback — within five business days.
Buddy lunch on day three. Standing weekly 1:1 between buddy and new hire scheduled for the next 12 weeks.
Friday: first weekly check-in via Lattice, 15Five, or your survey tool. Three questions only: what is going well, what is unclear, what do you need? AI summarizes responses across the cohort weekly so HR can spot systemic issues fast.
Phase 4: 30 days
Goal: the hire owns one workstream end-to-end and is integrated into the team.
Owner: hiring manager owns the review; HR coordinates.
Run a structured 30-day check-in. Use the AI assistant to draft the conversation guide based on the role-specific goals set on day one, and to summarize all check-in responses to date. The manager reviews and personalizes.
Confirm the hire has met everyone they need to: cross-functional partners, internal customers, key stakeholders. The buddy validates this from the peer side.
Reassess the 60- and 90-day goals based on what has actually changed. Goals set on day one are guesses; goals at day 30 are informed.
Phase 5: 60 days
Goal: the hire is producing at near-full capacity and giving feedback on the onboarding itself.
Owner: hiring manager and HR.
Run the 60-day review. Same structure as 30-day, deeper conversation. Add: peer feedback from 2 to 3 close collaborators, collected via a brief AI-summarized survey.
Critical step most companies skip: ask the hire to grade the onboarding. What worked, what did not, what was missing, what was wasted time. Feed those answers into the AI assistant with a prompt like "Summarize the onboarding feedback from the last 5 hires. Identify the top 3 themes and propose specific changes to the SOP." Update the SOP quarterly based on this.
Phase 6: 90 days
Goal: confirm a successful hire (or make the call early if not).
Owner: hiring manager owns the performance call; HR coordinates.
Run a full performance review against the role goals set at day one and adjusted at day 30. Score on outcomes, behaviors, and culture fit. The AI assistant pre-drafts the review from the manager's notes, weekly check-ins, and any objective metrics — manager edits and owns the final.
Confirm or course-correct. If the hire is on track, lock in the next 90-day plan and move them off the new-hire cadence. If they are not, this is the moment for a clear performance conversation, not month 6 when it costs more to unwind.
Send the official "you made it through onboarding" message. Move them off the new-hire onboarding tag in your HRIS so cohort emails and surveys stop.
Phase 7: Post-90 conversion review
Owner: HR and hiring manager.
Six to eight weeks after the 90-day mark, audit the cohort. Did retention hold? Did time-to-productivity actually hit the target? What did the onboarding miss?
Pipe these answers back into the SOP. The whole point of having an SOP is that it improves; otherwise it is just a checklist gathering dust.
SOP at a glance
| Phase | Owner | Key milestone | AI tool |
|---|---|---|---|
| T-7 to T-1 (pre-boarding) | People Ops + IT | Welcome email, accounts provisioned, laptop shipped | Claude or ChatGPT, Rippling/Workday |
| Day one | People Ops + manager + buddy | Mission, people, tools, AI assistant intro | Internal RAG chatbot, Lattice |
| Week one | Hiring manager + buddy | First deliverable shipped, daily check-ins | 15Five, AI standup summarizer |
| Day 30 | Hiring manager + HR | Owns one workstream, goals refined | Lattice or Workday |
| Day 60 | Hiring manager + HR | Peer feedback, onboarding feedback collected | AI feedback summarizer (Claude Project) |
| Day 90 | Hiring manager | Full performance review, conversion decision | AI review drafter, BambooHR |
| Post-90 | HR | Cohort audit, SOP improvement loop | Notion AI, internal analytics |
The AI assistant: the single highest-leverage piece
If you only deploy one piece of this SOP, deploy the internal RAG chatbot trained on your handbook. It absorbs the question volume that otherwise hits your HR coordinator and your buddy, makes policy answers consistent, and gives the new hire a 24/7 backstop.
Build it on Chatbase, Voiceflow, or a custom OpenAI or Claude RAG pipeline. Index the employee handbook, benefits docs, expense policy, security policy, IT runbooks, and any team-specific onboarding wikis. Update weekly. Track every question that returns "I do not know" — those are your content gaps.
Do not let the AI assistant replace the buddy or the manager 1:1. The bot answers process and policy questions; humans answer "is it normal that I feel underwater right now?" Confusing the two will tank your retention.
FAQ
How long should an AI-assisted employee onboarding SOP take to roll out?
A working v1 takes 4 to 6 weeks for a 50-to-200-person company: 1 week to map the current process and decide owners, 2 weeks to build the AI assistant and wire up the IT provisioning automation, 1 to 2 weeks to pilot with the next 2 to 3 hires, and ongoing iteration based on feedback.
What AI tool should we use for the internal HR chatbot?
For most teams under 500 employees, Chatbase or CustomGPT is the fastest path — pay $19 to $399 per month, point it at your handbook, ship it in a day. Larger teams or anyone with strict data requirements should build a custom RAG pipeline on Claude or OpenAI with their own vector database.
Can AI replace the human buddy in onboarding?
No, and trying is the most common mistake. The AI handles policy, process, and FAQ traffic. The human buddy provides culture transfer, social integration, and the kind of ambient support that makes someone feel like they belong. Both are needed; they do different jobs.
How do we measure if the onboarding SOP is working?
Track four metrics: 90-day retention rate (target above 80 percent for white-collar roles), time-to-first-meaningful-contribution (target inside week 1), engagement score from a 30-day pulse survey, and manager satisfaction with hire readiness at day 60. If three of four trend the right direction quarter over quarter, the SOP is working.
What is the biggest legal risk of AI-assisted onboarding?
Two areas. First, automated decisions about access or training assignment that effectively make protected-class distinctions — keep humans in the loop on anything affecting opportunity. Second, retention of employee data inside third-party AI tools — use enterprise tiers with no-training agreements and a clear data processing addendum, especially in EU jurisdictions under GDPR.
How often should we update the onboarding SOP?
Quarterly at minimum, with a hard rule that any 30/60/90 feedback theme that appears in 3 or more hires triggers an immediate SOP edit. Onboarding is one of the few processes where the new hire is the best critic — they are seeing it for the first time, just like the next person will.
