Bring your own model and pod
Your AI model, your runners. You pay for the platform.
AnnE thinks with the AI model your company already pays for, and your jobs run on your own servers. Work on your own model uses none of your plan's AnnE builds.
Four ways to run
Ours, yours, or both
Our models, included
Every plan comes with AnnE thinking on our models, within the plan's AnnE builds.
Your own model, no limit from us
Connect the AI model your company already pays for. AnnE's work on it uses none of your plan's AnnE builds or AI allowance: your model vendor bills you for it directly.
Your own pod, on your servers
A pod is the runner that does your jobs' work. Run it inside your own network or cloud account, next to your data, with on-premise runners.
Both: you pay for the platform
With your own model and your own pod, the AI and the computing are yours. Ductile AI charges only for the platform: your plan and its credits.
A model for each part of the work
Spend the strong model where it matters
Choose one model for everything, or a different one for each part of AnnE's work, to bring the cost down.
Planning
Reads what you asked for and plans the job. Give it your strongest model.
Filling in steps
Settings, formulas and field mapping for each step. A faster, cheaper model does this well.
Talking with you
The quick back-and-forth while AnnE works. The fastest model you have.
Your jobs' AI steps
Classify, extract and summarise inside your jobs, on your model unless a step names another.
Models you can bring
From the vendor you already use
Anthropic
Claude, with an API key
Google Vertex AI
Claude or Gemini, with a service account
Amazon Bedrock
Claude and more, in your AWS account
Microsoft Azure AI Foundry
Claude and the models you deploy
Any OpenAI-compatible service
With its address and key
Example: Google Vertex AI
Set it up in a few minutes
Other vendors work the same way: choose the vendor and paste the key it gives you.
First, in your Google Cloud account
- 1
A project with billing
In Google Cloud, a project with billing on and the Vertex AI API turned on.
- 2
The models you want
In Vertex AI's Model Garden, enable Claude (accepting Anthropic's terms there) in a region that offers it. Gemini needs nothing extra.
- 3
A key for Ductile AI
A service account with the Vertex AI User role, and a JSON key for it.
Then, in the Ductile AI console
- 4
Add your model
In the console, Settings > AI gateway > Add model. Give it a short name, choose Claude on Google Vertex AI (or Google Vertex AI for Gemini), the model, your project and region, and paste the JSON key. It is sealed and never shown again.
- 5
Check it
Press Check on the new row: the console makes one call to your model and says whether it answered.
- 6
Choose what AnnE uses
Under What AnnE uses, pick your model, and if you like a different one for planning, filling in steps and talking. Your jobs' AI steps use it too unless a step names another.
- 7
Set your own limits
Optional: under Limits and policy, a monthly budget or a daily cap of your own.
Questions about your own model and pod
Which plans can use their own model?
Every plan, Free included. An admin of your organisation sets it up.
Does using my own model cost credits?
No. AnnE's work on your model uses none of your plan's AnnE builds or AI allowance, and AI steps in your jobs already run on your own provider account. Steps of Premium apps use credits as usual, wherever they run; standard apps are free.
What reaches the model?
Names, settings and the shape of your data. Your records' values and your sign-ins never reach any model, ours or yours.
What if my model does not answer?
Check on the model's row shows straight away whether it answers. Your jobs' AI steps can turn to another of your models, named under If it does not answer, use, and you can put AnnE back on our models at any time.
Can I go back to your models?
Yes. Choose This console's own model and AnnE thinks on ours again, within your plan.
Which plans include on-premise runners?
Business includes one on-premise cluster and Enterprise includes them; on Team they are an add-on. See pricing.