The 6-tile Microsoft AI stack every SMB already owns (and the 4 they’re not using)

Most 30–300 person businesses on the Microsoft stack are paying for far more AI capability than they’re using. The average SMB we audit has activated maybe 20–30% of what’s already inside their Microsoft 365, Power Platform and Azure subscriptions — and is simultaneously evaluating standalone AI SaaS tools that duplicate the unused 70%.

This is the six-tile map we use to walk an SMB tenancy. Three of the six are almost always lit up but under-utilised. Three are almost never activated. Walk through it with a colleague over coffee and you’ll find at least $30–$80K of either activatable value or avoidable spend in the next 12 months.

The six tiles

1. Microsoft 365 Copilot

The familiar one. In-flow AI for Word, Excel, PowerPoint, Outlook, Teams and Loop. Sold per-seat at $360/user/yr (US list, May 2026) on top of a qualifying M365 plan. Most SMBs deploy it to 30– 80 seats and stop.

What’s under-used: the Outlook Prioritise My Inbox and Teams Recap capabilities, plus agent grounding via SharePoint and OneDrive. Most teams rate Copilot on whether it helps draft an email — the higher-leverage use is question-answering across enterprise content, which most users never try.

2. Copilot Studio

Microsoft’s low-code agent builder. Licensed via message-pack consumption or included in a Microsoft 365 Copilot plan (with consumption caps). Build branded conversational agents grounded on SharePoint, Dataverse, custom APIs, or a knowledge base, and deploy them to Teams, your website, or any channel.

This is the most-skipped tile in the average SMB audit — and almost always the highest-leverage one. Use cases that show up repeatedly: an HR-policy agent grounded on the staff handbook, a vendor-info agent grounded on procurement documents, a sales-enablement agent grounded on win-loss notes and proposal templates.

3. Power Automate + AI Builder

Power Automate is the cloud-flow engine that connects Outlook, Teams, SharePoint, Dynamics, Salesforce, ServiceNow and 1,000+ other systems. Most SMBs licence it via their M365 plan already. AI Builder layers pre-built AI models on top — document extraction, sentiment analysis, prediction, forms processing — invoked from inside a flow.

The under-use here is mostly a knowledge gap. Operations leaders know about Zapier; they don’t always know they have a Microsoft equivalent sitting in their tenant that handles identity, governance and audit cleanly. The migration from Zapier/Make/n8n to Power Automate typically saves 40–60% of automation tooling spend for a 100-person firm.

4. Azure AI Foundry (and Azure OpenAI)

Azure’s build-your-own-AI platform. Hosts foundation models, custom fine-tunes, embeddings, vector search, evals, and agentic frameworks. The right answer when Copilot Studio isn’t flexible enough — typically when you need fine control over retrieval, custom model selection, or production- grade evals.

Almost never deployed in SMBs because the team thinks of it as “the enterprise option.” In practice, even a 100-person firm can get value from Foundry on one or two custom workflows — e.g. retrieval over a 20,000-document SharePoint, a meeting-transcription pipeline with custom redaction, an embeddings index of your CRM win-loss history.

5. SharePoint Premium (with Syntex)

The most underrated tile on the map. SharePoint Premium adds AI-powered content services on top of standard SharePoint: Syntex document understanding, content assembly, e-signature, translation, image tagging, and metadata extraction.

Use cases: contract metadata extraction at intake, invoice OCR with line-item parsing, NDA classification, proposal assembly from a clause library, multi-language document translation. Each replaces a standalone tool that costs $400–$2,000/month on its own.

6. Microsoft Purview (with AI hub)

The governance tile. Purview is Microsoft’s data security and compliance platform; the AI hub specifically tracks AI usage across Copilot, third-party generative tools and custom agents, with DLP, audit, and risk insights.

This is the tile that turns an AI program from “using AI in Outlook” into “we have a defensible posture for the board.” Almost no SMBs deploy it. The two we’ve deployed it for in the last year reported a faster path through their enterprise customers’ vendor-risk reviews — Purview’s audit logs answered most of the questions directly.

The three tiles SMBs almost never activate

Copilot Studio, SharePoint Premium and Purview AI hub. Together these three are where 60–80% of the differentiated value sits for a 100-person company — and they’re the three the average customer has never even seen demoed.

The fastest way to add $100K of activatable value to your AI program isn’t to buy a new tool. It’s to walk these six tiles with someone who knows what they do.

The three-question test for which tile to use

Given a workflow, in what order do we ask “which tile?”

1. Is this in-flow productivity inside Office apps?If yes — e.g. drafting, summarising, analysing inside Word/Excel/Teams — the answer is M365 Copilot. Stop here.
2. Does this involve a repeatable question or a guided interaction? If yes — HR FAQ, vendor policy, sales playbook — the answer is Copilot Studio. Stop here.
3. Does this involve moving data, transforming documents, or chaining systems? If yes — invoice OCR, contract intake, multi-system handoffs — the answer is Power Automate + AI Builder, often with SharePoint Premium for the document-understanding layer.
4. Does this need custom models, retrieval over enterprise scale, or fine control? If yes — RAG over 50,000 docs, custom fine-tunes, agent orchestration — you’re in Azure AI Foundry territory.

Apply that ladder to ten of your real workflows. You’ll usually find seven that resolve at step 1–2 — cheaper, faster, lower governance burden — and three that legitimately need step 3–4. That mix is the basis for an honest Microsoft AI roadmap.

A 30-minute self-audit you can run before any vendor call

1. Open admin.microsoft.com. Pull your licence inventory.
2. For each licence type, mark which tile it covers (1–6).
3. Open Power Platform admin centre. Count active flows / apps / agents per environment.
4. Open Microsoft Purview compliance portal. Note whether AI hub is configured.
5. List the three workflows that consume the most analyst / ops / admin hours in your firm.
6. Run the three-question test against each.

You’ll come out of that audit with a list of activations that pay back inside one quarter and a list of new spend that was never going to. Most SMBs save more from the second list than from the first.

Next steps

If you want a structured version of that audit, our free AI Readiness Assessment runs the same diagnostic on your stack and produces a ranked opportunity list mapped to specific tiles. From there, a workflow activation sprint typically lights up the first one in production inside four weeks.