---
id: video/NE7NIjpkX3c
type: video
title: "What's New: Prepare Your Copilot Extension - Practical Considerations (2025 release wave 1)"
summary: "Practical guidance for partners building Business Central Copilot extensions in 2025 release wave 1: running the extension in the Microsoft cloud from a latest sandbox environment, GPT-4o and GPT-4o mini access, testing prompts against model version changes, handling fabrications by validating responses, grounding, and keeping to about five tools."
tier: official
language: en
tags:
  - copilot extension
  - cloud deployment
  - large language models
  - prompt engineering
  - model versions
  - hallucinations
  - fabrications
  - test automation
  - validation
  - responsible ai
  - function calling
  - grounding
system: copilot
review:
  state: reviewed
  by: opus
  at: "2026-10-07T22:59:25.993Z"
  flags: []
generated:
  at: "2026-10-07T22:59:26.044Z"
  pipeline: 0.2.0
  prompts:
    extract-video: 1
    summarize-video: 2
  input_hash: 67efa783f7295cde11c8bd14f66fb8f0f8e8f68961c28efa90f8e686106bced0
evidence:
  - kind: video
    url: https://www.youtube.com/watch?v=NE7NIjpkX3c&t=5s
    title: "What's New: Prepare Your Copilot Extension - Practical Considerations (2025 release wave 1)"
    date: "2025-04-01T15:00:34.000Z"
    commit: null
    t: 5
    quote: in this release we allow you to bring your capalot extension and finally run it in the Microsoft cloud based on our Microsoft II
  - kind: video
    url: https://www.youtube.com/watch?v=NE7NIjpkX3c&t=86s
    title: "What's New: Prepare Your Copilot Extension - Practical Considerations (2025 release wave 1)"
    date: "2025-04-01T15:00:34.000Z"
    commit: null
    t: 86
    quote: finally you and us have a same framework and infrastructure how do we build our compal experiences
  - kind: video
    url: https://www.youtube.com/watch?v=NE7NIjpkX3c&t=138s
    title: "What's New: Prepare Your Copilot Extension - Practical Considerations (2025 release wave 1)"
    date: "2025-04-01T15:00:34.000Z"
    commit: null
    t: 138
    quote: as a part of our def tools today we give you access to GPT for or and GPT for all mini
  - kind: video
    url: https://www.youtube.com/watch?v=NE7NIjpkX3c&t=219s
    title: "What's New: Prepare Your Copilot Extension - Practical Considerations (2025 release wave 1)"
    date: "2025-04-01T15:00:34.000Z"
    commit: null
    t: 219
    quote: before we as a business Central can give you access to those models we need to make sure that all the world where we
  - kind: video
    url: https://www.youtube.com/watch?v=NE7NIjpkX3c&t=528s
    title: "What's New: Prepare Your Copilot Extension - Practical Considerations (2025 release wave 1)"
    date: "2025-04-01T15:00:34.000Z"
    commit: null
    t: 528
    quote: new versions can be faster or sometimes cheaper and so on so forth so and because every model is not better is different sometimes
  - kind: video
    url: https://www.youtube.com/watch?v=NE7NIjpkX3c&t=601s
    title: "What's New: Prepare Your Copilot Extension - Practical Considerations (2025 release wave 1)"
    date: "2025-04-01T15:00:34.000Z"
    commit: null
    t: 601
    quote: it's much much more important for practical considerations to make sure your prompts will continue to work as we're going to update models
  - kind: video
    url: https://www.youtube.com/watch?v=NE7NIjpkX3c&t=721s
    title: "What's New: Prepare Your Copilot Extension - Practical Considerations (2025 release wave 1)"
    date: "2025-04-01T15:00:34.000Z"
    commit: null
    t: 721
    quote: large language models just copy those uh type of response if you will because it's trying to be helpful but it definitely lead to
  - kind: video
    url: https://www.youtube.com/watch?v=NE7NIjpkX3c&t=843s
    title: "What's New: Prepare Your Copilot Extension - Practical Considerations (2025 release wave 1)"
    date: "2025-04-01T15:00:34.000Z"
    commit: null
    t: 843
    quote: large language models always try to predict next words or talking and it always try to be helpful and finish the sentence and instead
  - kind: video
    url: https://www.youtube.com/watch?v=NE7NIjpkX3c&t=883s
