The Morning After AI Goes Dark: When Innovation Outruns Operational Discipline
Building resilient AI strategies requires operational discipline and recovery planning as much as innovation and speed.
In my last article, I wrote about what I called a culture of immediate—the pressure organizations face to move faster, adopt faster, and prove they are not standing still while AI reshapes the landscape around them. The question I asked then was simple: Are we getting AI right, or are we just getting it fast?
Damian Mathews and The Last Mile Team published a piece that brought me back to that question in a way I hadn’t expected. Their article wasn’t about a specific model. It was about what happens when a capability organizations depend on suddenly disappears.
And it highlighted something I don’t think we’re talking about enough.
Most organizations are building AI strategies before they’ve built recovery strategies. Security has become part of every AI conversation. Recovery too rarely is.
The Morning After
The events surrounding Claude Fable 5 and Mythos 5 created a scenario few organizations had seriously contemplated. One day, a frontier model was available. The next day, it wasn’t. Regardless of where anyone lands on the details, the event exposed something important: many organizations are not prepared for the morning after.
The morning after a model changes. The morning after a capability disappears. The morning after a feature your team built into a workflow behaves differently than it did the day before. Most importantly, they are not prepared for the moment someone asks a question that should have been answered long before deployment:
What do we roll back to?
Lessons We’ve Already Learned
For years in customer experience and contact center technology, we have operated in environments where change was expected. Platforms evolved. Releases happened. Features matured. Vendors adjusted roadmaps. Occasionally, a feature disappeared or a migration became more complicated than anyone anticipated (here’s looking at you, Intuity TUI). Yet through all of that, enterprise technology developed a discipline around managing change.
The technology industry spent decades learning how to operate in environments where change is inevitable. We built change-control processes. We built disaster recovery plans. We built redundancy. We built rollback procedures. We created governance frameworks and escalation paths—not because we expected failure, but because we respected the possibility of it.
The best technology organizations never assumed a platform would remain unchanged forever. They assumed it wouldn’t. What surprises me about many AI conversations today is how quickly we’ve forgotten those lessons.
The More Important Questions
Somewhere along the way, many organizations stopped asking operational questions and started asking only innovation questions. What model should we use? How quickly can we deploy it? What productivity gains can we achieve? How much cost can we remove? Those are important questions, but they are incomplete.
The more important questions are often the less exciting ones:
- What happens if the model changes?
- What happens if the vendor changes direction?
- What happens if regulations change?
- What happens if a capability becomes restricted?
- What happens if the experience you’ve designed today no longer exists tomorrow?
Those questions aren’t anti-innovation. They’re operational discipline. And operational discipline has never been more important.
What Feels Different Today
AI is not new to customer experience. Those of us who have spent our careers in this industry know that forms of artificial intelligence have existed in the contact center for decades. Speech technologies, predictive routing, workforce optimization, intelligent self-service, and automation have all played a role in shaping customer experiences long before generative AI became the topic of every board meeting.
Regulatory oversight isn’t new either. Organizations have spent years navigating requirements around privacy, accessibility, financial services, healthcare, security, and consumer protection. We understand compliance. We understand governance. We understand adapting technology to operate within evolving rules.
What feels different today is immediacy. Are we getting AI right, or are we just getting it fast?
Historically, organizations had time to respond to change. Today, regulatory decisions, vendor decisions, and model decisions can affect operations with unprecedented speed—not next quarter, not next year, but potentially tomorrow morning.
And if your customer experience is dependent on a single model or a single architectural approach, your customers won’t care why the capability changed. They will only experience the result.
Data Sovereignty and the Best Friend Experience
This brings me back to something I’ve spent years defining with my customers: the Best Friend Experience. Organizations are increasingly looking to AI to help deliver that experience, and rightly so. The technology is remarkably powerful when implemented thoughtfully.
But the Best Friend Experience requires something we don’t discuss nearly enough: data sovereignty. The model should not own the relationship. The platform should not own the relationship. The vendor should not own the relationship.
The organization should.
Because if customer understanding exists only inside a single model, then the relationship itself becomes dependent on that model’s continued availability. If the intelligence behind the experience changes, the customer relationship must survive the change. The context, history, and understanding that make the experience valuable should remain portable regardless of which technology is powering it at any given moment.
Automation Before AI
This connects to another conclusion I’ve reached after countless conversations with customers and industry peers.
If your organization has not yet mastered automation, it may not be time to jump directly into AI.
One of the most common patterns I see is organizations wanting to leapfrog foundational operational maturity in pursuit of AI transformation. Yet the most successful AI deployments I’ve observed are rarely the organizations that started with AI. They’re the organizations that first understood their operation well enough to automate it.
They knew their baselines. They understood where work flowed and where it stalled. They knew where customers became frustrated and where employees spent their time.
- They had reporting.
- They had governance.
- They had change control.
Most importantly, they understood what great looked like before asking AI to make it better. Because AI wasn’t their starting point. It was an accelerator.
That’s why I believe automation, measurement, governance, and change control (including your rollback plan) deserve more attention than they’re receiving in the current conversation. AI is not a substitute for operational maturity. It amplifies whatever maturity already exists.
Organizations with strong operational foundations can create extraordinary outcomes with AI. Organizations without those foundations often create complexity faster than they create value.
Who Owns the Intelligence?
Which brings me to what I increasingly believe is one of the most important architectural questions facing customer experience leaders:
Who owns the intelligence?
If the customer relationship, customer context, and operational knowledge exist primarily inside a single model, then the organization has outsourced more than technology. It has outsourced flexibility.
The Best Friend Experience depends on continuity. The customer should remain known even when the underlying technology changes. That requires data sovereignty. It requires portability. And increasingly, it requires modularity.
Bring Your Own AI
This is why I believe so strongly in a Bring Your Own AI approach.
- Not because every organization needs every model.
- Not because complexity is inherently good.
- But because resilience matters.
Employees already operate this way. Inside most organizations, people use different tools for different outcomes. They use Copilot, Claude, ChatGPT, Granola—they select the right tool for the right moment.
We’ve already accepted flexibility in the employee experience. Our customer-facing experiences should be designed with the same mindset.
The capability should belong to the business. The data should remain governed. The experience should remain portable. And the architecture should allow organizations to adapt when change inevitably arrives.
Innovation and Operational Discipline
Innovation and operational discipline should not be opposing forces.
The most successful technology organizations undertaking AI—those that will be written about in the future—will have understood both. They will move forward while maintaining the ability to recover. They will adopt new capabilities without becoming dependent on them. They will build flexibility into their architecture long before they need it.
AI doesn’t change those principles.
If anything, it makes them more important.
Because the question isn’t whether the technology will change.
It will.
The question is whether the customer experience you’ve built can survive the change when it does—including the morning after things have gone dark.
#ArtificialIntelligence #CustomerExperience #Leadership #AITransformation #ContactCenter
