Field Notes · Editorial draft · September 17, 2026

The platform beneath the promise.

AI can help you do a task. A working platform helps that task become a dependable part of how you operate.

By Govind Davis · Founder & Publisher, Strattegys

There is a moment when a promising experiment becomes an operating question. The first run worked. Now who starts the next one? Where do the results live? Which systems may it touch? And who notices when something goes wrong?

Those questions matter whether you lead IT, manage a finance function, build a company, or evaluate technology for a team. They are also where the conversation about AI needs to become a conversation about work.

A model, a harness, and somewhere to operate.

The model supplies capabilities such as interpretation and generation. The harness coordinates instructions, context, tools, and the steps of execution. The platform provides the operating environment: runtime, data, identity, triggers, visibility, and recovery.

These are responsibilities, not rigid product categories. A managed service may supply all three. A custom system may assemble them from several components. What matters is knowing who is responsible for each part.

Why Docker makes this concrete.

In my own work, Docker is a practical part of the platform. It gives services a defined runtime, and Compose lets me describe how multiple services, networks, and volumes fit together. That makes the system more explicit and easier to reason about. Docker Compose documentation.

Persistent volumes keep application data outside a container’s disposable writable layer. That is an important distinction when yesterday’s result needs to survive tomorrow’s container replacement. A persistent volume still needs an appropriate backup and recovery plan. Docker’s volume guide.

Container packaging alone does not supply a complete operating system for automation. Scheduling, secrets, access policy, monitoring, review, and recovery still need owners and implementation. Docker’s production guidance also calls for environment-specific configuration and restart policies. Compose in production.

Make the work observable.

Take a recurring research brief. It reads a defined set of sources, prepares a summary, stores the result, and sends it to a person for review. Before relying on that workflow, ask:

1. Trigger: What starts it, and what prevents the same run happening twice?
2. Access: Which sources can it read, and where is it allowed to write?
3. State: Where are inputs, outputs, and progress recorded?
4. Failure: What happens if a source, model, or database is unavailable?
5. Review: Who checks the result before it affects other people?
6. Value: Does it save useful time or improve quality after review and operating costs?

For this edition’s planned demonstration, we will show a successful run, a deliberately failed dependency, and the recovery path. The published version should include actual run evidence and costs—not a claim that one attractive output proves reliability.

Buy the responsibility you do not want to carry.

A hands-on builder may value Docker for control and a clear service structure. A team with limited operating capacity may get more value from a managed workflow platform. An IT organization may need an environment that meets existing identity, procurement, and data requirements.

Begin with one real workflow. Identify the data it touches, its consequence of failure, and the person accountable for it. Then compare options against those needs rather than the size of their feature lists.

• Build: when the required control or integration justifies ongoing operating work.
• Buy: when a managed service covers the requirements at an acceptable total cost.
• Combine: when managed services handle common capabilities and a small owned layer handles distinctive work.

The result is the point.

The platform earns its place when useful work happens repeatedly, exceptions are visible, and people can trust the process enough to use it. That is the connection worth exploring: strategy chooses the work, technology makes new approaches possible, and actual outcomes tell us what has value.