The Leaderboard, In The App They Already Have
Most of the people a contest is for will never open a contests dashboard. They're on a phone, between tasks, and the only question they have is "how am I doing?" Contests Agent answers it in chat — your rank, who you're chasing, what's still joinable, which picks lock tonight — then lets you join, withdraw, pick, or launch, under the same rules as the app.
Why Workplace Contests Quietly Fizzle
The scoring engine is rarely the problem — MangoApps contests already score themselves from the verified activity ledger. What kills contests is attention. People forget they're running, never see where they stand, and miss the one fact that would have made them try.
Nobody Remembers The Contest Is Running
It launched with a post on day one. By day four it's below the fold, and the people it was designed to motivate — the ones who don't sit at a desk — have no idea a two-week pickup challenge is live, let alone that they're already on the board.
The Leaderboard Lives Somewhere Frontline Staff Never Go
Standings refresh every five minutes, which is worth exactly nothing if you have to open a dashboard to see them. A warehouse associate on a phone between tasks isn't navigating to a contests page — so the leaderboard is invisible to most of the roster.
You Don't Know You're Two Points Off First
That single fact is the whole motivational engine of a contest, and it's the one thing an email digest and a break-room poster can't tell you. "Rank 4, two points behind Marcus, three days left" is what makes someone pick up one more shift.
Picks Don't Get Made Before They Lock
Bracket and pick'em pools die on the same failure: the pick window opens, nobody opens the app, and half the pool is locked out at tip-off. The participation collapse isn't apathy — it's a missed notification and a form nobody got to.
Admins Can't Tell Whether The Contest Moved Anything
The prize got paid and the win card got shared, but did participation actually rise? Was there daily-active lift against a control group? Did the behavior stick for four weeks? Without that, the next contest is a guess, and the budget conversation is a vibe.
Launching A Contest Feels Like A Project
A manager who wants a two-week safety sprint has to find the create form, pick a metric, set a window, build a prize table, and route it for approval. It's fifteen minutes and some uncertainty — so most managers simply don't, and contests stay an HQ-only activity.
Contests Agent At A Glance
AI Contests
Standings, your rank, joins, picks, and template launches — all in chat.
Inside Contests Agent — The Actual Capabilities
Eleven tools against the Contests app: seven read-only, four write tools that require explicit confirmation. Reads answer "what's running and how am I doing"; writes are limited to joining, withdrawing, making a pick, and creating a scheduled contest from a template. Everything else — approving a contest, entering prediction results, editing a live contest, posting manual scores, voting, judging — stays in the app on purpose.
How Am I Doing — Rank, And Who You're Chasing
The question the whole app exists to answer, asked the way people actually ask it. get_my_contests returns everything the asking user is enrolled in with their rank and score in each, plus what's still joinable. get_contest_standings goes deeper on one contest — the top entrants, the user's own rank, and specifically the entrant directly ahead of them.
- get_my_contests — the user's enrolled contests with rank and score in each, plus joinable ones they haven't entered.
- get_contest_standings — top entrants for one contest, the user's own rank, and who they're chasing. Accepts a contest name, fuzzy matched.
- Five-minute freshness — the agent reads the same standings snapshot the leaderboard and signage slides read.
- Scored from verified activity — metric contests come off the same ledger that powers Levels & Badges. No scorekeeper to ask.
What's Running — And Joining Or Stepping Out
list_contests surfaces what's live (or scheduled, or finished) with an optional name search. From there the agent can enroll the asking user in an individual contest, or withdraw them from one they joined — both confirmation-gated, both bound by the same eligibility rules the Join button enforces.
- list_contests — contests in the business, defaulting to live, filterable by status and searchable by name.
- join_contest — enrolls the asking user in an individual contest. Confirmation required. Team contests enroll the whole team automatically, so there's nothing to join.
- withdraw_from_contest — withdraws the user from a contest they joined; their scores stop counting. Confirmation required, and not available on finished contests.
- Strictly opt-in stays opt-in — the agent can only enroll the person asking, never a colleague, and only where the audience rules already allow it.
Picks Before Lock — Pick'em And Brackets In Chat
The pick window is the single most time-sensitive thing in the app, and it's exactly what chat is good at. get_my_picks shows open events still awaiting a pick, picks already locked, and results so far. make_prediction records or changes a pick — fuzzy matching both the event and the option — right up until it locks.
- get_my_picks — open events awaiting a pick, locked picks, and results to date for one prediction contest.
- make_prediction — records or changes the user's pick on an event until it locks. Confirmation required. Event name and option are both fuzzy matched.
- Changeable until lock, frozen after — the agent honors the same lock semantics as the app; crowd consensus still stays hidden until picks close.
- It cannot enter results — organizers post real-world outcomes in the app. The agent reads picks and outcomes; it never decides them.
Launch From A Template — Without It Feeling Like A Project
"Run a two-week safety sprint starting Monday" becomes a scheduled contest built from the gallery — the 32 OOTB templates plus your tenant's own saved ones. Pass the closest template slug; if it misses, the agent hands back the full catalog to pick from. Creator tier, the approval queue, and regional works-council policy all still apply.
- create_contest_from_template — creates a scheduled contest from a template (pickup-challenge, safety-sprint, learning-blitz, photo-contest, wheelbarrow-challenge, and the rest). Confirmation required.
- Sensible defaults — starts in 7 days unless told otherwise; duration defaults to the template's own.
- Policy applies unchanged — creator tier, approval routing, and works-council restrictions are evaluated server-side, same as the create form.
- Create only, never edit — the agent can't modify an existing contest. Live-contest changes and audited overrides stay in the app.
Admin Reads — Approvals Waiting, Budgets Left, Did It Work
Three admin-only read tools cover the questions a program owner asks between meetings: what's stacked up in the approval queue, how much of each delegated prize budget is actually left, and whether the contest moved any numbers — participation rate, DAU lift against a non-enrolled control, and post-contest retention.
- get_pending_contest_approvals — contests awaiting approval. The agent lists them; approving or rejecting happens on the approvals page.
- get_prize_budgets — delegated budgets with holder, allocated, reserved, spent, and remaining.
- get_contest_analytics — participation rate, daily-active lift vs baseline, and behavior retention for one contest.
- Admin-scoped server-side — a non-admin asking these questions doesn't get an admin answer.
Outcomes Teams Can Measure
Contests Agent is judged on whether more of the roster actually plays, and whether the contest moved a number. All six are measurable against your pre-agent baseline.
- Participation rate vs pre-agent baseline — share of the eligible audience enrolled per contest, before and after the agent.
- Share of standings checks via chat — how much of "how am I doing?" moved off the dashboard and onto the surface frontline staff actually use.
- Pick-completion rate before lock — share of pool entrants with a pick recorded at lock time, the cleanest read on prediction-contest health.
- Contest launch rate by managers — contests created from templates by non-admins per month, the signal that launching stopped feeling like a project.
- DAU lift vs control — daily-active lift for enrolled users against a non-enrolled control group, straight out of get_contest_analytics.
- Withdrawal rate — share of joiners who withdraw, the honest counter-metric that catches contests people regret entering.
Four Write Tools · Zero End-Runs Around Policy
Contests Agent's RISKY_TOOLS list is not empty — join_contest, withdraw_from_contest, make_prediction, and create_contest_from_template all write, and all require explicit confirmation. The governance point that matters more: chat is not a side door. Every eligibility and policy rule the web app enforces is enforced identically here, server-side. An employee cannot use the agent to enter a contest their audience rules bar them from. A manager cannot use the agent to skip the approval queue, exceed their creator tier, or launch an individual contest in a region running works-council mode.
- Four gated writes, named — join, withdraw, make a prediction, create from a template. Each one confirms before it acts.
- Eligibility is not re-implemented — audience rules and manager exclusion from ranking their own reports apply on the chat path exactly as on the Join button.
- Approval queue is unskippable — a manager-created contest from chat lands in the same queue, and the agent can only list that queue, never approve or reject from it.
- Works-council mode holds — where a region is restricted to team contests with cooperative goals, the agent cannot create an individual one.
- Only the asking user — joins and withdrawals act on the requester alone; the agent can't enroll or remove a colleague.
- Audit trail on every call — reads and writes alike log the requesting user, the tool, and the parameters.
WHAT TEAMS TRY INSTEAD
The four alternatives — and why none of them reach the frontline or score themselves
Every organization that runs workplace competitions has tried at least one of these. None combine self-scoring from verified activity, a chat surface the frontline already has on their phone, and eligibility policy enforced server-side on every path.
A spreadsheet leaderboard emailed weekly
Someone's Friday afternoon, rendered as a stale attachment
- Scoring comes off the verified activity ledger — shifts, inspections, courses — so there is no scorekeeper and no dispute about whose number is right
- Standings are current to five minutes, not current to whenever the spreadsheet owner last had a free hour
- Reaches people without email at their desk — the agent answers on a phone, in the app they already carry
Standalone gamification vendors (Bunchball, Ambition-style)
A separate sales-contest tool with its own logins and its own data island
- Scores from real workforce activity already in the platform — no CSV pipeline to keep a vendor's leaderboard fed
- No extra per-seat license for frontline staff who only ever check their rank twice a week
- Prizes settle into the same rewards-points ledger and badge wall as Recognition and Levels & Badges — one accounting story, not two
A custom leaderboard dashboard build
Two quarters of engineering, then forever maintenance
- Shipped already — nobody has to plumb per-capita team scoring, most-improved baselines, or first-to-target crossing detection
- Eligibility, creator tiers, delegated budgets, and works-council mode are enforced server-side on every path including chat, not re-implemented per surface
- Inherits new capability as the platform evolves — brackets, duels, programs, analytics — without a dashboard port
The manual fallback — a whiteboard in the break room
Honest, cheap, and invisible to everyone on the other shift
- Night shift, remote sites, and drivers see the same standings as the crew standing in front of the whiteboard
- Answers the fact a whiteboard never can: your rank, and the specific entrant you're two points behind
- Prediction picks get made before lock instead of dying on a form nobody walked over to fill in
PLATFORM ADVANTAGE
Contests Agent inherits everything the platform already runs
A gamification vendor or a DIY leaderboard has to build each of these. Contests Agent gets them for free.
The verified activity ledger
Metric contests score off the same ledger that powers Levels & Badges — shifts, inspections, course completions. The agent reads scores nobody had to enter.
Eligibility enforced server-side
Audience rules, manager exclusion, creator tiers, approval routing, and works-council mode apply on the chat path identically to the web app.
Prizes on the rewards ledger
Points, badges, or bragging rights — settled through Company Store and the badge wall, with delegated budgets reserved at approval. Never cash, never gift cards.
Translation in 100+ languages
A multilingual roster asks for their rank in their working language and gets the standings back the same way.
Frontline-ready by default
Same mobile app, no extra login, no extra license — the audience contests are designed for is the audience already holding the surface.
RubyLLM-grounded model tiering
A rank lookup runs on nano; a "what should we run next quarter" conversation routes up. Automatic per call.
INDUSTRY FIT
Industries where a contest agent moves the most weight
Contests Agent earns most where the workforce is deskless, shifts rotate, and the leaderboard would otherwise be invisible to the people it's meant to motivate.
Retail
Store-vs-store races scored per capita so a small format competes with a flagship, with associates checking rank on a phone between customers.
Manufacturing
Safety sprints scored straight off completed inspections, and works-council mode available where individual ranking isn't permitted.
Logistics / Transportation
Drivers and warehouse crews on opposite schedules see the same live standings — no break-room whiteboard only one shift walks past.
Healthcare
Training-completion and compliance blitzes scored from the learning ledger, with prizes that stay points and badges — never cash.
Hospitality
Shift-pickup challenges that fill open shifts, scored the moment someone grabs one from the scheduling ledger.
Financial Services
Culture and learning contests with compliance-safe prizing — prediction pools limited to badges and bragging rights, never points or cash.
WHY MANGOAPPS WINS
An embedded contest agent beats a gamification vendor, a horizontal chatbot, or a DIY leaderboard on every axis
The argument frontline staff, program owners, HR, and IT all share — and the one a standalone gamification tool structurally cannot answer.
Cheaper than the alternatives
No per-seat gamification license for a roster that checks rank twice a week, no dashboard build, no coordinator's Friday afternoon spent updating a spreadsheet.
More secure
Every write confirmation-gated. Eligibility, creator tier, approval queue, and works-council policy enforced server-side. Nothing leaves the tenant.
Easier to deploy
Already deployed if Contests is enabled. Turn the agent on and the roster can ask for their rank the same day.
Easier to use
"How am I doing?" in chat, on a phone, in the app they already have — instead of a dashboard most of the workforce will never open.
Easier to manage
Creator tiers, prize budgets, works-council mode, and audit logging live in the same admin console as every other app. One access model.
Easier to extend
New tools ship as agent capabilities against the same contest engine — duels, programs, seasons — with no vendor roadmap to wait on.
AI is actually better
A vendor copilot can read the leaderboard it built. Only Contests Agent can tell you your rank, the entrant you're chasing, which pick locks tonight, and then launch next month's safety sprint from a template — in one thread, under one policy engine.
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
MangoApps Customers
Organizations that trust MangoApps to run every workflow across their workforce.
Customer Case Studies
Customer Case Studies
Customer Case Studies
Video Case Studies
Webinars
Customer Case Studies
Frequently Asked Questions About Contests Agent
11 tools. Seven read-only: list_contests, get_contest_standings (top entrants plus your rank and who you're chasing), get_my_contests, get_my_picks, and three admin-only reads — get_pending_contest_approvals, get_prize_budgets, and get_contest_analytics. Four confirmation-gated writes: join_contest, withdraw_from_contest, make_prediction, and create_contest_from_template.
No — that's the point of the design. Every rule the web app enforces is enforced on the chat path server-side. join_contest honors audience rules and the exclusion that keeps managers out of ranking their own reports. create_contest_from_template honors creator tier, routes non-admin contests through the approval queue, and respects regional works-council policy. An employee cannot use chat to enter a contest they're barred from, and a manager cannot use chat to skip approval.
No to both. get_pending_contest_approvals only lists what's waiting — approving or rejecting happens on the approvals page. And organizers enter real-world prediction results in the app; the agent records picks and reads outcomes, it never decides them.
It cannot edit an existing contest (only create a scheduled one from a template), post manual scores or import a CSV, vote in a submission contest, or judge entries. Joining is for individual contests only — team contests enroll the whole team automatically, so there's nothing to join.
No. Prizes across the whole app are points, badges, or bragging rights — never cash, never gift cards. Prediction contests are tighter still: badges and bragging rights only, never points, so office pools stay compliant. The agent inherits both limits; it has no path to a prize type the app doesn't support.
Everyone, but scoped. Any member can check what's running, their own standing, and their own picks, and can join, withdraw, or make a pick for themselves. The approvals queue, prize budgets, and engagement analytics are admin-only — a non-admin asking doesn't get an admin answer.
Participation rate against your pre-agent baseline, share of standings checks coming through chat rather than the dashboard, pick-completion rate before lock, contest launch rate by managers, DAU lift vs a non-enrolled control, and withdrawal rate as the honest counter-metric.
Let's Talk
Since 2008, we've been building the employee platform for the frontline, earning the trust of 2 million+ users and an NPS of 78.
Why Choose Us?
- AI-Ready Platform: One intelligent place for every employee and workflow.
- Top Security: HITRUST, ISO & SOC 2 certified.
- Exceptional UX: Delightful on mobile and desktop.
- Proven Results: 98% customer retention rate.
Trusted by Legendary Companies:
Prefer to explore first? Ask AI about Contests AI Agent →