Excelgoodies logo +48 73 2084245

LEARN THIS HANDS ON

Power Apps & Power Automate

. Live Online FILLING FAST
View all upcoming batches
Copilot in Power Automate Doesn’t Speak Polish Well – Here’s How to Prompt It Anyway

Copilot in Power Automate Doesn’t Speak Polish Well – Here’s How to Prompt It Anyway

Most Polish teams trying Copilot in Power Automate hit the same wall: you describe a flow in Polish, Copilot answers partly in English, misreads your intent, or silently ignores key details. Copilot is optimised for English, but you still want it to help you build real flows. This article shows how to structure prompts so Copilot behaves predictably, even when your day‑to‑day work is in Polish.

Scenario: HR team automating onboarding emails

Imagine an HR operations team in Poland.

They need a cloud flow that:

  • Triggers when a new item is added to a SharePoint list OnboardingRequests
  • Sends a personalised email in Polish to the new hire
  • Creates a task in Planner for the HR specialist
  • Logs the action to a Teams channel

The HR specialist opens Power Automate (cloud), goes to Create → Describe it to Copilot, and types:

"Kiedy dodaję nowy wiersz do listy SharePoint z danymi pracownika, wyślij mu mail powitalny po polsku i dodaj zadanie w Plannerze dla działu HR."

What happens:

  • Copilot responds in English
  • It correctly picks SharePoint as the trigger
  • It suggests Send an email (V2) and Create a task in Planner
  • The generated email body is in English, and some field mappings are wrong

The flow is not useless, but it’s far from what the team wanted. We’ll use this scenario to walk through concrete prompt patterns that make Copilot more reliable in this mixed-language situation.

What Copilot in Power Automate actually supports today

Before we tweak prompts, it helps to be clear on what Copilot can and cannot do as of late 2026.

Language support constraints

Based on current Microsoft documentation and product behaviour:

  • Optimised for English: Copilot in Power Automate is primarily trained and tuned for English instructions.
  • Other languages have limited support: You can type in other languages, including Polish, but:
    • Intent recognition is less reliable
    • It may misinterpret domain terms or business acronyms
    • It may return answers in English even when you ask in Polish
  • Generated artefacts can be localised manually: Triggers, actions, and expressions work regardless of language; you can edit labels, email bodies, and messages to Polish after Copilot scaffolds the flow.

Where Copilot sits in Power Automate

For practitioners, the key Copilot touchpoints today are:

  • Describe it to Copilot (Create pane)
    • You type a natural language description of the flow
    • Copilot proposes a flow with trigger, actions, and basic configuration
  • Copilot in the flow designer
    • You can ask for changes ("Add a condition", "Use a filter array")
    • You can ask for help with expressions (e.g. composing concat, formatDateTime)
  • Copilot for cloud flows only
    • It currently applies to cloud flows (automated, instant, scheduled)
    • Desktop flows and RPA have different AI assistance patterns and are outside this article

None of these depend on your tenant language being Polish. The limitation is in the natural language layer, not in the flow engine itself.

Strategy: Separate “instruction language” from “business language”

The main pattern that works reliably for Polish teams is:

Give Copilot instructions in clear English, but describe business data and final content in Polish inside those instructions.

This sounds awkward at first, but it plays to Copilot’s strengths:

  • It parses the flow logic correctly (trigger, actions, conditions) because the instructions are in English.
  • It keeps your business terms and user-facing text in Polish because you embed them as literal examples.

We’ll apply this pattern to the HR onboarding scenario.

Before: pure Polish prompt

Original prompt:

"Kiedy dodaję nowy wiersz do listy SharePoint z danymi pracownika, wyślij mu mail powitalny po polsku i dodaj zadanie w Plannerze dla działu HR."

Typical Copilot output:

  • Trigger: When an item is created (SharePoint) – correct
  • Action: Send an email (V2) – correct, but English body
  • Action: Create a task – often missing Polish field names in dynamic content

After: mixed-language structured prompt

Revised prompt:

"Create a cloud flow in Power Automate.

  1. Trigger: when a new item is created in a SharePoint list named OnboardingRequests.
  2. Use columns: Imie, Nazwisko, EmailPracownika, DataStartu.
  3. Send a welcome email in Polish to EmailPracownika. Example subject: Witamy w firmie. Example body in Polish: Cześć {Imie},\n\nWitamy na pokładzie. Twój pierwszy dzień pracy to {DataStartu}.\n\nPozdrawiamy, Zespół HR.
  4. Create a Planner task in the plan Onboarding and bucket Nowe osoby. Task title example (Polish): Onboarding dla {Imie} {Nazwisko}.
  5. Post a message in a Teams channel (Polish) confirming that onboarding has started."

Result:

  • Copilot keeps responding in English, but:
    • It correctly uses the specified SharePoint columns
    • It populates the email and task fields with your Polish examples
    • It wires dynamic content into the right places more reliably

Prompt patterns that work for Polish-heavy flows

Below are patterns you can reuse, with notes on why they help Copilot in the current product.

1. Explicitly name triggers and actions

Copilot is more accurate when you reference the actual connector names.

For the HR scenario:

"Use the SharePoint trigger When an item is created on the list OnboardingRequests. Then use the Office 365 Outlook action Send an email (V2). After that, use the Planner connector to create a task in the plan Onboarding."

Why this helps:

  • Copilot matches your English names to known connectors and actions
  • It avoids guessing based on Polish verbs like "wyślij" or "dodaj"

Use this pattern whenever:

  • You know the connector and action names already
  • You want to avoid Copilot picking the wrong email connector or trigger variant

2. Use structured steps instead of a single paragraph

Copilot handles numbered lists better than long narrative text.

