The one-line difference
CoModeler helps you build the model. AI Gateway helps you ask questions of the model.
What CoModeler does
CoModeler is an AI assistant inside Anaplan for model builders. You describe what you want in plain language, for example “add a headcount module by department with a monthly roll-up.” It proposes the changes, and you approve them before anything is built.
Changes show up in the normal model history under the name of the person who directed them, so nothing happens in the dark. It’s a separately subscribed service, switched on per workspace, and you need to be a workspace admin to use it.
If you’re a planner, you’ll rarely open CoModeler yourself. You’ll feel it indirectly, because the model you rely on gets built, fixed and extended faster.
What AI Gateway does
AI Gateway is a governed connection (via MCP, the Model Context Protocol) between your Anaplan models and the AI tools your team already uses, such as Claude, Gemini or Microsoft Copilot. You stay in your AI assistant and ask:
- “What’s our forecast variance for Q4 by region?”
- “Which cost centres are over budget this month?”
- “Summarise what changed between the last two forecast versions.”
The answer comes from your live model, not from a stale export in a spreadsheet. Access is read-only to begin with. Anaplan says write access is coming, so the assistant could later update data as well as read it.
Gateway uses your existing Anaplan permissions. You can only get back what you’re already allowed to see.
Side by side
- Who it’s for: CoModeler is for model builders and consultants. Gateway is for planners, analysts, finance partners and anyone who needs answers.
- Where you use it: CoModeler runs inside Anaplan. Gateway runs in your external AI assistant.
- What it does: CoModeler creates and changes model structure. Gateway retrieves data from models and, later, updates it.
- Who’s in control: With CoModeler, the builder approves every proposed change. With Gateway, Anaplan permissions decide what you can reach.
- The risk to manage: CoModeler needs review of what it proposes. Gateway needs clean permissions.
How this helps your day-to-day
For planners and end users. Gateway cuts the “can you pull me that number?” loop. Instead of waiting on a report or a colleague, you ask your AI tool and get an answer grounded in the model. It also turns your assistant into a quick sanity check before a meeting: ask for the biggest movers, then dig into the detail in Anaplan.
CoModeler helps you less directly. When you need a new view, a new calculation or a new dimension, the builder has a faster path from your request to a working model.
For consultants. CoModeler takes a good share of the repetitive build work: scaffolding modules, drafting formulas, extending structures. You spend more time on design and validation. Gateway changes the conversation with clients, because users can query models in the tools they already live in. It also raises a new question for every engagement: who can see what?
One thing to do before switching on Gateway
This is the part people skip. Gateway is enabled per workspace, not per app. Anaplan’s own guidance puts it plainly: “it isn’t in an App” is no longer a valid reason to assume data is hidden. If a user has access to a model in a Gateway enabled workspace, their AI assistant can reach it.
That means the old safety net of “nobody sees it because it’s not on a page” is gone. Before enabling it, Anaplan recommends a short hygiene review:
- Audit your models and their owners.
- Archive models nobody uses.
- Check Model Roles, Selective Access and Dynamic Cell Access.
- Move models to the right workspaces.
- Validate any moved models.
- Check users in every model.
Keep Production models in their own workspace, and add new users with No Access by default.
Which one do you need?
If your pain is “building and changing models takes too long,” look at CoModeler. If your pain is “getting answers out of the model takes too long,” look at AI Gateway. Most mature teams will end up wanting both: one to build well, the other to put the result in people’s hands.
Have you tried either one yet? I’d like to hear what’s worked and what hasn’t.
~Bhumit Jain
Sources: