Redesigning hiring and onboarding for a restaurant network

The problem

The people function had commissioned research with three questions: how do the current services enable the end-to-end team-member experience; how are recruitment and onboarding enabled; and what is expected of managers at each level. The evidence said the pain was structural. Systems were disconnected across channels. Candidates and managers re-keyed the same information. Nothing reported on recruitment performance. Processes were bypassed and varied by franchisee group. Interviews happened in dining rooms and food courts. First shifts began with trainees standing in hallways because uniforms had not arrived.

Two internal mantras framed the work, that the customer experience never exceeds the team-member experience, and that people come first and sales and profit follow. Neither had ever been operationalised for hiring. Manager churn was rising, which compressed the long tenure path from crew member to restaurant manager that fed the pipeline.

What I did

• More than 50 episodic interviews: in restaurants with managers, assistant managers, shift supervisors, crew members and trainees, and above the restaurant with the people function, technology, operations, marketing, finance, data and franchisee leaders. Quotas oversampled first-job young people, parents and franchisee groups of different sizes.
• Observation, including sitting in on a live interview and a promotion-board review.
• A field survey of restaurant teams, a document review across operational, recruitment and technology material, and the funnel and turnover data the business already held.
• Synthesis into evidence-grounded personas, seven moments that matter with a pass or fail verdict on each, and a layered current-state blueprint across nine stages.
• A manager competency framework across three areas and three proficiency levels, to rebuild the top of the pipeline rather than only the bottom.
• The target-state workflow: screening questions in a conversation, auto-scheduled interviews for high-fit candidates, reminders and dispositioning, work-rights verification with a human review lane, and automated employee-record setup, all written as a versioned specification.

What the research found

The admin burden is the experience. Candidates uploaded the same documents more than once even when everything was correct. Managers re-keyed candidate data across systems. Offers landed in spam. The wait for a first human touch ran from days to weeks.

Five of the seven moments that matter failed basic expectations. Finding the job, applying, the interview, the offer and starting all failed; considering the job and accepting it were merely weak. The experience was optimised for administration, not for the person going through it.

Two moments had no data at all. Nobody measured whether a hired crew member turned up to their first shift, and nobody measured retention at ninety days. The absence became a documented finding with an owner, not an invented number.

Work rights were the hardest part of the experience. First-job candidates did not know what documents they needed; consent forms expired; payroll discovered wrong documents at pay time. This is the step that most needed a person and least had one.

The manager pipeline needed a standard, not a rescue. Managers carried recruitment inside a full day, capability varied, and nothing said what was expected at each level. The competency framework answered the third research question directly.

Three decisions

What was delivered

The workflow was written as a specification the technology team could version, review and test, rather than as tribal knowledge. This is the core of it.

sop: HIRE-001   Crew hiring workflow, application to first shift   version 2.0
owner: people_function_recruitment_lead
compliance_locked: true        # age and parental-consent changes need sign-off
screening:
  questions: [minimum_age, work_rights, availability, shifts_per_week, high_demand_shift]
  auto_schedule_if: high_demand_shift == YES AND shifts_per_week in [3, 4, 5]
  review_queue_otherwise: true
  underage_branch: parental_consent_flow_first   # never disposition before consent is attempted
interviews:
  scheduling: auto_scheduled, calendar_sync
  manager_actions: [confirm, invite, reject]
  reminders: [24h, 48h, 72h]
  disposition: after_10_days_no_action
work_rights:
  verification: automated_checks            # document, biometric, registry lookup
  human_review_lane: payroll_admin          # accept or reject with a reason emailed
onboarding:
  employee_record: auto_created
  pushes: [scheduling_12pm, scheduling_6pm]
  payroll_validation: full_pack_push_back_on_failure
rehire_branch:
  same_franchisee_group: skip_work_rights_and_onboarding, transfer_record
  other: full_workflow, new_records
kpis: [application_to_interview_days, funnel_dropoff, time_to_offer, first_shift_attendance, retention_90d]

Also delivered: the seven-moments verdict board, the persona set, the current-state blueprint below as a living document, the manager competency framework (three areas, three proficiency levels), conversation guides for the moments that stay human, and a manager one-pager for day one.

Nine journey stages across the network: click any cell to expand the detail; fail points F-01…F-09 are marked and owned.