    title: "What's New: Prepare Your Copilot Extension - Practical Considerations (2025 release wave 1)"
    date: "2025-04-01T15:00:34.000Z"
    commit: null
    t: 883
    quote: always validate l l m response
  - kind: video
    url: https://www.youtube.com/watch?v=NE7NIjpkX3c&t=944s
    title: "What's New: Prepare Your Copilot Extension - Practical Considerations (2025 release wave 1)"
    date: "2025-04-01T15:00:34.000Z"
    commit: null
    t: 944
    quote: whenever we never surface any llm response from a system to the user unless we verify they're grounded and they true
  - kind: video
    url: https://www.youtube.com/watch?v=NE7NIjpkX3c&t=964s
    title: "What's New: Prepare Your Copilot Extension - Practical Considerations (2025 release wave 1)"
    date: "2025-04-01T15:00:34.000Z"
    commit: null
    t: 964
    quote: you might use 40 mini as a cheaper faster model to validate its responses and this is like a very simple pattern
  - kind: video
    url: https://www.youtube.com/watch?v=NE7NIjpkX3c&t=1065s
    title: "What's New: Prepare Your Copilot Extension - Practical Considerations (2025 release wave 1)"
    date: "2025-04-01T15:00:34.000Z"
    commit: null
    t: 1065
    quote: having that small technique just to validate llm for this particular feature allow us to increase accuracy meaning it works from 6
  - kind: video
    url: https://www.youtube.com/watch?v=NE7NIjpkX3c&t=1321s
    title: "What's New: Prepare Your Copilot Extension - Practical Considerations (2025 release wave 1)"
    date: "2025-04-01T15:00:34.000Z"
    commit: null
    t: 1321
    quote: exactly same approach we can use in business Central to make sure we can answer the question about business data in a trustful and
  - kind: video
    url: https://www.youtube.com/watch?v=NE7NIjpkX3c&t=1523s
    title: "What's New: Prepare Your Copilot Extension - Practical Considerations (2025 release wave 1)"
    date: "2025-04-01T15:00:34.000Z"
    commit: null
    t: 1523
    quote: tools are awesome it's a best tool for a job to retrieve data from a system to enrich your responses to make sure it's
  - kind: video
    url: https://www.youtube.com/watch?v=NE7NIjpkX3c&t=1544s
    title: "What's New: Prepare Your Copilot Extension - Practical Considerations (2025 release wave 1)"
    date: "2025-04-01T15:00:34.000Z"
    commit: null
    t: 1544
    quote: if you look at Microsoft features we use around five fiveish and we see pretty pretty good U results so we would suggest you
  - kind: video
    url: https://www.youtube.com/watch?v=NE7NIjpkX3c&t=1618s
    title: "What's New: Prepare Your Copilot Extension - Practical Considerations (2025 release wave 1)"
    date: "2025-04-01T15:00:34.000Z"
    commit: null
    t: 1618
    quote: if you want to build like a gut capil which can help you with everything uh you probably don't have enough technology for now
links:
  learn: []
  objects: []
  features: []
  topics: []
  localizations: []
  videos: []
  posts: []
  guidelines: []
video_id: NE7NIjpkX3c
channel: yt-microsoft
source_name: Microsoft Dynamics 365 Business Central (YouTube)
url: https://www.youtube.com/watch?v=NE7NIjpkX3c
published_at: "2025-04-01T15:00:34.000Z"
duration_s: 1674
captions: full
audience:
  - developer
  - partner
  - functional consultant
chapters:
  - t: 0
    title: Copilot Extension in Microsoft Cloud
  - t: 138
    title: Available Models and Access
  - t: 257
    title: Prompt Engineering and Quality
  - t: 421
    title: Model Versions and Testing
  - t: 601
    title: Testing and Model Upgrades
  - t: 660
    title: Hallucinations and Fabrications
  - t: 782
    title: LLM Response Validation
  - t: 1004
    title: Validation in Marketing Text and Documentation
  - t: 1106
    title: Grounding Copilots in Data
  - t: 1219
    title: Tools for Copilot Extensions
  - t: 1462
    title: Optimal Tool Quantity and Best Practices
  - t: 1584
    title: Recommendations for Copilot Extension Development
features:
  - name: Copilot Extension Cloud Deployment
    status: unclear
    t: 5
    verified: false
    status_source: video
  - name: Shared Infrastructure for Copilot
    status: unclear
