All articles

  1. Governance & security

    Automatic external forwarding in Exchange Online: three controls, and one value that tells you nothing

    Whether mail can leave a Microsoft 365 organisation by automatic forwarding is decided by three separate controls, and the default value of the first one means On in some tenants and Off in others. How to read each control with PowerShell, how to find the mailboxes that forward today, and how to block forwarding with named exceptions.

    ·7 min read

    Keep reading →
  2. Governance & security

    Documentation is a claim until the system is read against it

    A configuration document governs a Microsoft 365 or Azure estate only while it is bound to the system it describes: an owner, a documented reading of the actual state, a comparison and a date to read it again. What NIST SP 800-53 and Microsoft's guidance ask of documentation, how Compliance Manager keeps a recorded status apart from a read one, and three Microsoft 365 settings where the document and the tenant can disagree unnoticed.

    ·12 min read

    Keep reading →
  3. Governance & security

    The last Graph page is not proof of tenant coverage

    A successful response, an exhausted page chain and a complete population are three different claims. A small offline experiment shows where an apparently reliable collector can silently change one into another.

    ·4 min read

    Keep reading →
  4. Governance & security

    SharePoint HTML pages are governed by a setting your custom script audit does not read

    HTML pages run JavaScript inside a sandbox, are on by default, and are switched off site by site by each site owner. Microsoft states there is no tenant-level restriction, and the custom script setting is not part of the answer. What the sandbox stops, what it leaves to governance, and what to measure before general availability reaches your tenant.

    ·9 min read

    Keep reading →
  5. Governance & security

    A report can look complete and still be impossible to verify

    A report that looks complete and a report that can be verified are different objects. The three things a report has to carry so that its reader can check it without its producer, and the specimen we publish that carries them.

    ·2 min read

    Keep reading →
  6. Governance & security

    The confident wrong answer is worse than missing evidence

    A governance report told us a site had external sharing disabled. The site permitted anonymous links. Nothing was broken: a correct API returned a default value for a property it had never populated, and the default happened to be the safest word in the enum. Here is the mechanism, the code that reproduces it, Microsoft's own documentation of it, and the discipline that catches the whole class.

    ·7 min read

    Keep reading →
  7. Migrations

    SharePoint 2016, 2019, Subscription Edition and Online: what actually changes

    2016 and 2019 both left support on 14 July 2026. Subscription Edition has no end date and cannot be bought outright. Online has features the others do not have, and Subscription Edition has two that are no longer a simple no. A practical comparison, with the licensing that decides it.

    ·21 min read

    Keep reading →
  8. Development & automation

    We are still measuring engineers by the part that got automated.

    In 2025 DORA threw out its own scoreboard. The low, medium, high and elite tiers that the industry had quoted for a decade were replaced by seven team archetypes across eight measures. That is not a methodology footnote. It is the most used measurement framework in software admitting that the instruments no longer read the thing that decides whether a system survives.

    ·9 min read

    Keep reading →
  9. AI & agents

    One agent is an integration. A hundred is a governance problem.

    Agents got production identities before they got a stable way to record what they did. Microsoft Entra Agent ID is generally available, with sponsors and expiry dates. The OpenTelemetry conventions for agent traces are still experimental. That gap is not a detail: it decides what you can responsibly put into production this year, and it turns an AI question into a governance one.

    ·10 min read

    Keep reading →
  10. AI & agents

    Agents decide. Code executes. Knowing the line is the job.

    The promise is that agents replace applications. In the systems we build, they do not. An agent is very good at understanding a request, weighing options and choosing a tool. It is the wrong place for a payment, a tax rule or a permission check. This is where the line falls, and why putting it in the wrong place is expensive.

    ·7 min read

    Keep reading →
  11. AI & agents

    AI solved syntax. It did not solve judgement.

    Three independent measurements, taken between 2025 and 2026, all point the same way: AI-generated code passes over 95% on syntax and 55% on security, and that second number has not moved in two years. The part you write got cheap. The part you decide did not.

    ·4 min read

    Keep reading →
  12. AI & agents

    What to put in order before switching Copilot on

    Copilot does not create permission problems. It makes the ones already there findable, and you no longer need to know what you are looking for.

    ·2 min read

    Keep reading →

Categories

Send us a message

or write to hello@ph7x.com