Synergy · Architecture

Brief us once. Same brand voice on every surface — marketing here, AI workspace next door.

You don't want one supplier who knows your strategy and another who has to be re-briefed every week. The marketing side and the AI workspace share the canonical brand voice, the same niche data, the same regulator rules — so your brand stays in family across both surfaces.

Each side feeds the other. Insight from your marketing side strengthens the AI workspace. How the workspace gets used tells the marketing side what's working. Every signal between them is checked. Every crossing waits for your approval.

Start with a free profile → See what's shared → Visit the AI workspace →

Why this matters:

You've worked with two suppliers — one for the marketing, one for the AI tools — and the brand felt like two different businesses. The data on one side never reached the other. The voice drifted between them. You paid twice and got half.

What you land on:

A brand that gets remembered.
Know what works. Do more of it.
Cheaper than a hire. Deeper than an agency.

What's shared.

Seven libraries are shared in code. Both products read from the same source — a new vendor name added here gets added on the AI side; a regulator-rule update here carries across; a new niche flows automatically.

Source catalogue

43 data sources (free + subscriber + tenant-OAuth). Same canonical list on both products; a new source lands once and benefits every client on either side.

Niche catalogue

68 niches with their regulator flags + benchmark surfaces. Same niche slug carries the same compliance rules + channel priors on both sides.

Channel priors

30 niche × channel importance entries (CRITICAL / HIGH / MEDIUM / LOW / SKIP). Marketing uses them for recommendation; AI side uses them for adjacent-feature suggestions.

Quant canon

29-entry reading-list registry across core / adjacent / practitioner. Both products cite the same canon refs on every typed output, methodology is the family bond.

Vendor registry + regulator policies

Same 32-entry vendor name guard + same FCA / ASA / GDC / MHRA / INFLUENCER policies. A vendor leaked on one product would leak on the other, so we share the registry; it's physically impossible to drift.

Brand archetype taxonomy

Jung's 12 archetypes (Hero / Sage / Magician / etc.) shared. Same brand classified into the same archetype family on both products → consistent voice across surfaces.

What stays separate.

The boundary is enforced in code, not a written policy. The shared package literally can't carry tenant data — it's built that way from the start. Beyond that:

Tenant data (ICP, brand voice, VOC corpora, strategy artefacts)
Credentials, tokens, secrets, NEVER cross the boundary
Billing (separate Stripe per CLAUDE.md)
Deployment boundaries (separate AWS accounts per ADR-0181)
Per-tenant graph / relational / document / vector storage (per ADR-0257)

How each side improves the other.

Synergy is a feedback loop, not a parallel. Each signal flowing between the two sides is a typed message through a bridge. Every crossing runs through the safety checks and waits for human approval before the receiving side acts on it.

Brilliant Marketing improves Brilliant AI

Awareness + acquisition. Marketing campaigns funnel prospects who'd benefit from an AI workspace.
VOC → AI feature priorities. Marketing-side VOC corpora surface 'what would help here' themes; AI side gets the typed signal under HITL approval.
Brand cohesion. Shared brand archetype + voice from atelier/shared/archetypes means AI side ships in the same family voice as the marketing side.
Engagement-tier signal. When a marketing client upgrades tier, the AI side receives a typed workspace-tier upsell signal.
Case-study credibility. Marketing-generated case studies cite AI workspace use (with consent + HITL approval). The AI side gets credibility evidence flowing back.
MMM applied to the AI side's own GTM. The marketing-side MMM machinery (Tier 3-4 quant) can be applied to the AI side's own marketing spend.
BrandScript → AI workspace voice. When Sten refreshes a tenant's BrandScript on the marketing side, a typed signal flows to the AI workspace so its content suggestions stay in family with the current voice (no re-briefing).
Pain → Bridge → Desire scaffold → AI suggestions. Sten's PainDesireBridge for the tenant becomes a typed grounding signal — the AI workspace anchors suggestions to the same villain shape and desire library the marketing side ships against.
Lore audit → AI workspace voice guard. When Lore finds high-severity drift in the tenant's existing copy, the workspace knows which patterns to avoid suggesting (so it doesn't repeat the off-brand sentence the audit just flagged).

Brilliant AI improves Brilliant Marketing

Workspace usage → marketing strategy. When an AI tenant hits a capability ceiling (e.g. needs a campaign), marketing receives the signal and Aria-Copilot proposes the matching engagement.
Knowledge-graph entities → reference data. AI side surfaces entity relationships that improve marketing's L2-L3 graph writes.
Onboarding handshake. New AI tenant → optional warm handoff into Aria-Copilot intake. The bridge envelope carries the typed handshake; HITL approves the handoff.
Attribution / measurement. AI side can compute lift / attribution for marketing campaigns it consumed. Shared quant canon = same method on both sides.
Content + creative raw material. AI-generated long-form / research / drafts become marketing's brief inputs (with HITL approval before publication).
Shared methodology updates. When the AI side updates a canon ref or playbook, the marketing side pulls the change via the shared registry version contract.
Existing-copy discovery → Lore ingest. When the AI workspace surfaces something the client wrote (a doc, an email, a deck), a typed signal flows to Lore for the next ExistingCopyAudit cycle — so Sten's next BrandScript synthesis has the freshest source material.

The discipline that holds the synergy honest.

Five principles that keep the shared layer + the improvement loops from drifting into "we say we share" hand-waving.

Shared shapes are read-only registries. The atelier/shared/ public surface exports no tenant-id parameters; no credentials; no per-tenant overrides.
Tenant data stays separate by construction. The boundary is the type system, not docs.
Every cross-suite signal flows through a typed envelope (CrossSuiteSignal) with the safety stack composed at the boundary, vendor guard + regulator gate + HITL approval.
Every loop has a counter-signal. When a loop misfires (e.g. AI tenant declines the marketing handoff), the typed defection signal records the rejection so synergy is honest, not assumed.
Version bumps to the shared registry are ADR-level decisions. Both products pin a known version; the bridge envelope verifies cross-product version compatibility before crossing.

Procurement teams: the architecture is auditable.

Five architectural decision records (ADR-0181 tenant isolation; ADR-0225 cross-suite envelope shape; ADR-0259 safety framework; ADR-0261 cross-suite synergy architecture; ADR-0263 shared registries across Brilliant products) make up the evidence pack. The shared code + the bridge envelope make the boundary auditable in code, not just in policy.

Request the evidence pack → See the safety framework → See the workflow library →