← back to blog
    Berner SetterwallSeptember 25, 20265 min read

    Your AI Marketing Team, Drawn as an Org Chart

    The idea for this one came from a customer, and it started with LinkedIn.

    Everyone on that feed wants to show they're operating at the frontier — that they've got a whole team of agents doing the work. The observation was: that's literally what Cogny is. A Cloud customer with a dozen scheduled reports has a dozen specialists auditing their channels every week. They just never see it that way, because we show it to them as a list of report names.

    So: an org chart. Not a metaphor bolted onto a list — the same rows, drawn the way a team is drawn.

    What it looks like

    It's the Org chart tab on the Reports page — the same scheduled reports the Reports tab lists, drawn as a team. The top node is you. Below you is the Co-CMO — the ticket agent that turns findings into tickets and runs the ones you approve. Below that, your specialists, grouped by department.

    Here's a real workspace, unedited:

    23 specialists on duty · 122 analyses in the last 30 days · 89 findings filed · 66 shipped $503K estimated impact from shipped work

    And a smaller one:

    10 specialists on duty · 36 analyses in the last 30 days · 60 findings filed · 44 shipped $114K estimated impact from shipped work

    Each card is one scheduled report, rendered as a hire:

    SEO Specialist SEO Performance Signals Mondays at 09:44 UTC · ran 4d ago 19 analyses · 12 findings · 9 shipped $68K est. impact

    Nothing there is new work being done. It's the work that was already happening, described in the vocabulary people actually use for it.

    How a scheduled report becomes a job title

    Every template in templates/role-based/ maps to a role a human chose. seo → SEO Specialist. sem-manager → SEM Manager. content-intelligence → Content Strategist. data-quality → Analytics Engineer. landing-page-funnel → Landing Page Analyst.

    That mapping is a lookup table, not something inferred at runtime, and a test reads the templates directory and fails the build if a template has no role. Otherwise a new template would silently show up as a generic "Growth Analyst" on every customer's chart, and nobody would notice until a customer did.

    Roles group into five departments:

    • Paid Media — every ad platform audit
    • Growth & Performance — search, site, lifecycle
    • Creative & Audience — creative, content, competitor and customer-conversation mining
    • Strategy & Finance — revenue, budget, forecasting, churn
    • Data & Tracking — the measurement layer

    Prompts you wrote yourself

    Custom prompts have no template to look up, so the role is inferred from the name — and the interesting part is what happens when that inference is only half-confident.

    A real workspace has prompts called "Dutch Branding RSA — complete pause + 30-day monitor" and "Special Software — ad-group tROAS guardrail". Those are unmistakably paid-search work. But "Paid Media Operator" is not a useful job title, and three cards in a row all saying it reads like a broken classifier rather than a team.

    So the classifier distinguishes two outcomes. When it can identify the person, it shows the role. When it can only identify the department, it files the card under the right department and leads with the name you gave the prompt — which is more specific than anything we could infer anyway. Either way both lines are on the card: our classification and your actual prompt name.

    Every number counts only what shipped

    The tempting version of this feature inflates. We deliberately didn't:

    • Impact counts only tickets with status done. Open tickets and rejected tickets contribute nothing. A finding that nobody acted on isn't impact, and one workspace's rejected tickets alone carried more nominal "value" ($641K) than its shipped ones ($505K) — publishing that number would be a lie with a dollar sign on it.
    • Paused prompts show as off duty. They're not hidden to make the team look bigger, and they're not counted in "on duty".
    • Impact is labelled as an estimate, because it is — the agents' own estimate at ticket time. The chart says so in a footnote rather than presenting it as measured revenue.
    • Attribution is real. A finding is credited to a specialist by walking growth_tickets.source_context.reporting_agent_ids → the report run → the scheduled prompt that produced it. Where a ticket cites runs from two prompts, both specialists get credit on their own cards, and the headline total counts distinct tickets so it can't double-count.

    Sharing it

    Since the whole idea came from wanting to show this off, there's a Share button. It copies a 1200×630 image of the chart — departments, roles, the headline numbers — plus the post text.

    Two things it deliberately does not do. The impact figure has an explicit opt-out, because that's your revenue number and publishing it should be a decision, not a default you didn't notice. And your workspace name is never on the image: whoever posts it is already the attribution, and a company name on an image that then circulates isn't our call to make.

    Where to find it

    Open your workspace at app.cogny.com. Open Reports and switch to the Org chart tab. If you haven't scheduled anything yet, it tells you so and points at the template library — each template you schedule becomes another specialist on the chart.

    It's available to every Cogny Cloud workspace now.