Stage / lane	Identify	Search	Consider	Apply	Interview	Offer	Accept	Onboard	Start
Candidate (frontstage)	Where do I even find jobs? Careers site, word of mouth, walk-ins: with no clear path from customer to candidate.	Browses roles; hours unclear Role pages and store listings, but job posts stay live and duplicates confuse.	Asks family; weighs a first job Nothing answers "what is it really like?": no employer brand separate from the consumer brand.	Long form; repeats info per store The form asks more than it decides: many visitors never complete it.	Dining-room table; rushed No quiet space, no rapport: "we walked upstairs to the dining room, I gave my availability and that was it."	Offer lands in spam Slow confirmation; offers lost to spam filters; candidates vanish.	Documents repeated; consent expiry The hardest part of the experience: same documents asked repeatedly; consent forms expire after seven days.	Online training; no device Training quality varies with trainer availability: incomplete training spills into shifts.	Uniforms missing; hallway training First shift begins not ready; no data on first-shift attendance at all.
Backstage & people	Managers hire constantly; no campaigns Hiring is constant and uncoordinated: "too hard to coordinate" campaigns.	Job posts left live; duplicates Stale listings and duplicate roles across stores and franchisee groups.	No employer brand No dedicated employer content; the consumer brand stands in for it.	Managers re-key into systems Candidate data re-entered across systems: error-prone and slow.	Manager capability varies Interviews are rushed; the dining room is not a quiet space.	Slow confirmation Offers wait on manual paperwork and manager availability.	Payroll finds errors at pay time Wrong documents surface only at pay time: the admin of starting is the friction.	Trainer availability varies "It's hard to learn when my shifts are spread out."	Manager absorbs admin Recruitment sits inside an already full day-to-day role.
Systems & automation	Careers site; no employee channel No employee-dedicated channel from customer to candidate.	Careers site + recruiting platform Two surfaces, duplicate listings, no single source of truth.	No content layer Nothing answers "what is it really like?"	Recruiting platform; AI assistant Multi-system form today; conversational AI assistant in the future state.	Auto-scheduling, calendar Future state: auto-scheduled interviews with instant confirmation.	Automated offer Future state: offers generated and tracked automatically.	Work-rights check; payroll Automated checks with a human review lane; guided capture for first-job candidates.	Learning platform; employee records Employee record auto-created with two daily pushes to scheduling and payroll.	Time & attendance; scheduling First-shift and 90-day retention have no data: documented findings.
Failure points	F-01 No path from customer to candidate No clear route into the funnel; hiring is invisible to the customer base.	F-02 Stale roles; duplicates Stale listings and duplicate roles confuse candidates.	F-03 Nothing answers "what is it really like?" No employer brand separate from the consumer brand.	F-04 Asks more than it decides nearly half drop-off: the form's middle loses the most visitors.	F-05 No quiet space; no rapport Dining-room interviews; rushed and inconsistent.	F-06 Offers lost Offers land in spam; candidates vanish without response.	F-07 Hardest part of the experience Documents repeated; consent expiry; first-time bank and tax setup.	F-08 Incomplete training Training spills into shifts; quality varies with trainer availability.	F-09 + W No data on first shift First-shift attendance and 90-day retention: no data: documented findings.

The product the research argued for

Ten modules drawn from the nine-stage blueprint and designed in September 2026. Seven belong to the candidate, because the research found the admin burden was the experience, and three belong to the manager, because the wait that made it worse was invisible while it was happening. They are built on the same design system as the safety reporting and workforce products.

Module 01Finding a jobSearch by suburb, with the ad in the same place the search was. Module 02The five questionsScreening as a conversation, and an assistant that says what it is. Module 03Consent firstThe under-18 branch runs before anything is decided, not after. Module 04Booking her own interviewReal slots from the manager, and a room rather than a dining table. Module 05Work rights, sent onceRead by everything downstream. No check ends an application; a person does. Module 06The offerSomewhere to live other than an inbox, and a rate that is explained. Module 07Before the first shiftWhat to bring, which door, one name to ask for. Module 08The pipelineThe waiting time on every card, where it can still be acted on. Module 09The review queueEveryone the rule could not place, with their answers attached. Module 10The human laneA machine may not end an application. A person decides, with a reason.

Explore the full concept, with the design system and every decision traced to the finding behind it.

Working files, anonymised: the persona set. The brief, the journey map, the proposals, the research summary and the blueprint stayed working documents with the business.

How it would be measured

What I would do differently: set the first-shift attendance and ninety-day retention baselines before designing anything, since two moments had no data at all.

Sponsor, scope and gates

What AI did and what stayed human

Fieldwork stayed human: consent, rapport and the decision to observe a real interview are not delegable, and neither were the interview quotas or the reframe. After the recording, AI helped: transcripts were pseudonymised and processed into evidence items with a participant code and a line range, every quoted line was checked against the transcript before it was kept, and the funnel and application exports were normalised into one leakage view. The personas, the moments verdict, the blueprint and the workflow design were people's work, reviewed with the people function at every gate.