01 / Integrations
mktskills.ai connects to Google Ads, Meta Ads (Facebook and Instagram), Google Analytics 4, and DV360. Platform access is stored securely and never shared with the AI layer.
Once connected, all data queries route through a centralized proxy that enforces the client isolation boundary. An analyst running a skill inside Client A's workspace queries Client A's Google Ads account. Not Client B's. Not a combined view.
CONNECTION STATUS
BetWarrior Group
Argentina · Brazil · Mexico
3 analysts
🔒 isolated
Sportsbetting.co.za
South Africa · Kenya
2 analysts
🔒 isolated
Luxury Automóveis
Brazil
1 analyst
02 / Client workspaces
Every client gets a dedicated workspace. When an analyst opens a workspace, they see that client's platforms, that client's account structure, and that client's historical reports. Nothing from any other client bleeds through.
Within a client workspace, account managers can configure sub-accounts — markets, regions, or business units — each with their own account mappings and optional data filters. A single GA4 property can serve multiple markets with country-level segmentation applied automatically.
03 / Skills library
Each skill is a structured AI instruction set built around a specific marketing analysis task. It knows what data to pull, what questions to ask, what patterns to flag, and how to structure the output for a client presentation. You run the skill; the AI does the analysis.
04 / Client isolation
Most tools trust the user not to look at the wrong tab. mktskills.ai makes the wrong tab structurally impossible.
Layer 1
Every request is authenticated against the active client context. Sessions are short-lived and scoped — there is no way to access the wrong workspace by accident.
Layer 2
Client isolation is enforced at the data layer, not just the application layer. Even if a bug existed in the software, a query for the wrong client's data is rejected before any records are returned.
Layer 3
AI skills run in isolated environments. Every data request is scoped to the active client — the AI cannot reach another client's accounts, regardless of what it is asked to do.
Layer 4
Platform logins are stored separately for each client and each integration. The AI never has direct access to them — all data requests go through a secure layer that supplies the right access automatically.
Layer 5
Agency plan and above. Formal Chinese walls: an analyst assigned to the "iGaming A" isolation group cannot be assigned to "iGaming B". Enforced in software, not in policy.
Layer 6
Every skill run, every data query, every login is logged with user ID, client ID, timestamp, and action. Append-only. Queryable by admins. 90-day retention.
05 / Reports
The AI drafts each report section from live data. You review, edit where your context matters, and export. The output is a branded PDF formatted for the client's reading level — not an analyst's.
Reports are stored per-client with full version history. Share with a time-limited link (30 days by default, configurable). Every section cites the data source, date range, and query that produced it.
REPORT_READY
Monthly Campaign Performance Review
BetWarrior Group · March 2026 · 12 sections · 2.4MB