AI tools change. Sometimes gradually, sometimes with an announcement that sends you scrambling. The question isn’t whether you can anticipate every change. It’s whether your business has a practical way to respond when one affects the work you depend on.
Imagine you've spent months building a custom AI assistant for your business...
You’ve done all the right things. Refined the instructions, uploaded reference documents, and worked through all the heartburn of getting a your new tool to do what you need.
Amazing timesavers are happening. Maybe its preparing client proposals or researching or organizes information that used to take hours to pull together.
Then comes the announcement.
The provider is retiring the tool.
Now what?
If you’re a solopreneur or small business, these aren’t minor hiccups.
They’re trigger big business questions.
AI Tool is Being Retired. Immediate Questions.
- Can I keep my instructions and reference materials?
- Will the replacement perform the same tasks in the same way?
- What happens to all of my connected applications?
- How much rebuild time will this cost me?
It Happened On September 11, 2026
OpenAI announced plans to retire custom GPTs, with a scheduled retirement date of December 11 and a migration path to Plugins.
For businesses that have incorporated custom GPTs into recurring work, that announcement creates an immediate reason to examine what they’ve built and prepare for what’s coming.
But I think it also raises a larger question.
When we commit to an AI tool, are we thinking only about what it can do for us today? Or are we considering what happens when the circumstances around that tool change?
Because they will change.
And that’s where contingency planning comes in.
Tools May Change. The Business Need Doesn't.
Let’s consider this business use case.
The company wants the ability to (1) search multiple hiring platforms, (2) identify relevant opportunities, and (3) consume-n-consolidate the results into something useful.
Pretty straightforward goal: spend less time searching and more time evaluating promising leads. That’s definitely AI’s wheelbarrow.
The company was just about to hit the approve button when notice the retirement announcement appeared.
Does the business abandon the idea? Why should they?
The company still has the same needs list.
What has changed is their original way to get there.
This is the important distinction, particularly for organizations that are just beginning to incorporate AI into their ecosystems.
It’s easy to become attached to a tool.
You’re bought into what it can do. Before long, you’re designing your process around that particular technology.
But the tool and the business need are two different things.
In this research example, the company needs to understand which platforms must be searched, what information should be captured, how results will be verified, and what an acceptable output looks like.
Luckily the company had not purchased the new tool but, it’s an overarching takeaway that should be a part of any tool consideration. Moving to another AI tool doesn’t mean the new tool automatically has the same integrations, permissions, or capabilities.
When the company understands the work independently of the technology, it has a much better starting point for evaluating alternatives.
And if circumstances change, it isn’t starting over with the fundamental question of what the business needs.
Contingency planning starts with understanding the work that needs to continue, not simply identifying another product you might use.
It’s easy to become attached to a tool.
Before You Commit, Ask What Happens If the Tool Goes Away
When a business owner evaluates an AI tool, the questions usually center on immediate value.
Can it save me time?
Will it help me serve my customers?
Can I afford the subscription?
And perhaps most importantly, can I figure out how to use it without creating more work for myself?
All reasonable questions.
But they tell us very little about what happens after we’ve come to rely on the tool.
We need to think about that, too.
I’m not suggesting that every small business needs a comprehensive disaster recovery program before experimenting with ChatGPT. That would be overkill for most organizations.
What I am suggesting is that the more important an AI tool becomes to your operation, the more you need to understand what would happen if you lost access to it.
There are three questions I’d start with.
1. How important is this tool to the business?
Think about how you’re actually using it.
An AI assistant that occasionally helps you brainstorm social media ideas is one thing. An assistant that you use every morning to prepare client deliverables is another.
If the second tool became unavailable tomorrow, what would happen?
Could you complete the work manually? Would your customers be affected? Would you miss deadlines or lose revenue?
How long could you reasonably operate without it?
You may discover that some of your AI tools don’t need much contingency planning at all.
Others might deserve considerably more attention.
The point is to make that distinction before something happens.
