When Fable 5 and Mythos 5 went dark for about nineteen days under US export controls, firms that relied on them got an unplanned lesson in AI continuity. The reason was a genuine safety finding. Set that reason aside for a moment and look only at the effect, because the effect is the part that should concern a firm. A tool that many people relied on was unavailable, without notice, for the better part of three weeks.
Treat AI as infrastructure
The first shift in thinking is to stop treating AI as a novelty and start treating it as infrastructure. A firm that routes its drafting, its research or its document review through a single tool has built a single point of failure into the way it works. When that tool goes dark, the work stops, and the deadline does not move to accommodate it. Infrastructure is something you plan around, with the same seriousness you would give to your case management system or your email.
The trouble is that AI tends to earn its place quietly. A member of staff starts using it for one task, finds it saves an hour, and spreads the habit without anyone deciding that the firm now depends on it. By the time the tool is woven through the day, no one has treated the dependence as a decision, and so no one has asked what happens if it stops. The outage is the moment that unasked question arrives all at once.
Build in resilience
A few practical measures reduce the exposure. Keep a fallback tool that staff already know how to use, so that switching to it is a matter of minutes rather than a scramble on the day. Avoid leaning on one vendor for everything, since concentration is what turns a single outage into a firm-wide stoppage. Ask about continuity and notice when you negotiate a contract, so that you know what the vendor owes you if the service stops. And keep the manual process alive, because the ability to do the work by hand is what lets the firm carry on when the tool is not there.
None of this is exotic. It is the same discipline a firm applies to any system it depends on. The difference with AI is that adoption has been quick and the dependence has grown faster than the planning around it, which is how a firm ends up relying on a tool it has no fallback for.
There is a related point about the work itself. If the firm has let a tool take over a task so completely that no one remembers how to do it by hand, the outage does more than pause the work. It exposes a skill the firm has quietly lost. Keeping the manual process alive is partly about the day the tool is down and partly about making sure the people in the firm still understand the work well enough to judge what the tool produces on every other day. A skill you cannot perform is a skill you cannot supervise.
What a firm should do
Write down what happens on the day a tool is unavailable, and do it before that day arrives. Name the fallback, name the person who decides to switch to it, and set out how staff keep working while the primary tool is down. A short document of this kind turns an outage from a crisis into an inconvenience, and writing it takes an afternoon rather than a project. The suspension that prompted this piece lasted about nineteen days. The next one might be shorter or longer, and the firm that has thought about it in advance is the one that keeps its promises to clients while it lasts.
The wider resilience basics are set out in the NCSC's small business guide, which every firm should have worked through once.
If your firm would feel a nineteen-day outage of one tool, the fallback plan is overdue, and we write them with firms in a day: start with a conversation.