Stand Up A Realistic Tenant From One Spec
Setup Wizard Agent is the agent behind the MangoApps Console "new tenant" wizard. Give it an industry plus headcount and it generates a believable org — diverse names, departments, roles, locations, manager hierarchies, comp ranges — then layers on realistic shifts, recruiting pipelines, service desk tickets, EPMS data, and the rest. Every record carries is_sample_data: true so the 22 confirmation-gated cleanups can revert it cleanly.
Why Hand-Building A Realistic Tenant Eats A Week
Stand up a believable demo or QA tenant by hand and you'll spend a week clicking through admin pages — and you'll still end up with an org that looks fake. Setup Wizard Agent runs the same creation flow Ruby code already exposes, faster and reversibly.
Demo Tenants Look Fake And Buyers Notice
Three users named "Test User 1, 2, 3" all in the same department. No manager hierarchy. No comp ranges. A demo with that data spends the first ten minutes explaining what's wrong with the demo — not what's right with the product.
QA Tenants Take Two Days To Stand Up
Setting up a 50-user tenant by clicking through Console Admin: users, departments, locations, manager assignments, then shifts, then EPMS data, then a recruiting pipeline. Two days of tab-juggling for an environment that you'll throw away after one feature ships.
There's No Clean Way To Reset A Test Tenant
You stood up a test tenant, ran experiments, and now you want it back to a clean state. The records aren't flagged. Removing them by hand risks taking real data with them. The "reset" becomes "scrap the tenant and start over."
Sample Data Drifts From What The Product Actually Needs
The product ships a new feature that requires a column. The hand-built sample data doesn't have it. Every demo tenant breaks in the same place at the same time. The agent runs the same data-creation services the product ships, so sample data evolves with the schema.
Industry-Specific Demos Mean Re-Editing The Same 200 Records
A healthcare prospect wants to see nurse scheduling, a logistics prospect wants warehouse shifts, a retail prospect wants store managers. Every industry pivot means renaming departments, retitling roles, and rewriting the org chart by hand. The same demo tenant gets rebuilt three times a week.
Reversibility Becomes A Spreadsheet Of "Things To Delete Later"
The SE seeded users, shifts, and pipelines for a Tuesday demo. By Friday they need the tenant clean for the next one. Without a tracked manifest of what was created, the cleanup is grep-and-pray — and inevitably one stale recruiting pipeline survives into the next buyer's screen-share.
Setup Wizard Agent At A Glance
Setup Wizard AI
Spec-driven org provisioning + 14 domain sample-data generators + 22 reversible cleanups.
Inside Setup Wizard Agent — The Actual Capabilities
Every block below maps to real tools the agent runs against the same data-creation services the product ships. Reads count what's there; writes create realistic records flagged for cleanup; cleanups are confirmation-gated and only touch flagged rows.
Generate A Believable Org In One Call
Hand the agent a spec — industry, departments with role breakdowns, locations, skills, comp ranges — and Ruby handles diverse name generation, email creation, hierarchy wiring, and record creation. Max 200 users per org. Validate first, then create — the agent runs validate_organization_spec before create_full_organization to catch issues early.
- Validate before create — validate_organization_spec returns errors before any record is written.
- One-call org creation — create_full_organization handles departments, roles, users, locations, skills, manager relationships, and comp ranges from a single structured spec.
- Diverse name generation — Ruby picks from a curated diverse name pool when names aren't provided; emails generated to match.
- Industry-aware — pass industry: "home_services" and the spec defaults pull from sensible departments and roles.
Follow-Up Edits Without Re-Running The Whole Spec
Once the org is up, the agent runs surgical follow-ups — add a department, add a location, add users, add skills, set comp, assign users to locations, set manager relationships. Each is a separate tool with a focused parameter shape, so adjustments don't require rebuilding the whole tenant.
- Add departments / locations / skills — create_departments, create_locations, create_skills; each a focused upsert.
- Add users — create_users generates names from a diverse pool when none provided; honors role / department / location.
