- Feature
- Manage agent permissions easier
- Availability
- Generally available (October 2026)
- Industry focus
- Pharmaceutical and medical device distribution
Distributors of medicines, medical devices and laboratory supplies live under constant scrutiny. Regulators, auditors and customers all expect them to prove that only the right people can change the right data. As AI agents start working inside Business Central, the same expectation applies to them. Business Central 29.0 makes it easier to give agents exactly the access they need, at the moment they need it.
The industry challenge
Life sciences distribution combines ordinary wholesale activity with strict rules:
- Good Distribution Practice (GDP) and similar frameworks require documented controls over who can change records
- Lot and expiry data must not be edited casually
- Customer licences, such as a pharmacy or hospital authorization, must be checked before goods ship
- Audit trails must show who did what, and when
AI agents promise real efficiency in this setting. They can draft sales orders from customer emails, process supplier invoices and chase routine queries. But every agent needs permissions to do its work, and that creates a dilemma.
If you give an agent broad permissions up front, you weaken your control environment and create an audit finding waiting to happen. If you give it narrow permissions, it gets stuck the first time a task needs one more table or page. Then someone has to work out what is missing, usually by trial and error.
What Microsoft delivered in Business Central 29.0
Microsoft describes this release simply: administrators can assign additional permissions to agents when the agents request them.
In practice, this changes the default posture for agent security. Instead of guessing every permission an agent might ever need, you can start the agent with a tight set of permissions. When it meets a task it can't complete, the need comes to the surface as a request, and an authorized person decides whether to grant it.
This builds on how agents already work in Business Central. They run under their own identity, operate within the permissions assigned to them and have their actions logged and attributed to that identity.
Where the value shows up
| Business area | Impact for a regulated distributor |
|---|---|
| Control environment | Agents start with minimum access, and access grows only through explicit decisions |
| Audit readiness | Permission changes for agents become deliberate events, not set-up guesswork |
| Time to value | Less trial and error when configuring an agent for a new process |
| Risk management | Sensitive areas such as lot data or licences stay closed unless someone decides otherwise |
| Governance | Ownership of agent access sits with named administrators |
The central idea is that least privilege becomes practical. Most organizations agree with least privilege in principle. They abandon it in practice because tight permissions break processes. A request-and-grant model removes that trade-off.
A worked scenario
Consider a distributor of medical consumables that serves clinics and hospital pharmacies (an illustrative example). It activates an agent to help capture sales orders from customer emails. The quality manager insists that the agent must not be able to change item tracking or customer licence data.
The administrator configures the agent with permissions limited to reading customers and items and creating sales quotes and orders. A few days later, a customer asks for delivery to a new ship-to address. The agent needs access it doesn't have. Instead of failing silently or being given broad access by default, it raises a request.
The administrator reviews the request and agrees that maintaining ship-to addresses is in scope. They grant that permission and nothing more. The quality manager can see a clear, explainable history of how the agent's access was shaped.
Adoption considerations
- Start narrow on purpose. Use this feature as permission to be strict at the start. The agent will tell you what it needs.
- Name the approvers. Decide who can grant agent permissions. In regulated businesses, this is usually IT together with quality or compliance.
- Record the reason. Keep a short note of why each extra permission was granted. That makes audits and periodic access reviews quicker.
- Review agent access regularly. Include agents in the same quarterly access review you run for people.
- Ring-fence the sensitive areas. Agree up front which data, such as lots, expiry dates and licences, should never be granted to a given agent.
The bottom line
For pharmaceutical and medical device distributors, the question about AI agents isn't whether they're useful. It's whether they can be controlled well enough to satisfy a regulator. On-request agent permissions in Business Central 29.0 make the answer easier. Agents start small, grow only through deliberate decisions, and leave a story that an auditor can follow.
If you're planning to introduce agents in a regulated environment, it's worth designing your agent permission model with an experienced Business Central adviser before going live.
Source: Microsoft roadmap 573370