If you’ve spent any time in the Microsoft channel over the last twelve months, you’ve heard the word “Foundry” thrown around in everything from Ignite keynotes to MCAPS partner calls. And if you’re like most of the resellers I talk to, you’re probably nodding along in meetings while quietly wondering: what exactly is it, what does it do, and more importantly — how do I actually sell it to my SMB customers?

Fair questions. Let’s unpack them properly.

So what is Microsoft Foundry, really?

At Ignite 2025, Microsoft renamed Azure AI Foundry to Microsoft Foundry and re-positioned it as something bigger than a developer tool. The new framing puts Foundry alongside Microsoft 365 and Microsoft Fabric as the third pillar of the Microsoft cloud — the place where AI apps and agents are built, run, and governed. Microsoft calls it the “AI app and agent factory,” and that description is more accurate than it sounds.

Think of Foundry as a unified workbench. Inside it, your customers get:

•         A catalogue of more than 11,000 models from OpenAI (GPT-5, GPT-4.1), Anthropic (Claude Sonnet, Opus, Haiku), Cohere, Meta Llama, Mistral, DeepSeek, NVIDIA, xAI’s Grok, and Microsoft’s own MAI family for voice, transcription and image generation.

•         A model router that automatically picks the cheapest model capable of handling each prompt. Microsoft has reported up to ~50% cost savings on real workloads through routing alone.

•         Foundry Agent Service — the runtime that actually hosts agents in production, with multi-agent orchestration, memory, and direct deployment into Microsoft 365 and Agent 365.

•         Foundry IQ — a grounding layer (powered by Azure AI Search) that lets agents reason across data in SharePoint, OneDrive, OneLake, S3, Snowflake and a long list of others, without you stitching together a vector database yourself.

•         Foundry Tools — a catalogue of 1,400+ pre-built MCP connectors into business systems like SAP, Salesforce, Dynamics 365, ServiceNow, Shopify, and pretty well any SaaS app the customer is already paying for.

•         Foundry Control Plane — governance, identity (Entra Agent ID), security (Microsoft Defender), data protection (Purview), observability and policy enforcement, all in one portal.

Two things matter here for the channel. First, the Foundry platform itself is free to use — customers only pay for the services they consume (model calls, search, agent runtime, storage). Second, the whole thing runs on existing Azure subscriptions, meaning every Foundry workload pulls Azure consumption — which, as you well know, is exactly what your incentives, MCI and Azure Accelerate funding are aligned to.

How is it actually used?

Forget the buzzwords for a moment. In practice, Foundry is used to build three things:

1. Internal copilots. Customer-specific assistants that sit inside Teams, Outlook or the browser and answer questions grounded in the company’s own SharePoint, Dynamics, or ERP data. Think “an HR copilot that knows our leave policy” or “a sales copilot that can pull live pipeline from Dynamics 365 without opening it.”

2. Agentic workflows. This is where Foundry separates itself from a chatbot. An agent doesn’t just answer — it acts. A customer-returns agent, for example, can read the support email, look up the order in Shopify, validate the warranty in Dynamics, issue a refund through the finance system, and notify the customer — all without anyone clicking through five screens.

3. Domain-specific AI applications. Voice agents for call centres, document-intelligence agents for invoice and contract extraction, marketing agents that draft copy grounded in brand guidelines and live product data. These get deployed once and repurposed across many customers.

The five-pillar architecture (Models, Agents, Knowledge & Tools, Observability, Local & Edge) is genuinely modular. Customers don’t have to adopt everything on day one. Most start with a single agent on a single workflow and grow from there.

Why this is a real opportunity for SMB?

Historically, “enterprise AI” felt out of reach for SMB customers — too much infrastructure, too many specialists, too long a payback period. Foundry collapses all three of those barriers.

•         No infrastructure to manage. Hosted agents, managed memory, managed RAG. The customer doesn’t need a data engineer to stand up a vector database.

•         Pay-as-you-go economics. An SMB processing 200 invoices a month doesn’t pay enterprise-scale pricing. They pay per token, per hour of audio, per agent run. Cost scales with value.

•         Low-code on-ramp. Foundry plays nicely with Copilot Studio, Power Automate and Logic Apps. Your customers’ existing power users — the ones already building flows — can prototype agents without learning Python.

•         Trust baked in. The Control Plane integrates Entra, Defender and Purview by default. That matters when you’re selling AI into a 50-person legal firm or a regulated medical practice; the “but what about our data?” conversation has a credible answer.

•         It bundles. Foundry-built agents deploy straight into Microsoft 365, Teams, Outlook, Dynamics. So instead of selling AI as a separate line item, you’re enriching the Microsoft stack the customer already pays you for. Stickier customer, higher ARPU, longer contract.

In short: the SMB use cases that used to require a custom-build project are now repeatable Foundry deployments. Sell once, scale many times.

Actionable next steps for the reseller base

Here’s where I’d point our partners next week, not next quarter.

1.       Pick three repeatable plays and productise them. Don’t try to sell “Foundry” — sell outcomes. A meeting-summary-and-action-item agent, an invoice processing agent, and a customer support deflection agent will cover 80% of SMB conversations. Build each one once, demo it, and have a fixed-price offer ready.

2.       Lead with an AI Readiness Assessment. A two-hour discovery workshop is a much easier “yes” than a project quote. Partners with the right Azure specialisations can apply for Azure Accelerate funding to cover the cost of these workshops and proofs of concept, meaning the reseller’s first conversation with the customer could effectively be Microsoft-funded. Stack this with Solution Assessments where available.

3.       Mine the existing M365 base first. Every customer paying for Microsoft 365 Business Premium already has the identity, data and security plumbing in place for Foundry. Filter your CRM for customers with Dynamics 365, SharePoint-heavy deployments, or call-centre workloads, those are your warmest leads, full stop.

4.       Bundle the managed service. SMBs don’t want to “operate AI.” They want it to work. Wrap every Foundry deployment with a monthly managed service: monitoring, prompt tuning, model-router optimisation, responsible AI checks. This is recurring margin on top of Azure consumption, and it dramatically reduces churn.

5.       Train two people, not twenty. Partners don’t need an army of AI engineers. They need one technical lead who can deploy a Foundry agent end-to-end, and one solution seller who can confidently run the customer conversation. Microsoft Learn has free Foundry learning paths; the Agent Service quickstart is genuinely a one-afternoon exercise.

6.       Use the Microsoft Marketplace. Resale-enabled offers in the Marketplace let partners transact Foundry-based ISV solutions with Azure consumption commitments — meaning customer spend retires MACC, and your partners get co-sell credit where applicable.

The bottom line

Foundry is the most channel-friendly AI platform Microsoft has shipped to date. It’s free to enter, consumption-priced, governance-ready, and it lands inside the Microsoft estate that resellers already manage. The customers who win in the next eighteen months won’t be the ones with the cleverest models, they’ll be the ones whose partners helped them put practical agents into production first.

That customer can be yours. Start with one workflow, one customer, one funded assessment. The rest takes care of itself.

Reach out to us at channel@4sight.cloud to discuss how to position, implement, and monetise Microsoft Foundry.