- Set manager hierarchy — set_manager_relationships identifies people by full name; idempotent.
- Assign locations + set compensation — assign_user_locations and set_compensation with per-user salary / hourly history rows.
14 Domain Generators · Status / Create / Cleanup Triples
Each major product area has a status / create / cleanup triple — see what's there, generate a realistic data set, and roll it back when you're done. Generators honor pinned wizard targets (users / locations) so the data lands on the people you've already chosen as the demo's heroes.
- 14 status tools — get_current_status, get_epms_status, get_shifts_status, get_recruiting_status, get_service_desk_status, get_hr_operations_status, plus 8 more across knowledge / employee experience / workplace / advanced / onboarding / referral / talent / interview / broadcast / offer / background-check / skills-cert / compensation-approvals / workforce-bots domains.
- 14 create tools — paired with each status; e.g. create_epms_data populates goals, OKRs, reviews, feedback, competencies; create_shifts_data generates schedules, shifts, attendance, leave, timesheets, marketplace claims, AND post-shift feedback.
- Pinned wizard targets honored — the wizard's chosen users / locations become the heroes of every domain's generated data set automatically; override per-call if needed.
- LLM-assisted scenarios — themed parameters (theme / industry / role_types) trigger an LLM call to generate scenario-specific tickets, jobs, candidates, interview feedback; falls back to a static pool if the themed call fails.
Shifts Generators Cover The Whole Marketplace Surface
create_shifts_data is the marquee generator — it stands up enough realistic shift-marketplace activity to make the entire scheduling app demo on. Open shifts, listings, offers, swaps, claims, timesheets, attendance, post-shift feedback — every row flagged for cleanup.
- Full shifts package — create_shifts_data generates schedules, shifts, assignments, attendance (clock in/out), absence reports, leave requests, timesheets, schedule approvals, and Shift Feedback rows.
- Marketplace primitives — create_open_shifts, create_shift_listings, create_shift_offers, create_shift_swaps, create_pending_shift_claims — every state in the marketplace flow.
- Per-user timesheet histories — create_timesheets_for_user generates 4 past weekly timesheets with varied statuses plus 2 upcoming pending timesheets for one user.
- Date-window aware — pass start_date+end_date or specific_dates or a single bound (other side defaults to ~6 weeks).
Reversible · 22 Confirmation-Gated Cleanups
Every generator pairs with a cleanup_sample_* tool. Each cleanup is flagged RISKY_TOOLS, requires explicit confirmation, and only removes records where is_sample_data: true. Production records that don't carry the flag are never touched. cleanup_shifts_in_range is the one exception — it deletes by date range regardless of flag and is gated even more carefully.
- 22 risky cleanups — every domain generator has a cleanup; cleanup_sample_data for the org rollup, plus 21 domain-specific cleanups.
- is_sample_data flag is the gate — cleanups only remove flagged rows; production records are never touched.
- cleanup_shifts_in_range is the exception — deletes by date range regardless of flag; only used by admins explicitly resetting a window.
- Every cleanup confirms first — the agent never deletes without explicit "yes, proceed".
Outcomes Internal Ops Can Measure
Setup Wizard Agent's job is to make standing up a believable tenant a 10-minute job instead of a multi-day click-tour, and to keep test/demo environments reversible. Measure against your pre-agent baseline.
- Time to provision a believable demo tenant — median minutes from spec to "ready for demo" vs hand-built baseline.
- Demo tenant "looks-real" rate — share of demo tenants that pass an SE qualitative review without obvious fake-data tells.
- QA reset cycle time — median minutes to roll a tenant back to clean state between feature tests.
- Generator-vs-product drift incidents — count of demos broken by schema changes (generators that call the product's own data services should hold this at zero).
- Cleanup precision — share of cleanups that touch only flagged rows (canary on is_sample_data hygiene).
22 Risky Cleanups · 73 Safe Tools · All Internal-Ops Scoped
Setup Wizard has 95 tools — by far the heaviest tool count of any agent in the platform. The 22 cleanups are RISKY_TOOLS and require confirmation. The remaining 73 tools are status reads and create operations that flag every record with is_sample_data: true. This agent is intended for internal ops (Console / demo / QA setup), not end-user surfaces.