Example for the onboarding flow:

"Create a Power Automate cloud flow with these steps:

  1. Trigger: SharePoint, When an item is created, list name OnboardingRequests.
  2. Action: Office 365 Outlook, Send an email (V2).
  3. Action: Planner, Create a task.
  4. Action: Microsoft Teams, Post a message in a channel."

Then follow up with a second prompt:

"Update the email and Teams message content to be in Polish. Use dynamic content from the SharePoint item: Imie, Nazwisko, EmailPracownika, DataStartu."

Why this helps:

  • First prompt gets the structure right
  • Second prompt focuses Copilot on content localisation

3. Provide Polish examples as literal text

Copilot often uses examples verbatim, even if they’re in Polish.

For the email body:

"Set the email subject to this Polish text: Witamy w firmie. Set the email body to this Polish text, using dynamic content placeholders:

Cześć {Imie},\n\nWitamy na pokładzie. Twój pierwszy dzień pracy to {DataStartu}.\n\nPozdrawiamy, Zespół HR."

After Copilot generates the flow, you manually replace {Imie} and {DataStartu} with the actual dynamic content tokens in the designer.

Why this helps:

  • Copilot doesn’t need to "understand" the Polish text; it just inserts it
  • You stay in control of the final wording

4. Ask Copilot for expressions, then localise labels yourself

For advanced flows, Copilot is useful for generating expressions, even if labels stay English.

Example requirement in the HR scenario: only send the email if DataStartu is in the future.

Prompt:

"Add a condition that checks if the SharePoint column DataStartu is greater than or equal to today's date. Use an expression with utcNow() and formatDateTime if needed."

Copilot may generate an expression like:

@greaterOrEquals(triggerOutputs()?['body/DataStartu'], utcNow())

(You’d then adjust the context to your actual flow if needed, for example when the value comes from an Apply to each.)

You can then manually rename the condition to a Polish label like Czy DataStartu jest w przyszłości?.

Why this helps:

  • Copilot is good at wiring expressions
  • You keep the user-facing labels in Polish by editing them yourself

5. Use follow-up prompts to fix misinterpretations

Copilot often gets the first attempt "close but not quite".

Common issues in Polish scenarios:

  • Wrong column mapped (e.g. using Title instead of Imie)
  • English email content
  • Missing condition logic

Use follow-up prompts like:

"In the email action, change the To field to use the SharePoint column EmailPracownika. Keep the body in Polish and include Imie and DataStartu as dynamic content."

or:

"Replace the English email body with this Polish text and keep the dynamic content: Cześć {Imie},\n\nTwój pierwszy dzień pracy to {DataStartu}.."

Then you manually fix any remaining placeholders in the designer.

Handling English responses inside a Polish workflow

Even with careful prompts, Copilot will:

  • Respond in English
  • Use English names for conditions, scopes, and comments

You can work around this without breaking the flow.

Rename everything that users see

After Copilot scaffolds the flow:

  1. Rename actions from English to Polish:
    • Send an email (V2) → Wyślij mail powitalny
    • Create a task → Utwórz zadanie w Plannerze
  2. Rename conditions and scopes:
    • Condition → Czy DataStartu jest w przyszłości?
  3. Update comments and notes in Polish for your team.

These changes are purely cosmetic; they don’t affect execution.

Keep technical artefacts in English where it helps maintenance

For teams that mix Polish and English speakers, a hybrid approach works:

  • Keep connector and action names in English when documenting the flow (e.g. in internal wiki)
  • Use Polish labels inside the flow designer for business clarity

This makes it easier to:

  • Match documentation and Microsoft Learn articles (English)
  • Onboard colleagues who learned Power Automate in English

When to avoid Copilot and build manually

There are cases where Copilot’s limited Polish support makes it more of a hindrance.

In the HR scenario, consider building manually when:

  • You need precise error handling with multiple branches and scopes
  • You rely on complex expressions involving locale-specific formatting (e.g. Polish date formats in email text)
  • Your flow has strict compliance requirements and you don’t want Copilot to auto-generate connectors or paths you didn’t intend

Practical rule of thumb:

  • Use Copilot to scaffold simple flows and common patterns
  • Switch to manual editing for:
    • Complex conditions
    • Nested Apply to each loops
    • Anything involving sensitive data where connector choice must be exact

In our onboarding example, Copilot is great for:

  • Creating the basic trigger and actions
  • Suggesting expressions for date comparison

Then HR and IT can:

  • Manually refine the email body
  • Add Polish-only details like links to internal portals
  • Adjust Planner task fields to match their naming conventions

Concrete takeaway for Polish teams using Copilot

If Copilot in Power Automate doesn’t speak Polish well, don’t fight it; talk to it in English about structure, and give it Polish as literal content.

For the HR onboarding flow, that means:

  • Describe triggers and actions in English
  • List your Polish SharePoint columns explicitly
  • Paste example Polish email bodies and Teams messages as text
  • Use follow-up prompts to correct mappings
  • Rename everything user-facing to Polish after Copilot scaffolds the flow

This mixed-language prompting pattern keeps Copilot useful without forcing your team to abandon Polish in day‑to‑day work.

Editor's Note

This article reflects how Polish data and operations teams are adapting Copilot in Power Automate to English-centric language models by separating logic prompts in English from business content in Polish, especially in mixed-language environments.

Professionals who want to apply these patterns to their own data can explore Excelgoodies' Power Automate Course programme - taught live by instructors, with certification awarded once a real project is running at work.

Insights compiled through ongoing industry research and discussions within the Excelgoodies Analytics Community.

Power Automate

New

Next Batches Now Live

Power BIPower BI
SQLSQL
Power AppsPower Apps
Power AutomatePower Automate
Microsoft FabricMicrosoft Fabrics
AzureAzure Data Engineering
Explore Dates & Reserve Your Spot → Reserve Your Spot →