1. Ampler.io
  2. Resources
  3. Integrations
  4. AI integrations
  5. Connect Ampler AI to your own AI model

Connect Ampler AI to your own AI model

Ampler brings AI into PowerPoint and handles the actions, while the model behind it is entirely your choice. Connect the one you trust through Microsoft Foundry today. Direct connections to a wide range of AI providers are planned.

Your options at a glance

Connection pathStatusWhat you needWho bills you
Microsoft Foundry (Azure)Available todayA model deployed in Microsoft FoundryMicrosoft, per token
Direct API connection to your AI provider (OpenAI, Anthropic and more)PlannedAn API key from your providerYour provider, per token
Microsoft 365 CopilotNot supportedNot applicableNot applicable

Ampler does not resell AI usage. You bring your own model, and the provider you choose bills you directly for what you use.

How it works today: connect through Microsoft Foundry

Microsoft Foundry is the place in Azure where you deploy and manage AI models. You deploy a model there, and Ampler AI uses that deployment.

  1. Open Microsoft Foundry and deploy the model you want to use.
  2. Copy the deployment’s endpoint URL, API key and deployment name.
  3. In PowerPoint, open Ampler > Company Settings > Integrations and find the Azure OpenAI section.
  4. Paste the endpoint URL, API key and deployment name, and save.

The endpoint URL must end with /responses. The full step-by-step guide, with screenshots of where to find each value, is in Integrate Ampler with Microsoft Foundry / Azure OpenAI.

We currently recommend GPT 5.3 Codex or GPT 5.4, which give a good balance between performance and price. Pricing is set by Microsoft and published on the Azure pricing pages. Anthropic models are available in Foundry as a preview, but Ampler does not support them yet. Support is coming soon.

What is coming: connect directly to your provider

We are working on letting you connect Ampler AI directly to your AI provider through its API, whether that is OpenAI, Anthropic or another provider. You would create an API key with the provider and enter it in Ampler, without going through Azure. This is a good fit if your organization already works directly with one of these providers. We have not set a release date.

Subscriptions and APIs are not the same thing

A subscription to a chat product, such as ChatGPT or Claude, gives people access to that provider’s own chat app. It does not include API access, which is what Ampler AI needs. API access is set up separately and billed per token, meaning by the amount of text the model reads and writes.

This means the subscriptions your people already have do not limit your choice. You can have a Claude subscription and still run Ampler AI on an OpenAI model through Microsoft Foundry, because the two are independent. Ampler AI works well no matter which model you choose in Foundry.

Why Microsoft 365 Copilot is not an option

Copilot is a finished assistant that Microsoft runs inside Microsoft 365. It is not a model that other applications can connect to, so it cannot power Ampler AI. If your organization wants Microsoft-hosted models, Microsoft Foundry is the route to use.

Ampler AI can run alongside Copilot. Your people can keep using Copilot and use Ampler AI in PowerPoint with the model you have connected through Foundry.

Frequently asked questions

Which AI models can I use with Ampler AI?

Today you can use the OpenAI GPT models you deploy in Microsoft Foundry. We recommend GPT 5.3 Codex or GPT 5.4. Anthropic models are coming soon, and direct connections to a wide range of AI providers are planned.

Do I need a subscription to ChatGPT, Claude or Copilot?

No. Those subscriptions give access to the providers’ chat apps, not to the API that Ampler AI connects to. You need a model deployment in Microsoft Foundry.

What does it cost?

The model provider bills you per token. For Microsoft Foundry, prices are listed on the Azure pricing pages, and they differ between models. Your Ampler license is separate.

Who sets it up?

Typically an IT administrator or whoever manages your Azure resources. They deploy the model in Foundry and enter the endpoint URL, API key and deployment name under Ampler > Company Settings > Integrations. The step-by-step guide covers each step.

Can we use Microsoft 365 Copilot instead?

No. Copilot is a finished assistant, not a model endpoint that Ampler can connect to.

Can we connect directly to OpenAI or Anthropic without Azure?

Not yet. It is planned, and when it is available you will be able to connect with an API key from the provider.

If I have built up knowledge in my Claude subscription, will the model have it when it runs through the API?

No. A subscription does not train or change the underlying model. What you build up in the chat app, such as projects, uploaded files, saved memory and custom instructions, stays in that app. A model called through an API, or through Foundry, starts from the provider’s standard model and does not have access to that material. Anything the model needs to know has to be part of the request it receives.

Why does Ampler not include an AI model?

Because you stay in control of which provider and model you use, and under which agreements your data is handled. Your organization chooses the model that fits its own security and compliance requirements.

It also keeps pricing simple. You pay only for your Ampler product, and all AI consumption is kept out of the Ampler price. What the model costs is between you and your provider.

Can we switch model later?

Yes. You change the deployment name, endpoint and key in Company Settings, and you can deploy a different model in Foundry at any time.

What are the advantages of running the model through Microsoft Foundry?

You get a model that runs in your own Azure environment, under your existing Microsoft agreement. The main advantages are control over where data is processed, Azure’s security and access controls, and one place to manage models and cost. The next questions go through each.

Can the model run in Europe?

Yes. When you deploy a model in Foundry, you choose a deployment type that decides where your data is processed. A Data Zone deployment for the EU processes data only within the EU data zone, and data stored at rest stays in the Azure geography you chose. Global deployment types can process data in any Azure region, so choose a Data Zone or regional type if you have a European data residency requirement. See Microsoft’s page on deployment types for details.

Can we use our existing Azure security and access controls?

Yes. The model is a resource in your Azure environment, so your IT team can manage it with the tools they already use for other Azure services, such as role-based access, managed identities and network controls.

Do we get one bill and one place to manage usage?

Yes. Usage is billed by Microsoft per token, together with the rest of your Azure consumption, so your IT and finance teams see AI cost where they already see cloud cost.

Can we choose between several models?

Yes. Foundry gives you access to several models in one place, and you can deploy a different one at any time. Ampler AI works well no matter which model you choose in Foundry.

Where can I get help?

Contact us at hello@ampler.io and we will help you get connected.

Was this article helpful?

Related Articles

Try free