Source-risk report — JulyNeeds review
3 source gaps are affecting answer quality. Highest risk: stale or permission-limited knowledge being treated as complete context.
4 citationsproof log attachedmock_local
Workers
Practical AI workers. Not an agent swarm.
Shared Worker Core preview: five focused workers use the company brain as context, produce business outputs, and leave a proof trail behind every run. Controlled, useful, and boring in the best way.
Worker 01 · Knowledge
Keeps the brain honest.
Review source health and recommend the next knowledge-cleanup actions before trusting automated answers.
- Flags stale and conflicting documents before they mislead an answer
- Maps knowledge gaps against the questions your team actually asks
- Every recommendation cites the sources that triggered it
completed4 sources checked3 gaps surfaced
Output · saved to Reports
Local cited brief: prove trustworthy workers before autonomyCited
Research Worker used approved local/demo sources to produce a cited brief. It did not browse the web, call Onyx, sync connectors, use customer data, schedule work, or take external actions.
2 citations4 limitationslocal brief
Worker 02 · Research
Turns hours of digging into a cited brief.
Works through approved local company context and returns a structured brief — claims, evidence, limitations, and open questions clearly separated.
- Every finding carries a citation and source quote
- Separates verified facts from inferred or unknown items
- Briefs save as reusable reports with proof attached
Worker 03 · Watchdog
Watches what you can’t.
Raises calm, sourced alerts when source health changes. No noise, no drama, and no real monitoring daemon in this mock/local slice.
- Alerts show what changed and which source triggered it
- Weekly watchlists roll up into a saved report
- Scope stays fixed; it never expands autonomy on its own
Worker 04 · Support
Drafts replies. Humans hit send.
Grounds a support draft in policy and source context, then routes the idea for review. Nothing reaches a customer without approval.
- Drafts cite the policy behind each claim
- Flags stale or ambiguous source material
- Sending remains blocked behind human review
Worker 05 · Sales / Ops
Briefs your team before the moment that matters.
Assembles account and ops briefs from approved context, with confidence, limitations, and approval boundaries visible before anyone acts.
- Briefs pull from contracts, threads, tickets, and reports
- States what it could not verify
- Customer, money, or record-touching recommendations route to approvals first
app.innexios.com/workers/run
Worker run console · mock/local
Proof trail before autonomy.
These runs come from the existing Innexios server worker services. They recommend, summarize, and alert only — no live Onyx call, customer data, email, export, billing, scheduling, connector sync, or external action is enabled.
Mode
mock_local
Alerts
3
Findings
3
Source-risk report — July
3 source gaps are affecting answer quality. Highest risk: stale or permission-limited knowledge being treated as complete context.
4 citationsproof log attachedmock_local
Local cited brief: prove trustworthy workers before autonomy
Research Worker used approved local/demo sources to produce a cited brief. It did not browse the web, call Onyx, sync connectors, use customer data, schedule work, or take external actions.
2 citations4 limitationslocal brief
Local watch report: source-health risks before automation
Watchdog Worker reviewed local/demo Source Health state and flagged stale, planned, and permission-limited sources. It did not schedule monitoring, call live services, sync connectors, browse the web, or take external actions.
3 alerts3 checksno scheduler
Support draft #4821 — refund request
Draft grants a prorated refund per the annual-plan policy, cites the exact clause, and notes the policy doc is stale — reviewer prompted before sending.
3 citationsrouted to: support leadno email sent
Account brief: Meridian — renewal call
Contract terms, open tickets, decision threads, and unresolved commitments assembled into one brief with source references attached.
14 citationsconfidence 89%saved to reports
Mock/local boundary
Worker run console is visible, but not autonomous.
Sources: mock_local_knowledge_manager_worker, mock_local_research_worker, mock_local_watchdog_worker. External actions taken: false. No live Onyx API call, no customer data, no email, no export, no billing, no scheduler, no connector sync, and no external action are enabled here.
Support Draft and Sales/Ops are visual/product states in this slice; live execution, email, account updates, exports, schedules, and billing remain blocked.