2. What needs to be preserved?
This is where I think people can underestimate what they’ve built.
Your AI workflow may involve much more than an active subscription.
Think about the instructions you’ve refined over time. The templates you’ve developed. The reference documents you’ve uploaded. The applications you’ve connected.
And what about the knowledge you’ve accumulated while getting the tool to work the way you want?
Some of that may be easy to transfer. Some may not.
So where are those materials stored? Which ones do you own and control? Do you have accessible copies outside the AI tool?
You should also consider whether you’re using something you created or relying on an AI assistant built by someone else.
Those are different situations.
Having permission to use another person’s tool doesn’t necessarily mean you have access to its underlying instructions, configuration, or a way to replace it.
You need to know what you can preserve before you’re faced with having to preserve it.
3. How will work continue?
Sometimes the alternative is another AI tool.
Sometimes it isn’t.
You might have a manual procedure that can keep things moving. Another application might handle the task temporarily. Or perhaps there’s someone who can help while you work through the transition.
What matters is that you’ve thought about a workable response.
And if the workflow is important enough, test it.
There’s a difference between believing you have a backup and knowing that your backup works.
You don’t need to eliminate every technology dependency. That’s not realistic.
But you should understand which dependencies you’re accepting and what they mean for your business.
A Contingency Plan Should Be a Living Process
Now, here’s where I want to take this a little further.
Let’s say you’ve done everything we’ve discussed.
You’ve identified your critical AI workflows, preserved the important materials, and documented what you’ll do if a tool becomes unavailable.
That’s a good start.
But what happens six months later?
Maybe the provider has introduced new capabilities. Maybe its pricing has changed. Maybe a contractor who once needed access to your AI tools is no longer working with you.
Or perhaps one of your connected applications has changed its permissions, and a workflow that worked perfectly when you implemented it isn’t functioning quite the same way anymore.
Your original contingency plan may still be sitting there, looking perfectly reasonable.
The problem is that it no longer reflects how you’re operating.
This is where my background as an Agile coach influences how I think about AI adoption.
When we introduce a new business process, we don’t consider the work finished simply because the process has been implemented.
We establish opportunities to inspect what’s happening.
Is the process still working as intended? Have circumstances changed? Are there adjustments we need to make?
Then we adapt.
Why wouldn’t we apply that same thinking to AI?
At TAP, I describe this as treating contingency planning as a living process.
And I mean something more specific than keeping a document up to date.
A living process involves actual, recurring interaction with your operating environment. Someone checks what’s happening, determines whether anything has changed, and takes action when necessary.
That might mean reviewing who currently has access to your AI tools.
It might mean checking your connected applications to confirm that they still have the right permissions.
It might mean reviewing a provider announcement and asking, “Does this affect anything we’re doing?”
These are active checks.
Not reminders to open a document and change the date at the bottom.
Now, if you’re running a small business, you might be thinking this sounds like another layer of administration you don’t have time for.
I understand that concern.
But a living process doesn’t have to be complicated.
You need a cadence that makes sense, someone accountable for the review, and a practical way to follow through when something needs attention.
That’s it.
The objective is to make adaptation part of how you operate, rather than something you only think about after a problem occurs.
Build a Review Cadence That Fits Your Business
For a solopreneur or small team, I’d suggest starting with a monthly AI workflow review.
Fifteen to thirty minutes may be enough, depending on how many tools you use and what you’re using them for.
That’s a starting point, not a rule.
If you’re relying on an AI workflow for sensitive or business-critical work, you’ll probably want to review it more frequently.
The review itself can follow a simple Agile-inspired pattern.
Inspect → Evaluate → Adapt
A recurring review of the tools and workflows your business depends on.
Inspect
What has changed?
Look at provider announcements, tool availability, current users, connected applications, permissions, and relevant policies.
Don’t assume that everything is still configured the way it was when you implemented it.
Evaluate
Does the change affect our workflow?
Maybe an integration needs attention. Maybe a new feature changes how a task is performed. Or maybe nothing significant has changed at all.
