Make vs Power Automate comes down to where your business runs: choose Power Automate when your work lives in Microsoft 365, you need desktop automation on Windows or IT wants tenant level control, and choose Make when your processes cross many non Microsoft apps and you want a visual builder. If you came to learn how to make a flow in Power Automate, the invoice example below builds one in both tools.
The wrong choice rarely fails on day one. It fails months later, when a premium connector changes the license math, an approval must wait days for a manager or the person who understood the scenario leaves. Plan names and prices change often, so this page explains each cost model and leaves current numbers to the vendors.
Where teams get stuck choosing between Make and Power Automate
The license math changes after the build
A flow that uses only Outlook and SharePoint can run under many Microsoft 365 licenses. Add one premium connector and every user who runs it may need a paid Power Automate license.
Usage bills grow with success
Make meters usage against a plan allowance tied to module actions. A scenario that loops over every invoice line uses far more allowance than one that batches the data first.
Approvals that must wait for a person
Power Automate has a built in approval action that pauses the flow. Make needs a webhook or a second scenario for long waits.
Desktop apps with no API
Old accounting software that only works on a Windows screen needs desktop RPA. Power Automate has it. Make does not.
The short answer on Make vs Power Automate
Pick Power Automate if your documents sit in SharePoint and OneDrive, your team works in Outlook and Teams and IT manages users in Microsoft Entra ID. Approvals, connectors and admin policies come inside the tenant you already pay for. Pick it too if a clerk spends hours typing into a Windows desktop application, which Power Automate Desktop can drive.
Pick Make if your processes run through Shopify, HubSpot, Airtable, Slack and other web apps, and Microsoft is only one of them. Make's canvas shows each module and the exact data it received, which makes branching, multi app scenarios easier to build and debug.
Then ask who maintains it. A business with an in house Microsoft admin should usually stay in Power Automate. An operations lead who maps data well but avoids admin centers often moves faster in Make.
Power Automate licensing: standard vs premium connectors
Power Automate splits connectors into standard and premium. Many Microsoft 365 business licenses cover cloud flows that use standard connectors such as Outlook, SharePoint, Teams and OneDrive. Premium connectors, the HTTP action, custom connectors and on premises data through the gateway generally need a paid Power Automate license.
Paid licensing comes in two broad shapes: per user, where each person who runs premium flows is licensed, and per flow or process, where one automation is licensed however many people trigger it. Unattended desktop RPA is licensed separately, and AI Builder capacity for reading documents is sold on top.
So your cost depends on the connectors you touch, how many people run the flow and whether a robot runs unattended. Check Microsoft's current licensing guide before you design, because one premium step can move the whole flow into a different tier.
Make scenarios: operations, routers, iterators and error handlers
In Make an automation is a scenario made of modules. A trigger module watches for new data and later modules act on it. Make meters usage against a plan allowance that follows module actions in each run, so cost depends on how many modules run and how many items pass through them. Check Make's current pricing page for the exact unit.
Four blocks do most of the work. Routers split one stream into filtered branches. Iterators turn an array, such as an email's attachments, into separate bundles. Aggregators collect bundles back into one item. Error handlers decide what happens when a module fails: retry later, skip, roll back or stop.
An iterator in front of five modules multiplies their usage by the item count. Filter early, aggregate before writing and poll less often.
Same workflow both ways: invoice email to SharePoint to approval
The workflow: supplier invoices arrive in a shared mailbox. Each PDF is saved to SharePoint, its supplier, number, amount and due date captured, and the invoice sent to the budget owner. Approved invoices are marked ready to pay. Rejected ones go back to accounts payable with a reason.
In Power Automate, the Outlook trigger watches the shared mailbox for messages with attachments. An Apply to each loop saves every PDF with the SharePoint create file action. AI Builder's invoice model, which uses paid capacity, reads the fields, or a person enters them in a SharePoint list. Start and wait for an approval sends a card to the owner in Teams and Outlook and pauses the flow until they answer. A condition then updates the list item.
In Make, the Microsoft 365 email module watches the mailbox, an iterator splits the attachments and a filter keeps only PDFs. A SharePoint module uploads each file, and an AI or document parsing module, or an HTTP call to an extraction service, reads the fields. The approval is the hard part: Make sends a Teams message with approve and reject links that point to a webhook, and a second scenario receives the click and updates the record. It works, but you build the waiting yourself.
In both versions a person stays in the loop twice: the approver decides, and anything extraction cannot read with confidence goes to a review queue instead of being guessed. Power Automate wins this one on native approvals; Make wins if the scenario must also post to a non Microsoft accounting tool. Measure the share of invoices needing manual review, approval wait time and failed runs.
- Same risk in both: a supplier who sends links or scanned images instead of PDFs breaks the happy path, so plan that exception first.
Governance and security: Entra ID, DLP policies and Make teams
Power Automate inherits Microsoft's admin model. Users sign in with Entra ID, environments separate testing from production and data loss prevention policies decide which connectors may be combined, for example blocking a flow that copies SharePoint files to a personal Gmail. For firms that already govern Microsoft 365 centrally, that is a strong reason to stay.
Make organizes work into organizations and teams with roles, and stores connections per team. Single sign on and stricter controls depend on the plan, and Make cannot be self hosted. On either platform, build in a service account or shared team the business controls, never one employee's personal workspace, and list every connection and whose credentials it uses.
When neither fits and n8n or Zapier is the better call
Zapier is often fastest for simple, linear automations between popular web apps, such as a form that creates a CRM contact and posts to Slack. It bills per task, so high volume multi step work gets expensive.
n8n fits when volume is high, logic is complex or data must stay on infrastructure you control. It can be self hosted, its cloud plans count whole workflow executions rather than steps, and it asks more of whoever maintains it.
Sometimes you should not automate yet. If the invoice process changes monthly, or nobody can say who approves what, run it by hand for a few weeks first. An automation locks in whatever process you give it.
Getting the flow built in accounts you own
Benian's Workflow Automation work maps the process and its exceptions first. We then build and test in the client's own n8n account, with connections, approval rules and failure handling defined in scope, and credentials the client holds. If your stack is Microsoft 365 and Power Automate is the right tool, we say so on the first call instead of selling you a rebuild. You keep the workflow files and an operating guide.
Do not hire anyone for a two step flow inside Microsoft 365; a template will do it in an afternoon. Outside help pays when work crosses several systems, needs approvals and error handling, or must outlive its builder. The free Opportunity Map or a 30 minute call will tell you which case you are in.