Human Resources

Pilot wave 3 (M9). Impact: medium. See the hub. HR runs late-wave because its use cases handle personal data (GDPR-sensitive) and depend on the knowledge base being mature enough to answer policy questions reliably. HR also carries a second, program-wide role: co-owner of the change management and training plan.

Current state (assumptions to validate)

HR administers ~109 employees across HQ, two branches, and the warehouses: contracts and hiring paperwork (ERGANI submissions), payroll inputs (payroll possibly outsourced or run by accounting — to validate), leave management, training records (including the statutory OHS training obligations of a company that itself sells OHS training), and recruitment for hard-to-fill roles (experienced ship-supply salespeople, warehouse staff, drivers).

Assumptions — to validate in discovery: team ~1–3; policies live in scattered Word files or in heads; employees ask HR the same questions repeatedly (leave balances, allowances, procedures); onboarding is improvised per hire; training records in Excel. Pain points: interruption load from repetitive questions, manual document production per employment event, slow onboarding to productivity in knowledge-heavy roles.

Target AI-first operating model

By M12 every employee self-serves policy and procedure answers from the assistant (Greek, with citations to the approved policy notes in the knowledge base); HR produces employment documents from templates with an agent pre-filling drafts; every new hire follows a structured AI-first onboarding path in the knowledge base with an onboarding assistant answering “how do we do X here” questions; training attendance and certification records update automatically from the training calendar. Hiring decisions, appraisals, and anything touching an individual’s employment terms stay entirely human.

AI use cases

Use case	Pain addressed	Data needed	Complexity	Impact	Pilot
HR policy Q&A assistant (leave, benefits, procedures) with citations	Repetitive interruptions	Approved policy notes in KB	L	M	Y
Onboarding assistant + structured onboarding paths per role	Slow ramp-up, improvised onboarding	KB department hubs, role guides	M	H	Y
Employment document drafting (contracts, certificates, announcements)	Manual document production	Templates, employee master data	M	M	N
Job-ad and interview-kit drafting for open roles	Slow recruitment starts	Role profiles, past ads	L	L	N
Training administration (records, reminders, certification tracking)	Excel tracking, statutory exposure	Training calendar, attendance	L	M	N

Process transformation opportunities

• Policy self-service: publish-once-answer-forever — every recurring question becomes an approved KB note the assistant cites; HR stops being a lookup service.
• Onboarding as a product: a standard, per-role onboarding path in the KB, improved after every hire’s feedback.
• Training compliance by default: statutory OHS training tracked with automated reminders instead of spreadsheet memory.

Required data sources

• Policy documents (Word/PDF → KB migration during M5–M8); employment templates.
• Employee master data (SoftOne HR module or payroll system — to validate); training records (Excel).
• Role descriptions, org chart (company profile).

Potential AI agents

• Onboarding Agent — guides new hires through their path, answers company-procedure questions; HR reviews escalations weekly.
• Knowledge Assistant (company-wide) — HR policy scope with strict ACL: individual employee data never enters RAG; only anonymized, approved policy content.

Automation opportunities

• n8n: joiner/mover/leaver checklist flows (accounts via IT, equipment, training enrolment); training-expiry reminders; leave-request routing in Teams.

Required integrations

• M365/Entra ID for joiner-leaver automation (integrations); employee-data source read-only for document drafting (strictly access-controlled).

KPIs

KPI	Baseline	M12 target
Repetitive HR questions answered by assistant	0	≥60%
Time-to-productivity for new hires (self-assessed + manager)	TBD (M2)	−30%
Employment document turnaround	TBD	−50%
Statutory training records complete & current	TBD	100%

Risks

• Personal data leaking into RAG → hard rule: no individual employee data ingested; DPIA covers HR use cases (REQ-008); security/data owner sign-off before pilot.
• Assistant giving wrong policy answers → only approved KB notes ingested; answers always cite the source note.
• HR capacity: the department both transforms and co-runs change management → TPL support explicitly budgeted in the change plan.

Training needs

• HR team: KB authoring and policy-note curation (M6–M8); onboarding-path design workshop (M8).
• All employees: “ask the assistant first” habit — covered by general training (M4–M5).

Deliverables

• Migrated, approved policy corpus in the KB; per-role onboarding paths; policy Q&A live (M9 pilot); training-records automation.

12-month execution milestones

Month	Milestone
M1–M2	Discovery: question-load logging, policy inventory, onboarding walkthrough
M4	Assistant onboarding (general use); HR co-launches all-staff training
M5–M8	Policy corpus migrated to KB and approved; onboarding paths drafted
M9	Wave-3 pilot: policy Q&A + onboarding assistant with next new hires
M10	Pilot evaluation; document-drafting workflow live
M11	Joiner/mover/leaver automation in production
M12	KPI review; onboarding path v2 from feedback