- 22 risky cleanups — every cleanup confirms before running; only removes flagged rows.
- 73 status + create tools — reads count what's there, creates produce realistic sample data tagged for cleanup.
- Calls the product's own data services — HrOperationsSuiteService, ShiftsDataService, ServiceDeskSampleService and similar; the same code paths the product uses, so generators don't drift from product schema.
- Audit trail on every tool call — reads, creates, and cleanups all log requesting admin, tool, parameters, and outcome to the Console audit log.
WHAT TEAMS TRY INSTEAD
The four alternatives — and why none of them produce a believable tenant that cleans up without leftover sample data
Internal sales-engineering, implementation, and demo teams have built their own provisioning routines for years. The honest gap is that most options produce thin demo data, run through code paths the product itself doesn't use, or leave sample rows scattered after cleanup.
Pasting specs into ChatGPT, Claude, or Copilot
General-purpose AI describing what sample data should look like
- Generates the tenant directly through the product's own data services — not a CSV import
- Tags every record with <code>is_sample_data: true</code> so cleanups can revert it cleanly
- 22 confirmation-gated cleanups guarantee the demo tenant doesn't leak into a real-customer environment
Implementation consultancies and demo-data services
Outside vendors building tenants that drift from product reality
- Calls the product's own services (<code>HrOperationsSuiteService</code>, <code>ShiftsDataService</code>) — no schema drift between generator and product
- Hands off the same code path the product uses, so a fresh feature is generatable the day it ships
- Internal team owns the wizard — no consulting retainer for each new industry or headcount profile
Custom seed scripts and SQL fixtures
An internal team's evergreen project that lags every new app launch
- 73 status and create tools cover every domain — no per-app seed file to maintain
- Calls live data services — no per-app fixture update when a column or validation changes
- Audit trail on every tool call — reads, creates, and cleanups logged to the Console audit log
The manual fallback — click through every screen to build a tenant
A two-day setup ritual the sales-engineering team runs every demo
- A believable tenant — diverse names, departments, roles, manager hierarchies — generated from one spec
- Realistic shifts, recruiting pipelines, service desk tickets, EPMS data, and more layered automatically
- Cleanup is 22 confirmations instead of two days of orphaned-record hunting
PLATFORM LEVERAGE
Setup Wizard Agent inherits everything the product runs
A standalone provisioning tool has to plumb every domain. Setup Wizard calls the product's own services — there is no provisioning code path that the product itself doesn't already use.
Confirmation-gated cleanups
22 cleanups, each confirmation-gated, only removing rows flagged is_sample_data: true. No way to wipe a real customer's data.
Calls the product's own services
HrOperationsSuiteService, ShiftsDataService, ServiceDeskSampleService — the same code paths the product uses, so generators never drift from product schema.
73 status + create tools
Reads count what is there. Creates produce realistic sample data tagged for cleanup. Every domain covered without a per-app seed file.
Spec-driven realism
Industry plus headcount produce a believable org — diverse names, departments, roles, locations, manager hierarchies, comp ranges. Not a uniform fixture.
Audit trail on every call
Reads, creates, and cleanups all log requesting admin, tool, parameters, and outcome to the Console audit log.
Console-scoped
Runs from the MangoApps Console — only internal admins can invoke. Customers never see Setup Wizard from inside their tenant.
INDUSTRY FIT
Industries the wizard generates with the most realism
Setup Wizard is industry-aware. Sales-engineering and implementation use it most for the verticals where demo realism matters most.
Retail
Store-level org charts, seasonal hiring patterns, and shift coverage modeled on real retail operations — district managers and store associates included.
Healthcare
Clinical roles, credentialing requirements, and shift patterns that match real provider operations — charge nurses, hospitalists, and floor staff included.
Manufacturing
Plant-floor crews, supervisors, and shift patterns that match real manufacturing operations — safety certifications and toolbox-talk participation included.
