Keep Work Moving When AI Drops: A 10-Minute Plan

The outage is the test, not the excuse

You are halfway through a client deliverable, the model you used to draft, summarize, or check the work stops answering, and the deadline is still on the calendar. That is the moment your reliability gets tested.

AI tools are now part of many knowledge workflows, but they are also a shared dependency. When one fails, the work still has to move, and reliability means handing over a draft, a plan, or a status update when the model is silent.

Problems in four of the most popular chatbots began around 15:00 Spain peninsular time.

Around 17:00, ChatGPT, Claude, Gemini, and most known Western AI companies stopped functioning almost at once, and service was not fully restored until after 19:00, with Anthropic's chatbot most affected.

OpenAI's chatbot was the first to recover normal service, while ChatGPT, Claude, Gemini, and Grok suffered an outage that affected thousands of users for several minutes.

77% of reported ChatGPT problems were directly related to the chatbot, 10% to OpenAI's code assistant Codex, and 7% to the application.

In some cases, Claude denied having problems but was then unable to load additional responses or search the web.

That sequence matters: the failure is not just a blank screen. The tool can look alive, give a confident answer, and then stop producing anything useful.

For a freelancer, consultant, or knowledge worker, the risk includes lost time and the client or manager wondering whether you can deliver without the shortcut.

The fallback is the real deadline protection

A fallback is a way to keep the deliverable moving with what you already have: notes, templates, a manual method, and a clear message.

Start with the task that would hurt most if it stopped. If you use AI to draft client emails, the fallback is a short email template you can fill in by hand. If you use it to summarize long documents, the fallback is a simple summary structure with headings you already trust.

You do not need to rebuild your whole workflow. You need one reliable path for the task closest to a deadline.

The manual path should be boring. It should take less time to start than it takes to explain why you are late.

Keep the fallback in a place you can reach without the tool: a plain-text file, a local document, a printed page, or a note in your calendar.

The checklist turns the plan into a habit

The plan takes about ten minutes, and each item is small enough to finish in a minute. Run it once this week, not only when something breaks.

• Identify an AI-dependent task and write the first manual step.
• Save a plain-text copy of your last AI-assisted draft so you can edit it without the tool.
• Draft a brief update that says what changed, what you are doing, and when the client will hear back.
• Do a quick dry run: open your fallback method and confirm you can produce one usable line.
• Set a short timer this week to repeat the checklist before a deadline.

The first two items protect the work itself, the next two protect the relationship, and the last one makes the plan repeatable before a deadline.

When a tool fails, silence reads as uncertainty. A short update protects the relationship and shows you are still in control.

Keep it short. Say the tool is down, say what you are doing instead, and give a realistic next update time. Do not over-explain the outage. Do not promise a recovery time you cannot control.

If the deadline is still safe, say so. If it is at risk, say what you can deliver now and what needs a short extension. The client does not need a technical report. They need a plan.

After the tool comes back, do not delete the fallback. Keep it for the next failure.

Do the checklist this week.