TL;DR:
- Microsoft Power Automate with AI Builder makes most sense if you’re already paying for M365 E3/E5 — the licensing value is real, especially for document processing
- Copilot integration is genuinely useful for non-technical users building flows; it’s the best no-code flow builder in the platform
- For cross-platform integrations or teams not embedded in the Microsoft ecosystem, Zapier or Make will be faster to implement and maintain
Microsoft Power Automate has been the enterprise automation platform that practitioners either swear by or struggle with since its Flow-era rebrand. In 2026, the AI Builder additions and Copilot integration have meaningfully changed the value calculation — particularly for UK organisations already deep in the Microsoft 365 stack.
What AI Builder Actually Adds
AI Builder is the umbrella term for Microsoft’s pre-built and custom AI models that you can use as steps in Power Automate flows. The most production-ready capabilities are distinct enough to cover individually.
Document Processing (Form Processing) trains a model on your documents — invoices, purchase orders, contracts — and extracts structured fields. You upload sample documents, label fields in a visual UI, and publish a model. Quality is solid for consistently formatted documents; it degrades meaningfully with highly variable layouts.
Receipt Processing is a pre-built model with no training required. Point it at a receipt image or PDF and it returns vendor, total, date, and line items with reasonable accuracy. Practical for expense report automation.
Text Classification lets you train a custom model to categorise text into your own label set — useful for routing support tickets, tagging feedback, or classifying leads. You’ll need 50–100 labelled examples minimum for useful accuracy.
Prediction is a structured data classification/regression model. Less commonly useful than the document and text models, but valuable for approval workflow routing based on historical patterns.
| AI Builder Model | Training Required | Best For |
|---|---|---|
| Form Processing | Yes | Invoices, contracts, structured docs |
| Receipt Processing | No | Expense automation |
| Text Classification | Yes | Ticket routing, lead tagging |
| ID Reader | No | Identity verification workflows |
| Business Card Reader | No | Contact capture |
Copilot Integration: The Real UX Change
The Copilot integration in Power Automate is the most significant usability change in years. You describe what you want in natural language — “When a new row is added to my SharePoint list, send an email to the manager in column D and create a task in Planner” — and Copilot generates the flow structure with triggers and actions pre-configured.
For power users it’s useful for scaffolding quickly. For non-technical users it’s a genuine step change — they can build flows that previously required an IT request. The generated flow usually needs cleanup (error handling, edge cases) but it’s a solid starting point.
Where Copilot falls short: complex conditional logic, cross-system flows with many steps, and anything involving custom connectors. It understands Microsoft products well and everything else adequately.
Licensing and Pricing for M365 Users
This is where Power Automate’s value proposition gets interesting for UK enterprise buyers. The included M365 tier gives you limited flows and no premium connectors — fine for simple Office 365 automation. Power Automate Premium (roughly £12/user/month) adds premium connectors (Salesforce, Workday, etc.), process mining, and a base allocation of AI Builder credits. Additional AI Builder credit packs are available for high-volume document processing.
For a 500-person organisation on M365 E5 already, adding Power Automate Premium for the team that actually builds flows is often cheaper than an equivalent Zapier Teams plan — especially when document processing volume is high. Run the maths against your specific connector requirements; premium connectors are the wildcard.
How It Compares to Zapier and Make
| Dimension | Power Automate + AI Builder | Zapier | Make |
|---|---|---|---|
| M365/Azure integration | Excellent | Good | Good |
| Non-Microsoft integrations | Good (1000+ connectors) | Excellent (7000+) | Very Good (1500+) |
| AI/ML capabilities | Strong (AI Builder) | Basic (AI steps) | Limited |
| Document processing | Best in class | Requires external AI | Requires external AI |
| Learning curve | High | Low | Medium |
| Pricing for enterprise | Competitive with M365 | Expensive at scale | Best price/volume |
| Debugging / observability | Moderate | Good | Good |
The pattern that emerges: Power Automate wins on Microsoft-adjacent workflows and document processing. It loses on integration breadth for non-Microsoft systems and on time-to-first-working-flow for teams without existing Power Platform experience.
Practical Limitations
The Power Automate experience still has rough edges. Run history can be difficult to navigate for debugging at scale — filtering and tracing failures across many flow runs is slower than in Zapier or Make. Custom connector setup requires more technical overhead than equivalent integrations in competing platforms. Version control and team collaboration on flows improved in 2025 but still lags developer-grade tooling. And AI Builder credits deplete faster than anticipated at scale; audit consumption before rolling out document processing broadly.
Bottom Line
Microsoft Power Automate with AI Builder earns a strong recommendation for organisations already embedded in M365, especially where document processing automation is a priority. The licensing maths often makes it the most cost-effective option at enterprise scale, and the Copilot integration has genuinely lowered the floor for non-technical builders. If your team isn’t in the Microsoft ecosystem, or if you need broad third-party integration coverage from day one, Zapier or Make will get you to production faster.