Every company’s AI strategy is also a decision about where its next outage will land. Most companies make that decision without anyone responsible for AI business continuity planning in the room.
McKinsey’s December 2025 report, The AI reckoning: How boards can evolve, asks boards to pick an AI posture along two lines. Will AI create new products and revenue, or improve the business you already run? Will it be embedded across the enterprise, or used in targeted workflows? The answers produce four postures: business pioneers, internal transformers, functional reinventors, and pragmatic adopters.
McKinsey frames posture as a strategy and governance choice. It is. It is also a dependency map, and your business continuity program should be reading it.
The short answer: AI business continuity planning starts by reading your company’s AI posture as a dependency map: the more deeply AI is embedded, the more of the business stops when it fails. Find where AI has replaced a manual workaround, and plan for that failure before an outage finds it.
Posture Decides What Failure Looks Like
Each posture puts AI in a different place in the business. That changes what breaks, who notices first, and what recovery means.
| Posture | Where AI sits | What failure looks like | Start here |
| Business pioneers | In the product | A public failure in front of customers and regulators | Crisis communications and a product incident playbook |
| Internal transformers | In the operating backbone | Operations stall, with no manual process left to fall back on | AI in every business impact analysis, with recovery targets |
| Functional reinventors | In vendor tools, workflow by workflow | One provider’s outage or model change breaks several workflows at once | Third-party mapping and provider concentration |
| Pragmatic adopters | Wherever employees put it | A dependency nobody knew existed | Find the shadow AI first |
Business pioneers sell AI. McKinsey’s example is a medical-device company that evolves into delivering AI systems that interpret scans and suggest treatments. When that AI fails, it isn’t an IT incident. It’s a product failure with customer harm attached. “Our AI did it” won’t satisfy a customer, a regulator, or a plaintiff’s attorney. Your crisis team, and your board, need to have rehearsed this before launch, not after. (I laid out the questions directors should ask in your board oversees AI; has it ever watched AI fail?)
Internal transformers run the business on AI. McKinsey tells their boards to be “particularly diligent in probing resilience, observability, and explainability,” and describes one manufacturer’s board that asked whether AI-driven planning, supply chain, and maintenance systems were “stress-tested before approving new capital.” That is a board asking a continuity question. Your business continuity program should be ready to answer it.
Functional reinventors buy AI tool by tool. McKinsey notes they are more likely to buy than build and flags vendor lock-in, including whether the company keeps “internal support capabilities.” The continuity risk is concentration. Five workflows on three tools can still sit on one model provider. When the AWS outage hit in October 2025, plenty of companies learned their dependency map was shallower than they thought; I talked through that on Managing Uncertainty, Episode 314. AI providers are next.
Pragmatic adopters wait for proven tools, so they assume they have little AI exposure. They usually have more than they think. Employees adopt AI on their own. In healthcare, a Vanson Bourne survey for Imprivata found that 72% of organizations deploy AI tools or agents without formal IT approval, at least occasionally. The risk isn’t the AI you approved. It’s the AI you don’t know about.
Most large companies straddle more than one posture, and McKinsey says so. That’s fine. Map each business unit to its own column.
The Manual Workaround Is Disappearing
Here is the continuity problem nobody budgets for.
When AI absorbs a process, the manual version starts to decay. The people who knew it move on. The forms get retired. The headcount that could run it by hand gets reallocated, which was often the point of the investment.
Then the AI goes down, and the workaround in your continuity plan describes a team that no longer exists.
This is not an argument against AI. It’s an argument for knowing, before the outage, how long the business can run without it and who will do the work. McKinsey suggests boards track “backup drill results” as an AI resilience indicator. That’s the right metric. Most companies have never run the drill.
AI Business Continuity Planning: Four Moves for This Quarter
1. Get your posture in writing. Ask whoever owns AI strategy which posture the company is pursuing, by business unit. McKinsey tells boards to review posture at least annually. Make sure the business continuity program hears the answer, not just the board. If your board is talking about AI without you, get into that conversation.
2. Add AI to every business impact analysis. In each BIA interview, ask: which AI tools or agents does this process rely on, who provides them, and what happens if they disappear tomorrow? Then validate the answers. Process owners routinely underestimate their dependencies, which is why validating BIA results matters more than collecting them.
3. Set recovery targets for the process, not the model. Nobody needs a recovery time objective (RTO) for a model. They need one for the claims process, the scheduling desk, or the planning cycle that now runs on it. Add one more question your DR runbooks probably skip: if the AI quietly produced bad output for two weeks, when was the data last right? I covered that failure mode in your AI agent is now part of your incident.
4. Turn it off on purpose. Run a business continuity exercise where a critical AI tool is unavailable for a full business day. Watch whether the manual workaround still exists, who knows how to run it, and how long the backlog takes to clear. For functional reinventors, bring the vendor in; exercising with your key vendors exposes contract and notification gaps faster than any questionnaire.
Read the Posture, Then Plan for It
A pioneer and a pragmatic adopter don’t need the same continuity program. The pioneer needs a crisis team that can handle a public AI failure. The transformer needs recovery targets and a manual fallback that still works. The reinventor needs to know which provider sits under everything. The adopter needs to find the AI it didn’t approve.
Your strategy team chose an AI posture. Whether they meant to or not, they also chose where your next outage will land.
Keep Going
A few ways to go deeper if this was useful.
- Read more. Resilience, crisis management, and continuity writing at Bryghtpath Insights, or the structured Ultimate Guide to Business Continuity.
- Map your AI dependencies. Our business continuity services build AI tools and providers into your BIA, recovery strategies, and exercises.
- Get a maturity score. Our Resiliency Diagnosis® is a standards-based review that produces a maturity score and a prioritized roadmap.
- Talk to us. Set up a call to think through your program with us.





