Skip to content
ONE3.ai

One3 Ecosystem · Module

Human Expert Cloud

Agents and users escalate hard cases to vetted experts inside your workspace

Planned · Year 3 (W4)Sold as: MeteredC-169
Ask about Human Expert Cloud

Human Expert Cloud will turn agent and user dead-ends into completed, accountable work. Scoped tasks go to verified experts such as chartered accountants, lawyers and designers, with a redacted context bundle, a deliverable schema and an SLA.

Human Expert Cloud is planned for W4 as a closed beta, with GA in W5. It is sold per task: the tenant accepts a quote and a platform fee applies per task.

Why it exists

When a Digital Employee has low confidence, a policy requires professional judgement, or a tenant rule demands a licensed human, the work usually leaves the system. Someone emails a CA, re-explains the context and the audit trail is lost. Human Expert Cloud keeps that work inside the tenant's workspace.

How an escalation works

  1. An agent run (or a user, from a ticket, invoice, contract or Copilot conversation) creates a task with a task type, a context bundle draft, a deliverable schema, an SLA tier and a budget. The agent run pauses.
  2. The redaction policy masks sensitive personal data by default and conflict-of-interest rules filter experts.
  3. A tenant approver reviews exactly what the expert will see. Sharing a bundle with personal data always needs human approval; agents cannot approve their own bundles.
  4. The task goes to the tenant's preferred expert pool or to matched experts, who accept a fixed price or send a quote.
  5. The expert works through a time-boxed guest access grant: downloads off by default, rendered documents watermarked, every session audited into the tenant's log.
  6. The deliverable is validated against its schema, then the tenant accepts it, asks for a revision or disputes it.
  7. On acceptance, the deliverable resumes the paused agent run as typed output. Any Actions taken from it still follow normal risk-tier approvals.

Experts

Initial categories include chartered accountants and tax, lawyers, company secretaries, designers and video editors, ads and SEO specialists, developers, HR and payroll specialists and translators. Experts are verified with KYC, licence checks against registries where access is permitted (otherwise document review), test tasks and portfolio review. Experts cannot run high-risk Actions in your tenant; they deliver documents and structured outputs.

Profession-specific presentation rules, liability statements and payouts with tax withheld where applicable are designed in, and each needs confirmation with counsel (verify).

Features

What Human Expert Cloud includes

Expert application and verification

Application by category with KYC, licence and membership checks where registry access is permitted, and periodic re-verification.

Test tasks and portfolio review

Category-specific test tasks scored by senior reviewers, plus portfolio review for creative work.

Conflict-of-interest controls

Experts declare employers, key clients and blocked organisations; matching excludes conflicts and the expert confirms on each task.

Task type catalogue

Task types with a deliverable schema, required inputs, default SLA tiers and allowed expert categories, such as GST notice response review.

Agent escalation Action

Agent skills call a typed escalation Action when confidence is low or policy requires an expert; the originating run pauses.

Ask an expert from any record

Create a task from a ticket, invoice, contract, campaign, workflow run or Copilot conversation.

Context bundle builder

Select records, fields and files, apply redaction by classification, and preview exactly what the expert will see.

Budget, SLA tier and deadline

Standard, priority and urgent tiers with declared response and delivery windows and a budget cap.

Preferred expert pools

Invite your own CA, lawyer or designer, who must still pass verification; routing prefers your pool.

Quotes and approvals

Fixed price per task type or a capped quote, approved in the Approvals Inbox or on WhatsApp with step-up above a threshold.

Expert workspace

Context bundle, clarification thread, deliverable editor, time-boxed guest access, downloads off by default and watermarking.

Deliverable validation and acceptance

Schema validation and an agent pre-check, then accept, request revision or dispute within an acceptance window.

Resume originating run

The accepted deliverable returns to the paused agent run or workflow as typed output.

Expert score and outcome signals

Score from acceptance, on-time rate, outcome quality and ratings minus disputes, shown only after a minimum sample.

Workflow

Agent escalation to an expert

  1. 1

    Agent run creates a task with type, bundle draft, schema, SLA and budget

  2. 2

    Redaction policy and conflict-of-interest filter are applied

  3. 3

    Tenant approver approves the bundle preview and routing

  4. 4

    Expert accepts the fixed price or sends a quote for approval

  5. 5

    A time-boxed guest grant opens the bundle to the expert

  6. 6

    Expert asks clarifications and submits the deliverable

  7. 7

    Schema validation and agent pre-check run before tenant review

  8. 8

    On acceptance the agent run resumes, the earnings entry is recorded and the grant is revoked

Workflow

SLA breach and reassignment

  1. 1

    SLA monitoring flags an at-risk task

  2. 2

    Tenant chooses to extend, reassign or cancel

  3. 3

    Reassignment returns the task to matching with tenant approval

  4. 4

    Partial compensation rules apply and the expert score updates

AI

AI skills and Digital Employees

Escalation decision

Recognises inside an agent run that a case needs a licensed human and raises the ask; the task itself is approval-gated.

Task drafting from a record

Turns a ticket, notice or contract into a typed task with a deliverable schema.

Context-bundle redaction proposal

Proposes which fields to withhold or mask so the bundle stays minimal.

Expert matching and explanation

Ranks experts by category, credentials, outcomes, SLA history, conflicts and language, and explains the ranking.

Deliverable completeness check

Validates the deliverable against its schema and flags gaps such as a missing rule citation.

SLA risk prediction

Predicts at-risk tasks and proposes reassignment.

Digital Employees that use this module

They act through the same governed Actions as people; risky steps wait for approval.

Frequently asked questions

Who are the experts?

Initial categories include chartered accountants, lawyers, company secretaries, designers, ads and SEO specialists, developers, HR and payroll specialists and translators, each verified before taking tasks.

What data does an expert see?

Only the approved context bundle, with sensitive personal data masked by default, through a time-boxed grant that closes when the task closes. Every session is audited.

Can an expert change records in my tenant?

No. Experts cannot run high-risk Actions. They deliver documents and structured outputs, and your users or agents act on them under normal approvals.

Can I use my own CA or lawyer?

Yes. You can invite them to your preferred expert pool. They still pass verification, and routing prefers them.

When is it planned?

A closed beta is planned for W4 (Year 3) and GA for W5.

Previous moduleOne3 Academy

Next: Freelancer & Gig Management

Contracts, timesheets, invoices and TDS for your freelancers in one record

Explore

See Human Expert Cloud in your workflow

Tell us about your team and the outcomes you want. We will walk you through the apps, the Digital Employees and the plan for your industry.

Request early access How One3 works