The software selloff in early February raised a question that goes beyond a week of share prices: where will enterprise customers find value as AI takes on more of the work?
Axios reported on February 7, 2026 that investors had lost over $400 billion during that week’s software selloff.
A change in market value reflects changing expectations. It is not a transfer of that amount in cash to another company, and it can’t prove what caused every investor’s decision.
My read is that leaders need to examine the layer of their business where customers experience value. If AI connects work across tools, the value of organizing that work may change. That is a strategic question to test, not a forecast a share-price move settles for us.
The mechanism behind the repricing
Software valuations depend partly on expectations about future revenue and margins. When the future becomes harder to assess, investors may pay less for those expected results even if the current business has not changed.
The product changes behind the question
There's a meaningful difference between AI that helps you do your work and AI that does the work.
Two announcements on February 5 made the enterprise use case more concrete: Anthropic’s Claude Opus 4.6 and OpenAI’s Frontier.
Anthropic introduced Opus 4.6 with expanded capabilities for coding and knowledge work, including a one-million-token context window in beta. The useful comparison for a business is how it performs on its own tasks, with the information and review those tasks require.
OpenAI introduced Frontier as a platform for building, deploying and managing enterprise agents. Its announcement emphasizes shared business context, onboarding, evaluation and permissions. Those are operating requirements around the model, not simply a better answer in a chat window.
That is the shift I want enterprise leaders to examine: from software that stores or organizes work toward systems that can carry selected tasks across it.
Where the moats are moving
Proprietary data, distribution, scale and switching costs still matter. AI can change how a customer accesses or uses those assets. The question is which part of the relationship remains valuable when more of the task can happen through an agent.
Established systems carry years of process knowledge, integration and customer trust. Replacing them is a different problem from generating a working prototype. That distinction matters when you assess exposure.
A system that works across several applications could become the place a customer starts. Individual applications could remain necessary while becoming less visible in the daily workflow.
This is the structural possibility I find most useful: a business can remain important infrastructure while another layer captures more of the customer relationship. It is one scenario to plan for, rather than an inevitable outcome for every software company.
Ask where your product is a convenient interface and where it is a deeply embedded system of record. Then ask what an AI-enabled workflow would need from each. The answers should shape what you build and how you connect with other tools.
Execution matters, but speed alone is not the moat. The useful combination is customer understanding, reliable delivery and a way to learn from the result.
That is what I mean by compounding: building a process that carries evidence from one cycle into the next. Improvement needs to be observed and acted on; it is not automatic just because agents are involved.
Capability and adoption are different questions
I wrote about the AI capability overhang in my last article. The concept is simple: AI capability has leapt forward. Adoption hasn't. And the distance between those two things is growing, not shrinking.
A production workflow gives a team evidence that a pilot can’t: how people use it, where exceptions occur and what it costs to operate. That evidence can improve the next decision if someone owns the learning.
What this means and where the value moves next
The strategic question is where your customers will want the work to happen, and how your company can remain useful as that changes.
Look at a real workflow. What information does it require? Which decisions can a system support? Where do users need a person, a reliable record or an accountable provider?
For a technology company, there may be several valuable positions: coordinating the workflow, supplying an essential capability or maintaining the trusted record beneath it. Know which role you are trying to earn.
If you provide established software, examine whether customers value access, usage or an outcome. Test pricing and integration choices against that evidence. Adding an AI feature is useful only if it improves what customers are trying to do.
If you are building a new capability, be precise about where it fits. Some companies will coordinate the work. Others will supply a critical component. Both need a clear understanding of the customer, the handoffs and the result they are responsible for.
The work is to understand the changing architecture well enough to make a deliberate choice. That is a more useful response to a market signal than either dismissing it or treating it as a complete strategy.
Start with the part of the customer’s work where your company has the strongest reason to matter.
From the portfolio
Board of Innovation helps enterprises respond to this repricing by redesigning what they sell, how the work gets done, and where durable value can still be built. The goal is not another AI feature. It is a business that becomes more valuable as intelligence gets cheaper.
