batteriesincluded.com · Questions & Answers

If our AI provider goes down during the workday, how should a forward-deployed engineer keep our operations moving?

Require a business continuity plan for the workflow, not just an uptime promise for the application. Your tool can remain accessible while its AI provider, database, or connected business system is unavailable.

Start by identifying which work must continue, which can wait, and which must stop safely. An AI outage might reasonably delay drafting an internal summary. It should not cause a tool to invent missing information or mark unfinished work as complete.

Ask your forward-deployed engineer to define:

- How the application detects and communicates an unavailable dependency.
- Which functions remain usable without AI.
- Where pending work is stored and how employees can see its status.
- The manual fallback for time-sensitive tasks.
- Who authorizes resuming automated processing after recovery.

Use existing systems where possible. A fallback in the business software employees already know usually creates less operational friction than an entirely new emergency tool. It must still preserve the necessary permissions and a record of what happened.

Keep EOS as the operational backbone. Continuity responsibilities should align with the Accountability Chart, and unresolved business risks should reach the appropriate Issues List. The engineer can implement recovery behavior, but leadership must decide which interruptions are tolerable.

For a Lovable or Replit build, test dependency failure deliberately in a safe environment. Do not assume switching to another AI provider is seamless; output behavior, data terms, and capabilities can differ.

The recommendation is graceful degradation: keep safe work moving, make blocked work visible, and ensure recovery does not depend on employees guessing what the system completed.

Category: Forward-Deployed Engineering

← All questions