Where appropriate, run a representative task to confirm that the workflow still produces acceptable results.
The question isn’t simply whether the tool is available. It’s whether it’s still doing what the business needs.
Adapt
What do we need to do about it?
You might need to adjust access, reconnect an application, update instructions, retrain someone, or investigate an alternative.
Sometimes the answer will be to activate your contingency plan.
Whatever you decide, identify who is responsible and confirm that the action gets completed.
There’s one more consideration.
Your review cadence shouldn’t become a reason to ignore something important between reviews.
If your AI provider announces today that an essential feature will disappear in 60 days, you don’t wait until next month’s scheduled review to think about it.
You assess the impact when you learn about it.
The recurring cadence provides consistency. Significant events create additional opportunities to respond.
And remember, the process itself should be proportional to your business.
You don’t want to spend more time managing AI than the tools are saving you.
You want enough structure to recognize meaningful changes, make informed decisions, and keep your business moving.
A Real-World Opportunity to Put the Process Into Practice
Let’s bounce back to OpenAI’s custom GPT announcement.
According to OpenAI’s migration FAQ, GPT instructions can move into a plugin skill, and knowledge files can be copied into plugin reference files.
But not everything transfers.
- Existing conversations don’t migrate.
- Custom actions don’t transfer automatically.
- Sharing settings need to be reviewed. And
- A migrated plugin may behave differently than the original GPT.
That’s quite a bit to consider, especially after you’ve spent months developing the assistant.
So what should next steps look like?
- Start by identifying which of your current workflows depend on custom GPTs. Next,
- Look at the materials and capabilities you need to preserve.
- Review the migration options available to your account
- Test the replacement before making it part of day-2-day operations.
For the latest account-specific details, refer to OpenAI’s official Custom GPT Retirement and Migration FAQ .
But here’s what I don’t want us to lose in all the discussion about this particular transition.
The larger lesson isn’t about moving from one OpenAI product to another. It’s about how we manage change after we’ve incorporated AI into our businesses.
If you already had a living contingency process in place, the announcement could have been identified during a routine review.
You would have had an opportunity to assess which workflows were affected, determine what needed to happen next, and assign responsibility for the transition.
Without that process, you might not discover the impact until someone tries to use a tool and encounters a problem.
And this isn’t just about existing custom GPT users.
Imagine you’re considering building your first customized AI assistant.
You can apply the same thinking now, before you commit.
Define the business requirements. Understand what depends on the provider. Identify what you need to preserve. Decide how you’ll monitor changes once the tool is in use.
Will that prevent every disruption?
Of course not.
But it gives you a way to respond deliberately instead of having to figure everything out under pressure.
Responsible Adoption Means Staying Engaged Post Implementation
AI offers SMBs and nonprofits useful opportunities.
But implementation isn’t the finish line.
Once a tool becomes part of your ecosystem, it’s important to remain engaged with the work it supports.
This doesn’t mean watching every product announcement or chasing every new AI capability.
It does means understanding what you depend on and establishing a manageable way to notice when something important changes.
Before you commit, know what the tool needs to accomplish and how you’ll continue the work if it becomes unavailable.
During implementation, preserve what matters.
And after implementation, establish a living process for inspecting your AI environment, evaluating changes, and making adjustments.
You can’t predict every unexpected event.
That’s how AI adoption becomes more than simply putting a new tool to work.
It becomes part of building a business that can continue to operate as both its needs and the technology evolve.
Three Takeaways
1. Build your continuity before you commit. Understand what the AI tool will support, how your business depends on the tool, and how essential work could continue if the tool becomes unavailable.
2. Make contingency planning a living process. Establish a recurring cadence to review tool availability, user access, connected applications, provider changes, and whether your workflows still meet your needs.
3. Make adaptation someone’s responsibility. Even if you’re a one-person business, there needs to be a deliberate way to identify changes, evaluate their impact, and follow through on necessary adjustments.
