Zain's system advances deterministic work and stops when a payout rule, ownership decision or approval is missing. Every pause returns a precise request for the human decision needed to continue.
Evidence at publication: Maya Brunometi was verified as the active primary agent on 30 May 2026, with the Zain dashboard, ClawPort, Ollama and proxy services active. The workflow packet includes audits, handoff manifests and client action packets; operation remains supervised at consequential gates.
The company
Zain runs a high-volume agency and offer operation across client follow-up, content production, payout logic, closer and setter mapping, and several business workstreams. The operating surface includes Discord, Slack, ClickUp, Todoist, scraping, Google/service accounts and a persistent knowledge layer.
That mix produces more than task volume. It produces dependency: a content asset may be complete while its approval is not; a commission message may be drafted while the buyer-to-closer rule remains unknown; a client handoff may need information that exists only with Zain or Yasmine.
The constraint
Work could begin in one tool, wait for approval in another and require a human to assemble the final action packet. The tools were individually capable, but they did not behave like one operations department.
Blind automation would be worse than the original fragmentation. Moving a task or preparing a message can be deterministic. Deciding an unresolved payout rule cannot. The system needed to distinguish execution from judgment and expose the blocker instead of silently guessing.
Why existing tools were not enough
ClickUp and Todoist could store tasks. Discord and Slack could carry communication. Apify could collect data. Content systems could produce assets. None of those surfaces alone maintained the operating contract between them.
The missing layer had to retrieve business context, run repeatable steps, audit what actually completed, package remaining actions for a client or operator and enforce stop conditions around data and approvals.
Transformation map
Before
Tasks | chats | content | payout rules -> manual reconciliation -> client action packet -> follow-up
After - supervised operation
Workflow trigger -> Maya retrieves context -> deterministic steps execute -> audit -> blocker or approval gate -> action packet -> human decision
The audit is part of the workflow, not a retrospective check added after an agent says it is done.
One workflow, end to end
- Input: A client workflow is triggered with its available task, content and account context.
- Context: Maya uses the persistent knowledge layer and the corresponding ClickUp/Todoist records to establish current state.
- Execution: Deterministic actions - such as organizing a handoff, preparing a ready-to-post asset, scraping permitted data or moving a known task state - proceed through the configured route.
- Audit: Goal and credential-mode verifiers check whether the intended step completed and whether the required provider state exists.
- Stop condition: If a draft needs approval, or a payout/closer-setter mapping is unresolved, the system pauses rather than filling the gap with an assumption.
- Output: Maya generates a paste-ready message, commission client note or structured client action packet identifying the exact human decision required.
- Record: The handoff manifest and cadence checks preserve the blocker and next action for the following run.
What changed
- Built: Maya CSM workflows, client action packets, commission-message preparation, goal audits, handoff manifests, credential verifiers, cadence checks and paste-only outputs.
- Connected: OpenClaw, gbrain, local Ollama embeddings, ClickUp/Todoist automation, Apify, content generation, Discord, Slack and the agency dashboard environment.
- Tested: Deterministic workflow steps, blocker audits, credential modes, cron cadence and the handoff behavior around missing input were encoded and checked.
- Verified: On 30 May 2026, Maya was the active primary agent and the dashboard, ClawPort, Ollama and proxy services were active.
- Actively used: The retained record shows active operating services and a supervised primary agent. It does not provide a quantified volume of completed client workflows.
- Remaining dependencies: Approval-gated drafts, product/team payout rules and buyer/product closer-setter mappings still require Zain or Yasmine where the source is incomplete.
- Financial result reported by client: None retained.
Proof
- Runtime receipt: Active primary agent and core operating services observed on 30 May 2026.
- Control receipt: The workflow explicitly pauses for approvals and client-side data rather than treating uncertainty as permission.
- Audit receipt: Goal audits, handoff manifests, credential-mode verifiers and cadence checks are part of the packet.
- Output receipt: Client action packets and paste-only messages make remaining human work concrete.
- Claim boundary: The source supports supervised operation, not unsupervised autonomy or a quantified revenue result.
Technology - revealed last
The stack runs OpenClaw 2026.5.2 on a Hostinger VPS with Discord, Slack, gbrain, local Ollama embeddings, a Command Center dashboard, hosted 3D office, Caddy, a billing proxy, OCW bridge service, ClawPort, agency-os-zain, ClickUp/Todoist automation, Apify, content generation, Maya CSM, scheduled jobs and workflow-audit scripts.
Client quote
No source-backed verbatim client quote is published for this case. The control architecture and verified runtime are the evidence; no testimonial is reconstructed from internal project language.
Next transformation
See how ConvertSail encoded provider authority across a six-agent delivery system. Zain's case focuses on stop conditions; ConvertSail focuses on selecting the right source before work begins.
Get your AI blueprint
If your operation needs automation but the costly decisions remain ambiguous, the design target is guarded autonomy. Book a free AI audit to separate deterministic work from the decisions only your team should make.
Source record
OPENCLAW-DEEP-CLIENT-CASE-STUDIES.pdf, page 15.- Zain does not appear in
RECAST_CASE_STUDIES_2026_UPDATED.pdf; the deep case document is the retained case source. RECAST_WEBSITE_MASTER_BLUEPRINT.md, sections 10.4, 11 and 58.