Skip to main content
Technology

Microsoft Dynamics 365 AI Integration: Beyond Copilot

Four routes to AI in Dynamics 365: Microsoft's Copilot, low-code, a custom-built layer or a bought product. Requirements, risks and a safe first step.

Author: Agent360 editorial team
Published: 27 April 2026
Last updated: 7 November 2026
Reading time: 9 min read

How do you integrate AI with Microsoft Dynamics 365?

There are four routes, and they can be combined: using Microsoft's own Copilot, building it yourselves with Microsoft's low-code tools, having someone build a tailored layer on top of the system, or buying a finished product. Which one fits depends on how particular your workflows are, and how much you want to own yourselves.

Dynamics 365 is built on Microsoft's data platform, Dataverse, and has an open interface that other programmes can read from and write to. That is what makes it possible to connect AI without replacing the customer register.

TL;DR:

  • Microsoft's own Copilot is the easiest start if your tasks are email, meetings and notes.
  • Low-code (Power Automate and similar) is a good choice for simple, repeated flows, if you have an internal super-user who will own them.
  • A custom-built layer makes sense when the task is particular to you, and when it must fit your data model exactly.
  • Whichever route: the AI must work within the permissions and rules Dynamics already has, and start by proposing, not changing.

What does Microsoft's own Copilot typically cover?

Copilot for Sales works in Outlook and Teams and can be connected to Dynamics 365. Typical tasks are suggested email replies, background before a meeting, notes from Teams meetings and suggested updates to the customer register. It is useful and requires no development of your own. Product names and licences change, so check Microsoft's pages for what is current. Read more in Microsoft Copilot for Sales: What It Does, and What to Ask Before You Buy.

Where standard tools typically stop is with the workflows that are particular to you: your own fields, your own rules for when a deal moves a phase forward, or conversations that take place outside Teams.

The four routes to AI in Dynamics 365

Route Good for Requires Be aware of
1. Microsoft's own Copilot Email, meetings and notes in Outlook and Teams Licences and some training Check its depth for your workflows and channels
2. Low-code in Power Platform (Power Automate, Copilot Studio and others) Simple, repeated flows, for example moving data or sending a reminder An internal super-user who will own the flows Flows can become messy if nobody has an overview
3. Custom-built layer Tasks that are particular to you, and which must fit your data model Someone who can build and maintain it Requires a clear task description and approval rules
4. Bought specialist product Defined needs where a product already exists Assessment of supplier, price and data Rarely fits perfectly, and data leaves your hands

Most will end up with a combination: Copilot for the general, low-code for simple flows, and a custom-built layer for what makes your business distinctive.

What does it technically take to connect AI to Dynamics 365?

Without going into detail, these are typically the building blocks:

  • Access to data. Dynamics 365 can be read from and written to via Dataverse and its API. Access is set up as a technical user with specific permissions, and it should only have access to what the task requires.
  • Permissions. Role and field permissions in Dynamics must also apply to the AI layer. If a salesperson may not see a customer, the AI must not show it to them.
  • Test environment. Dynamics has separate environments. Build and test in a test environment before anything points at the real data.
  • Traceability. It must be possible to see what the AI has proposed or changed, and who approved it.
  • Error handling. What happens if the AI is in doubt or the information is missing? The answer must be that it asks, not that it guesses.

Which tasks give the most value?

The best first tasks have two qualities: they are repeated, and a mistake is easy to spot before it does harm.

  • After the meeting: Notes and suggestions for fields and next steps, which the salesperson approves in Dynamics 365.
  • Email sorting: Incoming enquiries are sorted and given a draft reply. The rest is sent to the right colleague.
  • Quote drafts: A draft based on your own rates and templates. A person sets the price.
  • Clean-up: Duplicates and missing fields are gathered for review, without anything being deleted by itself.
  • Overview: Open tasks, quotes and the latest contact, gathered with source and time.

As a first task, avoid anything where the AI changes important fields such as deal size or phase without approval.

