Scaling Webinars had capable tools, but acquisition, attribution, content, webinar delivery, onboarding and sales still operated through different instructions and handoffs. RECAST built the coordination layer that allowed those departments to share context without collapsing their roles.
Evidence at publication: eight named agents, 35 bindings and 29 active sessions; gateway, billing proxy, 3D operating view and secure tunnel verified running on 30 May 2026.
The company
Bryant Ward runs Scaling Webinars, a business built around webinar growth and client delivery. Its operating surface includes ads, registrations, show-up sequences, post-webinar nurture, client onboarding, sales coaching, scorecards, weekly reporting and creative production.
GoHighLevel held contacts and opportunities. Hyros held attribution. Google Workspace held documents and calendars. Slack and Discord carried instructions and team communication. Those systems were individually useful, but Bryant was still the place where their context came together.
The constraint
One generic assistant could not safely handle every operating mode. Ad strategy, VSL scripts, client onboarding, sales review and webinar nurture have different inputs, standards and approval rules. Splitting them into disconnected bots would create the opposite problem: specialist lanes with no shared memory.
The business needed department-level separation without returning coordination to Bryant.
Why existing tools were not enough
The automation stack could move records and trigger actions, but it did not carry operating judgment. A contact in GoHighLevel did not automatically bring the relevant client knowledge, attribution context, creative standards and approval policy into a task.
The missing layer had to route each request to the correct specialist, make shared context retrievable, expose the work in a visible command surface and stop at the right boundary. Spend changes, live sends, publishing, hiring or firing, and production CRM changes remained approval-gated by design.
Transformation map
Before
Ads | Hyros | GHL | Workspace | Slack | Discord -> Bryant reconciles context -> department acts
After
Shared operating context -> specialist agent -> approved tool path -> visible output -> Bryant/team gate when consequential -> system record
The eight lanes are separated by responsibility, not isolated by memory.
One workflow, end to end
- Input: A new client or campaign requirement enters the command surface with the relevant business context.
- Routing: Agent Jeff directs the work to the appropriate specialist - ads, content, webinars, sales, onboarding, ad/VSL script or creative pipeline.
- Context retrieval: The specialist uses shared gbrain knowledge and the appropriate client data rather than asking Bryant to restate the account.
- Production: The agent prepares the requested asset, analysis, sequence, scorecard or operational next step through its allowlisted tools.
- Human gate: Spend, production CRM changes, live sends, publishing and people decisions wait for explicit approval.
- Visible output: The result appears in the team's Discord/Slack channels or hosted Command Center rather than disappearing into a private model session.
- Record: Client pages, GHL panels and the system's active sessions preserve work state for the next department.
What changed
- Built: Eight departmental agents, a hosted Command Center, a 3D operating view, shared memory and specialist workflow skills.
- Connected: Discord, Slack, GoHighLevel, Hyros, Google Workspace, Zapier, Apify and governed Composio routes.
- Tested: Gateway health, model route, Discord and Slack status, local services, secure tunnel and server-side WebSocket proxy were inspected live.
- Verified live: On 30 May 2026, the OpenClaw gateway, billing proxy, Claw3D service and Cloudflare tunnel were running. The environment held 35 bindings and 29 active sessions.
- Actively used: The retained evidence shows active sessions and a live operating environment. It does not convert activity into an invented time-saving or revenue figure.
- Provider or client dependency remaining: Consequential actions intentionally remain subject to approval; that is a control boundary, not unfinished automation.
- Financial result reported by client: None retained for this case.
Proof
- System scope: Eight named specialist agents.
- Operational routing: 35 bindings at the verification point.
- Use signal: 29 active sessions at the verification point.
- Runtime receipt: Gateway, billing proxy, secure tunnel, Discord, Slack and Claw3D reported healthy or active on 30 May 2026.
- Control receipt: Approval gates explicitly cover spend, live sends, publishing, people decisions and production CRM changes.
Technology - revealed last
The system runs on Bryant's Mac mini with OpenClaw, Discord, Slack, GoHighLevel, Hyros, Zapier, Google Workspace, Apify, gbrain/memory-core, Composio, a hosted Command Center, Cloudflare Tunnel and Claw3D. The live environment exposes clients, agents, brain, skills and settings, while a server-side WebSocket proxy connects the visual office to the OpenClaw gateway. The interface makes the operating layer legible to the team; it is not the source of the automation itself.
Client quote
No source-backed verbatim client quote is published for this case. The public proof is the verified deployment scope and runtime evidence; RECAST does not write testimonial language on a client's behalf.
Next transformation
See how Ramon's multi-brand operating layer shares reusable knowledge without collapsing brand boundaries. It applies the same orchestration principle to stores rather than webinar departments.
Get your AI blueprint
If your tools are capable but every department still waits for one person to connect the work, the constraint is orchestration. Book a free AI audit to identify the first operating lane worth proving.
Source record
OPENCLAW-DEEP-CLIENT-CASE-STUDIES.pdf, page 12.RECAST_CASE_STUDIES_2026_UPDATED.pdf, page 3, documentRA-CASES-001, version 2.0.RECAST_WEBSITE_MASTER_BLUEPRINT.md, sections 10.4, 11 and 58.