An unavailable inbox can stop a Norwegian company from taking an order, reaching its employees or approving an urgent payment. If the contact list and the continuity plan sit behind the same login as the failed service, the problem spreads quickly. Management should decide which work must continue and test a way to do it before the next outage.
The Microsoft 365 disruption that began on 12 November 2026 affected services including Outlook, Teams and Copilot. On 14 November Microsoft said service had been restored to the remaining affected customers and regions. The Associated Press reported that some Norwegian government operations were affected. This does not mean every Norwegian customer experienced the same interruption.
Microsoft described a technical problem involving an earlier network capacity upgrade and a subsequent network change. It said there was no indication of a cyberattack and that the incident was not related to AI agents. The service update reproduced by a Japanese university identifies incident MO1550821 and says a full post-incident report would follow. That is a preliminary explanation, not a final independent investigation.
AI is relevant here for a different reason. An assistant that depends on the same files, identity system or online services as the rest of the business may be unavailable at the same time. The value of an AI tool should therefore be considered alongside the cost and feasibility of continuing the underlying work without it. This is a dependency to check, not a claim that AI caused the outage.
Start with one working day
Imagine a Norwegian wholesaler with ten employees. Orders arrive by email, delivery instructions are held in a shared cloud folder and staff use an AI assistant to summarise customer requests. This company is hypothetical. Its manager does not need a duplicate of every system immediately. The first decision is which orders and commitments cannot wait until tomorrow.
Agree on an outage plan for those few tasks:
- Contact: Maintain a current, access-controlled list of essential employees, suppliers and customers outside the affected platform. Agree on an alternate channel in advance and verify who receives messages. A second app is not independent if it needs the same unavailable login or device connection.
- Record: Define a simple way to note incoming orders, time, customer, delivery details and who accepted them. Use a controlled paper form or approved local file when suitable. Do not store more customer information than necessary.
- Decide: Name who can accept an order, change a deadline or approve a payment while normal systems are unavailable. Use an independent verification channel for payment changes and exceptional requests. If the facts cannot be verified, postpone the commitment.
- Reconcile: Assign someone to enter approved work into the regular system after recovery, check for duplicates and resolve conflicting versions. Keep the temporary record securely and remove it under the organisation's retention rules.
A retailer, consultancy and manufacturer will choose different priorities. Decide the maximum tolerable delay for each critical task and make the backup proportionate to its consequences. A small company may need a phone list and a small number of protected forms. A larger company may need tested alternative access, designated staff and supplier support.
A downloaded file is not a complete backup plan
Microsoft's guidance for working offline describes conditions for accessing and synchronising files on a mobile device. Its instructions differ by platform and require preparation. They do not establish that every file, permission, account or application will work during every cloud outage. Test the actual devices, documents and sign-in conditions your team relies on.
Offline copies create other obligations. Who updates them, who can read them, and how are they protected if a laptop or printout is lost? A copy that is six months old can be as misleading as no copy. Sensitive information needs suitable access controls and secure handling. Avoid automatically sending complete customer databases to personal email or consumer tools in a crisis.
A separate supplier can help for some important functions, but another subscription is not proof of resilience. It may share the same identity provider, network connection or personnel. Maintaining two systems also costs money and introduces reconciliation and security work. Test whether the alternative actually supports the task before paying for a permanent duplicate.
Practise the first hour
Run a short tabletop exercise with the people who do the work, the IT provider and anyone responsible for customer communication. Assume email, Teams and the normal cloud documents cannot be reached. Do not disable production systems for the exercise.
Ask someone to locate the contact list without the normal login. Give the team an urgent fictional order and ask how it will be received, approved and recorded. Then ask how the order will be checked and entered after service returns. Record each blocked step, the owner and a date to fix it. Where practical, run a controlled technical test later with IT and security approval.
Norway's National Security Authority recommends assessing dependencies and considering redundancy for important services. Its warning about dependence on foreign cloud providers is a reason to understand where control and recovery sit, not a claim that every company must leave the cloud.
Management owns the exercise, the supplier discussion and the cost of a workable alternative. The board should ask which critical tasks have a tested fallback, what interruption management considers acceptable and which unresolved gaps carry a material business consequence. It should receive the results and deadlines rather than manage the phone list itself.
The November outage is a useful prompt to test the first working day. Start with the order, customer or service your business cannot afford to lose, and prove that the people responsible can handle it without the normal tools. Connect this exercise to your broader business preparedness and your AI supplier exit plan.
Sources checked on 15 November 2026. The wholesaler and exercise are hypothetical; the recommendations are the author's analysis. The illustration is AI-generated.