An example: from meeting to updated customer register

This is what a simple, safe set-up can look like for the task "after the meeting":

  1. The meeting ends. There is a transcript or a summary, either from Teams or from the salesperson's own notes.
  2. The AI reads the summary and finds what normally goes into the customer register: who took part, what was agreed, what the next step is, and whether a time or a competitor was mentioned.
  3. It prepares a proposal, not a change. The salesperson sees: "Proposed next step: send quote before Friday. Proposed contact: Jens, purchasing."
  4. The salesperson approves or corrects with one click, and only then is it written to Dynamics 365.
  5. If the AI is in doubt, it marks this rather than guessing: "Not clear whether the amount includes VAT."

The benefit is that the quality of the customer register rises without the salesperson having to write everything, and that nobody discovers an error too late.

What must be in place in Dynamics before you start?

The AI cannot fix a messy system. Check these five things first:

  • Few and clear fields. Each field has an owner and a meaning that everyone agrees on.
  • Phase definitions that salespeople understand in the same way.
  • Reasonably clean data. Duplicates and outdated contacts spread.
  • Permissions that fit. It must be clear who may see and change what.
  • One person who owns the system. Without an owner it is impossible to change anything without breaking something else.

How do you ensure it is lawful and safe?

  • Data location and data processing agreements. Ask where data is processed, and get a data processing agreement with every supplier that touches personal data.
  • Transparency. If the AI writes to customers, you must be able to explain it.
  • Keep only what is necessary. Do not keep whole conversations if a short summary is enough.
  • Recording conversations requires a lawful basis and information to those who are recorded. See the Danish Data Protection Agency.

Have a lawyer or data protection adviser assess your specific set-up.

How do you get started safely?

  1. Choose one task. The one that takes the most time, and where a mistake is cheap.
  2. Write down what the AI may do on its own, and what it must ask about. Start strictly.
  3. Build in the test environment, and let 2 to 3 salespeople try it.
  4. Measure what matters: time per task, quality of the fields, how much the salespeople correct.
  5. Expand only once it works. Not before.

What can Agent360 do?

Agent360 builds AI solutions inside your own systems. We call it "Your AI department": a task you solve by hand today is built on top of the systems you already use, with you as the approver. We contact your suppliers and help with access, and you are copied throughout. The price follows the task and is agreed before we start.

If your customer register is Dynamics 365, we look at what the system allows, and say honestly whether the task can be built and what it requires of you.

Examples from use cases that suit a CRM task:

If you want to learn to get AI working in your everyday work yourselves, there is a personal AI mentoring programme.

Frequently asked questions

Q: Can AI be connected to Dynamics 365 without changing the customer register? Yes. Dynamics 365 has an open interface, so an extra layer can read and write data without replacing the system.

Q: Should I use Copilot or build my own? Start by finding out whether Copilot covers your tasks. If the workflows are particular to you, or if something happens outside the Microsoft tools, a custom-built layer may be better.

Q: How long does an integration take? It depends on the task and on how tidy your Dynamics environment is. Decide the timeline after a short mapping, not before.

Q: Can the AI change data in Dynamics by itself? It should not at the start. Let it propose, and let the salesperson approve. Only give it more freedom once you have seen that it gets things right.

Q: What about Danish language? Test with your own Danish conversations and emails. Quality varies between tools and cannot be judged from a demo.

Q: Who should own the solution afterwards? Agree that before you start. If nobody owns it, it becomes outdated. A mentoring programme can give an employee the skills to maintain it.

Conclusion

AI in Dynamics 365 does not have to mean a new customer register. Start with one task, keep the AI on a short leash, and choose the route that fits the task, not the hype.

Book 20 minutes of free consultation with Agent360 and tell us what takes up time in your customer register. We will also tell you if the task is not worth building.

Also read: Teams integration and sales meeting data and CRM automation.

Sources

Should your CRM keep itself up to date?

We build the layer that moves data into the CRM you already use, without anyone having to type it in. We look at your current set-up first.

How we built it · CRM & customer overview

Customer CRM that updates itself

An internal page per customer with contacts, the agreement, the projects and a timeline where emails, meetings, email drafts and tasks land by themselves. Above the pages sits the board with all customers, owner and latest contact, and the system suggests next steps based on fixed rules.

  1. Email, meeting, task
  2. Lands on the customer’s page
  3. You take the next step
See the customer’s own page →