    t: 86
    verified: false
    status_source: video
  - name: GPT-4o and GPT-4o Mini Model Access
    status: unclear
    t: 138
    verified: false
    status_source: video
  - name: Azure OpenAI Abstraction Layer
    status: unclear
    t: 178
    verified: false
    status_source: video
  - name: Additional Models Coming When Capacity Available
    status: unclear
    t: 198
    verified: false
    status_source: video
  - name: Prompt Engineering Guidelines
    status: unclear
    t: 257
    verified: false
    status_source: video
  - name: Model Version Variations
    status: unclear
    t: 421
    verified: false
    status_source: video
  - name: Test Automation for Prompt Validation
    status: unclear
    t: 560
    verified: false
    status_source: video
  - name: Hallucinations and Fabrications Management
    status: unclear
    t: 660
    verified: false
    status_source: video
  - name: LLM Response Validation
    status: unclear
    t: 863
    verified: false
    status_source: video
  - name: Marketing Text Generation with Validation
    status: unclear
    t: 1004
    verified: false
    status_source: video
  - name: Copilot Documentation Chat with Grounding
    status: unclear
    t: 1106
    verified: false
    status_source: video
  - name: Tools for Copilot Extensions
    status: unclear
    t: 1219
    verified: false
    status_source: video
  - name: Function Calling for Copilot
    status: unclear
    t: 1462
    verified: false
    status_source: video
objects_mentioned:
  - other Sales Order Page
  - other Production Order
  - other Credit Memo
  - other Sales Invoice
  - other Purchase Order
  - other code unit
quotes:
  - t: 5
    text: in this release we allow you to bring your capalot extension and finally run it in the Microsoft cloud based on our Microsoft II
    check: exact
  - t: 86
    text: finally you and us have a same framework and infrastructure how do we build our compal experiences
    check: exact
  - t: 138
    text: as a part of our def tools today we give you access to GPT for or and GPT for all mini
    check: snapped
  - t: 219
    text: before we as a business Central can give you access to those models we need to make sure that all the world where we
    check: exact
  - t: 528
    text: new versions can be faster or sometimes cheaper and so on so forth so and because every model is not better is different sometimes
    check: exact
  - t: 601
    text: it's much much more important for practical considerations to make sure your prompts will continue to work as we're going to update models
    check: exact
  - t: 721
    text: large language models just copy those uh type of response if you will because it's trying to be helpful but it definitely lead to
    check: snapped
  - t: 843
    text: large language models always try to predict next words or talking and it always try to be helpful and finish the sentence and instead
    check: exact
  - t: 883
    text: always validate l l m response
    check: exact
  - t: 944
    text: whenever we never surface any llm response from a system to the user unless we verify they're grounded and they true
    check: exact
  - t: 964
    text: you might use 40 mini as a cheaper faster model to validate its responses and this is like a very simple pattern
    check: exact
  - t: 1065
    text: having that small technique just to validate llm for this particular feature allow us to increase accuracy meaning it works from 6
    check: fuzzy
  - t: 1321
    text: exactly same approach we can use in business Central to make sure we can answer the question about business data in a trustful and
    check: exact
  - t: 1523
    text: tools are awesome it's a best tool for a job to retrieve data from a system to enrich your responses to make sure it's
    check: snapped
  - t: 1544
    text: if you look at Microsoft features we use around five fiveish and we see pretty pretty good U results so we would suggest you
    check: snapped
  - t: 1618
    text: if you want to build like a gut capil which can help you with everything uh you probably don't have enough technology for now
    check: exact
---

