Supervised Intelligence: what it does and where to start
Supervised Intelligence runs your business on your rules. You set the instructions, capability levels, data labels, and tools once, and it runs on that supervision. Maya, Genuin's orchestrator, routes each task to the best model, agent, or tool for the job.
Most AI today is unsupervised. One model runs the whole job and spins up agents in the background, without your control. Supervised Intelligence picks the right path for each task instead. A smaller model classifies your video when it does that job better and cheaper, and a larger one steps in only when the task calls for it. Closed models still work when you connect them through Marketplace.
Octo, Genuin's Supervised Intelligence interface
Octo is Genuin's Supervised Intelligence conversational interface, available across Genuin's website, Genuin's apps, and Brand Control Center (BCC). It helps creators and customers find their way around Genuin's products and features and put them to work. In BCC, Octo operates your account with you. It reads your brand context, proposes each change with the setup already done, and carries it out once you approve. For example, it proposes your autonomy settings with anything public, paid, or sent set to ask every time. Each kind of change runs at the autonomy level you choose, and every action is logged with the approval that allowed it. Octo operates your Brand Control Center explains the model, and Supervised Intelligence: the orchestration and generative commerce layer gives the full picture.
Maya, Genuin's orchestrator
Maya is Genuin's orchestrator. Maya harnesses business-level capabilities, with multi-model, multi-agent, and multi-tool orchestration, for the Supervised Intelligence you deploy, toward your growth and monetization outcomes.
Maya isn't a separate product to buy or set up. Maya ranks every feed from the placement's context and the viewer, and picks between the experiences a placement allows. It's measured on per-minute watch time per user and per-minute revenue, with a standing holdout isolating Maya's contribution. Personalization and relevance: how Maya ranks every feed explains the signals Maya uses and the parameters you control.
Get started
- Give Supervised Intelligence your brand context. Point the crawler at your properties and add guidelines, catalog, and policies.
- Set its autonomy and approvals. Choose which changes it asks about every time, and which it does and reports.
- Read Supervised Intelligence's analytics dashboard. The Relevance dashboard shows how well each feed matched its viewer, what that did to engagement and revenue, and what to change next.
Ask, act, and resolve
| Capability | What it does | Where it fits |
|---|---|---|
| Ask | Answers questions from your reports, with the chart and rows behind the answer | Any BCC user |
| Act | Turns instructions into drafted actions across products, run on your approval | Any BCC user with the role for that action |
| Resolve | Turns a tagged moment or a viewer's question into a commerce outcome at serve time | Commerce and retail models |
| Generative Commerce Routing | Routes intent your catalog can't serve to retail assistants and marketplaces | Retail and commerce media networks, media brands without a catalog, and brands sold through retailers |
| Generative Player for LLM Applications | Lets AI assistants call your own player, so viewers there watch and shop in your experience, under your rules, and the outcomes report to you | Early access, for brands that want their own player in AI assistants rather than a sponsored answer |
Run the network with Canvas
- Canvas: run your network with growth and monetization agents: how the fleets work, and the guardrails you set.
- Canvas Growth Fleet: trend scout, producer, editor, scheduler, and community manager agents.
- Canvas Monetization Fleet: packager, pricer, partner liaison, yield manager, and reporter agents.
Read the signals
- Relevance Dashboard: relevance, engagement, revenue, and recommended changes for every placement on your grid, with the lift from personalized ranking measured against a holdout.
- Audience Segments: cohorts from onsite, offsite, and offline signals, joined under consent.
- Attribution Models: last touch, multi-touch, incrementality, and exposure-to-purchase models.
- Margin Reporting: revenue, cost, and net margin per impression, by placement and partner.
Build it
Engagement data reaches your systems today in two ways. The SDKs send engagement events to your own analytics as they happen. Scheduled exports send event rows to BigQuery, Mixpanel, or Amazon S3. Query Genuin event data in BigQuery shows the first queries to run.
The Supervised Intelligence calls in the Business Account API are proposed, not live. Live API today and the Business Account API roadmap maps what you can call now to what's planned. The Supervised Intelligence SDK puts Supervised Intelligence chat on article, event, and video pages, Generative Destinations, and video posts, and in the Genuin player. Viewers there ask about what they're reading or watching.
Questions people ask
- How do I make Octo smarter about my brand? Asked by founders and business owners.
- Does Octo change our code, and who approves it? Asked by developers and engineering.
- Which routine work can Supervised Intelligence do without asking? Asked by operations and ad ops.
