Agency autopilot: running 3+ brands without diluting voice
Workspace-per-brand architecture, isolated credit pools, per-brand Persona Briefs, and network-level reporting. The multi-brand autopilot system that keeps voice tight across clients.
The direct answer
Agency autopilot works at 3-5 brands per Agency tier when each brand has its own workspace with isolated Persona Brief, asset library, credit allocation, and posting cadence. The architecture: one credit line for the agency, separate workspaces per client to prevent voice cross-contamination, and network-level reporting that rolls up per-client metrics. Going past 5 brands typically requires Enterprise credit pooling.
Agencies have a structural disadvantage with AI content: each client needs a different voice, but most AI tools assume one persona per account. Running 5 clients on 5 separate ChatGPT instances is unmanageable; running them all on one account dilutes voice across clients.
The right architecture is workspace-per-brand with shared infrastructure. This post covers how to set it up and the failure modes to avoid.
The agency multi-brand problem
Three structural issues with AI content for agencies:
Voice dilution. Running multiple clients through one workspace causes their Persona Briefs to bleed into each other. Output for Client A starts sounding like Client B.
Credit allocation. If one client uses 80% of monthly credits, the other clients suffer. Need isolated credit pools.
One Kompozy Agency-tier subscription for the agency (the legal entity).
One workspace per client brand (e.g. 4 workspaces for 4 clients).
Each workspace has its own: Persona Brief, asset library, sources, scheduling cadences, banned-word list, output buckets.
Credit pool is shared at the agency level (so unused credits from one client can flow to another).
Network-level reporting rolls up per-workspace metrics to the agency owner dashboard.
This is how Agency tier is engineered. Multi-workspace is the architectural unit, not a feature on top of single-workspace.
Persona Brief per workspace
Each client gets a dedicated Persona Brief in their workspace. The agency does the discovery work upfront:
60-minute intake call with the client. Discover voice, audience, banned phrases, reference posts.
Agency writes Persona Brief draft based on intake.
Client reviews and approves before any autopilot generation begins.
Run 14-day manual review ramp per client (parallelize across clients to speed up).
Flip autopilot on per-client only after the brief is validated.
Common mistake: skipping the per-client discovery and copy-pasting Persona Briefs across clients. Result: every client sounds the same. Audiences notice within 2 weeks.
Credit allocation strategy
Agency tier ships with 55,000 credits/month, shared across workspaces. Allocation patterns:
Pro-rata by retainer: each client gets credits proportional to their monthly retainer.
Even split: divide credits evenly across active clients.
Dynamic by demand: track each client's typical credit burn and allocate based on need.
Pro-rata is the simplest. Even split prevents one client from running others dry. Dynamic requires monitoring infrastructure.
Hard cap per client (default: 20,000 credits/month per workspace) prevents runaway burn from one client starving the others. Override on a per-month basis if a client is launching or running a campaign.
Reporting and dashboards
Per-client metrics agencies need:
Outputs shipped this month (broken down by bucket: video, image, text, blog, newsletter)
Credit spend this month + projected month-end spend
Engagement rate per platform (rolled up across all posts)
Quality-gate failure rate (proxy for Persona Brief health)
Approval-untouched rate (proxy for autopilot maturity)
Top-performing post + top-failing post (for client review meetings)
Agency dashboard rolls these up across all client workspaces in one view.
Client onboarding workflow
Standard 30-day client onboarding for agencies running Kompozy autopilot:
Week 1: Intake call, Persona Brief draft, source identification, sample-output generation for client approval.
Week 2: Manual review every output, edit aggressively, refine brief from edits.
Week 3: Approval rate climbs. Add platforms one at a time as voice stabilizes.
Week 4: Flip autopilot on for the most stable source. Spot-check daily.
Day 30+: Monthly client review call with metrics. Refine brief based on what worked.
Common agency failures
Reusing Persona Briefs across clients. Sounds the same. Audiences notice.
Flipping autopilot on day 1. Skips the ramp. Outputs sound generic.
Not surfacing AI use to clients. Eventually a client asks "is this AI?" — be transparent upfront.
Overcharging. Agencies sometimes bill $5k/month for AI-assisted content. Margin pressure builds when clients realize the AI cost.
Under-charging. Some agencies bill at AI-cost-plus-20%, eroding their own margin. Pricing should reflect strategy + voice work + monthly management, not just compute cost.
Pricing the service
Agency pricing patterns for AI-managed content in 2026:
Setup fee: $1,500-3,500 per client for Persona Brief discovery + ramp + onboarding.
Monthly retainer: $1,500-4,000 per client for 80-150 posts/month + monthly strategy review.
Performance bonus: 5-10% of attributed pipeline / revenue for B2B clients (optional).
The margin equation: ~$200 of credit cost per client per month → ~$1,200-3,500 of monthly profit per client. Agency tier ($399/month) covers ~5 clients comfortably.
Frequently asked questions
How many clients can one Agency tier handle?
Comfortably: 3-5 clients depending on output volume per client. Heavy-video clients burn credits faster; mostly-text clients stretch further. Past 5 clients, Enterprise credit pooling is the right architecture.
Can multiple agency team members work in the same client workspace?
Yes. Each workspace supports multiple seats. The agency owner has access to all workspaces; individual team members can be scoped to specific client workspaces.
How do we handle client churn cleanly?
Each workspace can be exported as JSON (Persona Brief, asset library, generated content history). Client takes their workspace with them or archives it. Credit allocation reverts to the agency pool.
What if a client wants to bring AI in-house after we set it up?
They can. Transfer the workspace to their own subscription. Agency loses the retainer but keeps the relationship. Most agencies position themselves as the strategy + brief + ramp layer, not the production layer — which is harder for clients to replicate alone.
How do we handle clients in regulated industries?
Manual review only for those clients. The Agency tier supports manual-review workflows alongside autopilot. Compliance review chain is preserved regardless of the AI generation layer.
AI Content Repurposing — The complete methodology for turning one source into 25-35 pieces of native-format content across every platform — without producing AI slop.
AI Brand Voice & Persona — Without a Persona Brief, every AI output averages to the LLM default voice. This is the 5-section methodology that makes 100+ AI-generated posts feel like one human author wrote them.