# What's New: Prepare Your Copilot Extension - Practical Considerations (2025 release wave 1)

> Practical guidance for partners building Business Central Copilot extensions in 2025 release wave 1: running the extension in the Microsoft cloud from a latest sandbox environment, GPT-4o and GPT-4o mini access, testing prompts against model version changes, handling fabrications by validating responses, grounding, and keeping to about five tools.

[Watch on YouTube](https://www.youtube.com/watch?v=NE7NIjpkX3c) · Microsoft Dynamics 365 Business Central (YouTube) · 2025-04-01 · 27:54 · tier official · reviewed (checked by Opus)

## Overview

The session explains that partners can upload a Copilot extension to a Business Central sandbox and run it in the Microsoft cloud on Microsoft AI resources, without managing LLM models, deployment or availability. Partners now use the same framework and infrastructure as Microsoft. GPT-4o and GPT-4o mini are available through an abstraction layer. Models such as Deep Seek R1 and other reasoning models are not yet available in Business Central, because capacity is needed in all Azure regions first.

The rest covers practical considerations. Model versions change behavior and are not necessarily backward compatible, so test automation should start from day zero. Fabrications are called the number one priority. The video shows validating LLM responses with a marketing text example (accuracy raised from 6%) and a documentation chat grounded in Microsoft Learn. It also covers tools defined in AL for grounding, with a recommended limit of around five.

## Key points

- Partners can upload a Copilot extension to a latest sandbox environment and run it in the Microsoft cloud on Microsoft AI resources, with no LLM deployment or availability to manage.
- GPT-4o and GPT-4o mini are available through an Azure OpenAI abstraction layer, so partners do not manage model versions themselves.
- Deep Seek R1 and other reasoning models are not yet in Business Central. They come only after capacity exists in all Azure regions where customers are, and no timeline was given.
- Model versions can change behavior overnight and are not necessarily backward compatible. Write test automation from day zero to catch prompt regressions, which matters more than prompt styling.
- Fabrications come from models predicting likely text. Validate LLM responses, for example with GPT-4o mini as a cheaper validator, and do not show them to users unless they are verified as grounded.
- Validation added to the marketing text feature raised accuracy from 6% to very high numbers. The documentation chat checks grounding against Microsoft Learn articles, which may add about 200 ms of latency.
- Define tools in AL to pull real Business Central data into responses. Keep to around five tools, since too many reduce reliability.

## Chapters

- [0:00](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=0s) Copilot Extension in Microsoft Cloud
- [2:18](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=138s) Available Models and Access
- [4:17](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=257s) Prompt Engineering and Quality
- [7:01](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=421s) Model Versions and Testing
- [10:01](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=601s) Testing and Model Upgrades
- [11:00](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=660s) Hallucinations and Fabrications
- [13:02](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=782s) LLM Response Validation
- [16:44](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=1004s) Validation in Marketing Text and Documentation
- [18:26](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=1106s) Grounding Copilots in Data
- [20:19](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=1219s) Tools for Copilot Extensions
- [24:22](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=1462s) Optimal Tool Quantity and Best Practices
- [26:24](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=1584s) Recommendations for Copilot Extension Development

## Features

| Feature | Status | At |
|---|---|---|
| Copilot Extension Cloud Deployment | status not stated | [0:05](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=5s) |
| Shared Infrastructure for Copilot | status not stated | [1:26](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=86s) |
| GPT-4o and GPT-4o Mini Model Access | status not stated | [2:18](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=138s) |
| Azure OpenAI Abstraction Layer | status not stated | [2:58](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=178s) |
| Additional Models Coming When Capacity Available | status not stated | [3:18](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=198s) |
| Prompt Engineering Guidelines | status not stated, demoed | [4:17](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=257s) |
| Model Version Variations | status not stated, demoed | [7:01](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=421s) |
| Test Automation for Prompt Validation | status not stated | [9:20](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=560s) |
| Hallucinations and Fabrications Management | status not stated, demoed | [11:00](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=660s) |
| LLM Response Validation | status not stated | [14:23](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=863s) |
| Marketing Text Generation with Validation | status not stated | [16:44](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=1004s) |
| Copilot Documentation Chat with Grounding | status not stated | [18:26](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=1106s) |
| Tools for Copilot Extensions | status not stated, demoed | [20:19](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=1219s) |
| Function Calling for Copilot | status not stated | [24:22](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=1462s) |

## AL objects mentioned

As heard in the captions. A name that matches one object page by exact type and name links to it; the others stay as named.

- other "Sales Order Page" at [12:42](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=762s)
- other "Production Order" at [11:20](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=680s)
- other "Credit Memo" at [11:20](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=680s)
- other "Sales Invoice" at [12:01](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=721s)
- other "Purchase Order" at [12:01](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=721s)
- other "code unit" at [22:41](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=1361s)

## Quotes

- [0:05](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=5s) "in this release we allow you to bring your capalot extension and finally run it in the Microsoft cloud based on our Microsoft II"
- [1:26](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=86s) "finally you and us have a same framework and infrastructure how do we build our compal experiences"
- [2:18](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=138s) "as a part of our def tools today we give you access to GPT for or and GPT for all mini"
- [3:39](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=219s) "before we as a business Central can give you access to those models we need to make sure that all the world where we"
- [8:48](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=528s) "new versions can be faster or sometimes cheaper and so on so forth so and because every model is not better is different sometimes"
- [10:01](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=601s) "it's much much more important for practical considerations to make sure your prompts will continue to work as we're going to update models"
- [12:01](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=721s) "large language models just copy those uh type of response if you will because it's trying to be helpful but it definitely lead to"
- [14:03](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=843s) "large language models always try to predict next words or talking and it always try to be helpful and finish the sentence and instead"
- [14:43](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=883s) "always validate l l m response"
- [15:44](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=944s) "whenever we never surface any llm response from a system to the user unless we verify they're grounded and they true"
- [16:04](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=964s) "you might use 40 mini as a cheaper faster model to validate its responses and this is like a very simple pattern"
- [17:45](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=1065s) "having that small technique just to validate llm for this particular feature allow us to increase accuracy meaning it works from 6"
- [22:01](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=1321s) "exactly same approach we can use in business Central to make sure we can answer the question about business data in a trustful and"
- [25:23](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=1523s) "tools are awesome it's a best tool for a job to retrieve data from a system to enrich your responses to make sure it's"
- [25:44](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=1544s) "if you look at Microsoft features we use around five fiveish and we see pretty pretty good U results so we would suggest you"
- [26:58](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=1618s) "if you want to build like a gut capil which can help you with everything uh you probably don't have enough technology for now"

## Disclaimers in the video

- [0:25](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=25s) other: in this release we finally give you ability to run your extension in our Microsoft cloud
- [3:39](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=219s) coming-later: Deep Seek R1 reasoning Model and other reasoning models we need capacity before we can give you access to those models
- [19:48](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=1188s) other: dealing with Fabrications should be a number one priority as you consider to bring copilot to your users just to avoid lack of trust frustration bad reputation brand damage
- [27:12](https://www.youtube.com/watch?v=NE7NIjpkX3c&t=1632s) other: over the last releases and this release we have a number of great great videos how to extend copilot in business Central and how to build your own copilot