Hospitality
Property-level org charts, F&B and front-office roles, and shift patterns that match real property operations.
Financial Services
Branch and team structures, license tracking, and supervision chains that match real financial-services operations.
Public Sector
Agency org charts, role-based access patterns, and audit-trailed sample data that match real public-sector operations.
WHY MANGOAPPS WINS
An embedded setup wizard beats a generic AI, an implementation consultancy, or a custom seed script library on every axis
The argument sales-engineering, implementation, and demo-ops all share — and the one ChatGPT or a consultancy structurally cannot answer.
Cheaper than the alternatives
No implementation consultancy retainer, no per-demo data services, no engineering team maintaining a per-app fixture set.
More secure
Confirmation-gated cleanups, console-only access, every action logged to the Console audit log. Sample data never bleeds into a real-customer environment.
Easier to deploy
Runs from the Console. Spec in, tenant out — no per-tenant onboarding, no manual setup.
Easier to use
Internal admins invoke from chat. A believable industry-shaped tenant returned in minutes — no two-day click-through.
Easier to manage
is_sample_data: true on every generated record. Cleanups are 22 confirmations. No orphan-row hunting after a demo concludes.
Easier to extend
New product features become generatable the day they ship — the wizard calls the same services the product uses, so adding a domain is one service call.
AI is actually better
A generic AI can describe what a tenant should look like. Only Setup Wizard can call the product's own data services to build it, tag every row for cleanup, and revert it on confirmation.
At our head office within HR, marketing, and finance, we use ChatGPT quite a bit, and MangoApps AI Helper is a useful tool for us. As I go around to the stores, I’ve been showing them how the AI Intranet Helper can find things and answer plain language questions about our resources, and they’re quite excited about that.
Alison van Zijl
Head of Human Resources
Full House
Customer Success
Related Customer Stories
Customer Case Studies
Video Case Studies
Customer Case Studies
Customer Case Studies
Video Case Studies
Customer Case Studies
Frequently Asked Questions About Setup Wizard Agent
95 tools across 14 product domains. Org provisioning (validate_organization_spec, create_full_organization + 8 surgical follow-ups), 14 status / generator pairs (org, templates, EPMS, shifts, recruiting, service desk, HR ops, knowledge, employee experience, workplace, advanced, onboarding, referral, talent, interview scheduler, broadcast, offer manager, background checks, skills certifications, compensation approvals, workforce bots), and 22 confirmation-gated cleanups.
The cleanups are gated by is_sample_data: true — they will not touch production records that aren't flagged. Generators that ADD data to a production tenant will tag every row they create with is_sample_data: true, so a follow-up cleanup can remove them cleanly. The one exception is cleanup_shifts_in_range, which deletes by date range regardless of flag — that one is for explicit window resets only and should be treated more carefully than the rest.
Because each product domain owns its own status / create / cleanup triple, and the platform has 14 product domains plus the org-level primitives. Bundling them into fewer tools would have produced very large parameter shapes that the LLM struggles to use reliably. The one-tool-per-action shape is the trade-off — wider surface for better reliability.
The MangoApps Console "new tenant" wizard, demo / SE workflows, QA tenant setup, and occasionally for adding sample data to a real tenant that's bootstrapping. It is NOT registered for end-user routing — it's scoped to internal ops via the Console.
Time to provision a believable demo tenant, demo "looks-real" rate (SE qualitative review pass), QA reset cycle time, generator-vs-product drift incidents (should be zero since generators call the product's own services), and cleanup precision (canary on the is_sample_data flag).
Let's Talk
Since 2008, we've been building the workforce platform — earning the trust of 2 million+ users and an NPS of 78.
Why Choose Us?
- AI-Powered Platform: The most unified workforce experience on the planet.
- Top Security: HITRUST, ISO & SOC 2 certified.
- Exceptional UX: Delightful on mobile and desktop.
- Proven Results: 98% customer retention rate.
Trusted by Legendary Companies:
