Platform user guide

Run your community's AI pod. Keep what your community earns.

Actionboard is a platform-as-a-service for community-run AI labs. Host it anywhere, train and evaluate action models in a closed loop with a verified human expert network — and route 100% of NGO earnings back to your community.

Set up a community pod → How the platform works
✓ Host anywhere — cloud, on-prem, or local provider ✓ Closed-loop evaluation, not open contribution ✓ 100% of NGO earnings stay local
01 · Platform overview

One platform, run by your community

An Actionboard pod is a self-contained deployment of the platform, operated by a local community or business. Each pod connects three layers: the platform itself, the people doing verified work, and the local providers who handle hosting and payouts.

Pod architecture
Actionboard Platform
Action-model training & evaluation engine, task queues, quality gates, pod admin console.
⇄
Community Pod
Verified expert network, pod operators, contributor onboarding, local governance.
⇄
Local Providers
Hosting (cloud / on-prem), payment rails, identity verification, physical AI facilities.
↓ every pod reports into one shared evaluation standard ↓
Closed-loop evaluation — the trust layer every pod shares
Pod operators
NGOs and local orgs who run the pod and govern membership.
Businesses
Host a pod to source verified training and evaluation work.
Expert contributors
Verified members who do paid annotation, evaluation, and physical AI work.
Developers
Integrate the pod's APIs with local systems and providers.
02 · Getting started

Set up your community pod

A pod goes from application to first paid task in four steps. Your Actionboard onboarding partner walks you through each one.

1 Choose your community type

Learn & Earn, Government, Research & Development, NGO Community, or Sponsored Community. The type sets your pod's default governance, payout, and evaluation templates.

2 Pick where you host

Actionboard is platform-as-a-service, hostable anywhere: a cloud region near you, on-premises hardware, or an integrated local provider. Data residency stays under your pod's control.

3 Connect your local provider

Link a local payment provider so contributors are paid in their own currency, and register your identity-verification partner for member vetting.

4 Onboard members & open the queue

Invite community members, run them through verification and calibration tasks, then open your pod's task queue. First payouts typically land within the first cycle.

03 · Closed-loop evaluation

Every contribution is verified before it counts

Unlike open contribution systems, nothing enters an Actionboard model unverified. Work flows through a closed loop: task → contribution → expert review → model evaluation → feedback. The loop only closes when quality gates pass.

The evaluation loop
STEP 1
Task issued
Pod queue assigns work to verified members only.
→
STEP 2
Contribution
Annotation, demonstration, or physical AI data is submitted.
→
STEP 3
Expert review
Network experts cross-check against calibration standards.
→
STEP 4
Model evaluation
Action models are retrained and scored on held-out tasks.
→
STEP 5
Feedback & payout
Scores flow back to contributors; verified work is paid.
↺ evaluation results re-calibrate the next round of tasks — the loop is closed
Quality gates
Contributions below the pod's agreement threshold are returned with reviewer notes, never silently merged.
Calibration tasks
Every member periodically completes known-answer tasks that keep reviewer standards aligned across the pod.
Auditable trail
Every accepted datum carries who made it, who reviewed it, and which evaluation run confirmed it.
04 · Expert network

Joining as a contributor

Contributors are real, verified members of a local pod — not anonymous accounts. Onboarding takes most members under a week.

1
Verify your identity
Through your pod's local identity provider. One person, one account — this is what keeps the network trustworthy.
2
Declare your expertise
Languages, domains, trades, and physical skills. Pods match tasks to declared and demonstrated expertise, not to whoever clicks fastest.
3
Pass calibration
A short set of known-answer tasks establishes your starting quality score and unlocks the paid queue.
4
Grow into expert reviewer
Sustained high agreement promotes you to reviewer — higher-paid work checking others' contributions and setting pod standards.
05 · Earnings & payouts

100% of NGO earnings stay in the community

Actionboard charges for the platform, not a cut of your community's work. Earnings from action-model training and physical AI work flow through your pod's local payment provider directly to the NGO and its contributors.

How money flows
Client pays for verified work into the pod escrow
Evaluation loop closes quality gates confirm the batch
Local provider pays out local currency, local rails
100% to NGO & contributors no platform commission
What the pod decides
  • The split between contributor pay and NGO programs
  • Payout schedule (per-task, weekly, or monthly)
  • Reviewer pay premiums
  • Which local payment provider handles disbursement
All splits are set in the pod admin console and shown transparently to every member.
06 · Physical AI training

Paid, hands-on work for physical AI

Beyond screen work, pods can run physical AI programs: recording real-world demonstrations, teleoperating robots, and validating embodied model behavior — coordinated through the same task queue and paid through the same local provider.

Demonstration capture
Members record everyday manual tasks — cooking, sorting, assembly — following pod capture protocols. Sessions upload directly into the evaluation loop.
Teleoperation shifts
Scheduled shifts operating partner robots remotely or at local facilities. Requires the physical-AI calibration track and safety briefing.
Embodied evaluation
Expert reviewers score model behavior against real-world outcomes — the physical half of the closed loop.
Safety note: physical AI tasks are only available at pods with a registered local facility provider and completed safety onboarding. Your pod admin controls who is eligible.
07 · Local provider integrations

Plug in the providers your community already uses

Each pod integrates four provider categories. Developers connect them through the pod admin console; each integration is scoped to your pod only.

💳 Payments
Local banking, mobile money, or payroll providers for contributor payouts in local currency.
🪪 Identity
Regional identity-verification services used during member onboarding — one person, one verified account.
🖥 Hosting
Cloud regions, national data centers, or on-prem hardware. The platform ships as containers your host runs anywhere.
🏭 Facilities
Local spaces and equipment partners for physical AI capture and teleoperation programs.
08 · Security & trust

Closed community, verifiable work

Open contribution systems accept work from anyone — which means unverified contributors, unreliable quality, and supply-chain risk. Actionboard pods take the opposite approach.

Open contribution
  • Anonymous, unverified contributors
  • Quality discovered after merge
  • No accountability trail
  • Security depends on volunteer review
Actionboard pod
  • Identity-verified community members
  • Quality gated before acceptance
  • Full audit trail on every datum
  • Paid expert review, every time
Your data stays in your pod
Pods are isolated deployments. Client data and community data never leave the hosting your pod chose.
Role-based access
Operators, reviewers, contributors, and developers each see only what their role needs.
Signed evaluation reports
Clients receive evaluation reports they can verify — proof the closed loop actually ran.
09 · Admin & pod management

Running the pod day to day

The pod admin console is the operator's home. Everything a pod operator manages lives in five tabs:

Members Invitations, verification status, expertise profiles, calibration scores, and role promotions.
Tasks Incoming client work, queue priorities, matching rules, and deadline health.
Quality Loop dashboards: agreement rates, gate pass rates, evaluation scores per batch, and reviewer workload.
Earnings Escrow balance, payout runs, contributor/NGO splits, and the transparent ledger members see.
Integrations Provider connections, hosting health, API keys for developers, and webhook logs.
10 · FAQ

Common questions

Who owns the models a pod helps train?
Client engagements define this per contract. The pod's evaluation records and community data always remain the pod's property, hosted where the pod chose.
How does Actionboard make money if NGOs keep 100%?
Actionboard charges a platform-as-a-service subscription to the pod host or sponsoring business — never a commission on contributor or NGO earnings.
Can one organization run multiple pods?
Yes. Each pod is a separate deployment with its own members, providers, and ledger. A regional NGO often runs one pod per community.
What happens if my work fails a quality gate?
It's returned with reviewer notes and doesn't count against you immediately. Repeated low agreement lowers your queue priority; calibration tasks are the way to recover.
Do contributors need special hardware?
Screen work needs only a browser. Physical AI tracks use equipment provided by the pod's local facility partner.
How fast are payouts?
As soon as the batch's evaluation loop closes and the pod's payout schedule triggers — per-task, weekly, or monthly, set by your pod operator.