AI Is Disposable
The whole AI is your copilot puts AI in the wrong seat and foils adoption. Instead, AI apps should be considered disposable.
Every startup and software company peddling AI apps and/or functionality pitches its wares as an indispensable agent to the modern-day worker. My long-term belief is that AI apps are just the next generation of applications. As such, when effectiveness is lacking, they should be disposed of and discarded promptly.
These would-be assistants promise to achieve incredible efficiencies and make work easier. In many cases, they do! From small fractional improvements to massive efficiencies, some apps help tasks quite a bit. But when they don’t or a better app comes along, users should get rid of them as quickly as possible.
Why pay for an inferior app? Remove it.
If an AI app can’t deliver the result after a reasonable number of tries, why not try a different one? I love Claude, but after three attempts, it’s on to ChatGPT or Perplexity. Midjourney isn’t working, let’s try Firefly or Dall.E (via ChatGPT). We just ripped out a CRM at CognitivePath because of its weak AI functionality.
I have no long-term loyalty to any AI app. They are disposable. Even developers are realizing this ethos.
Disposable Apps Are the Norm
Disposable apps have become the norm in the consumer space. It’s harder to achieve in professional environments, but monthly AI productivity app fees and software as a service (SaaS) licensing make cutting contracts on a 30-day basis easier.
As AI apps improve over the next two years, the general technology adoption focus will move from changing humans to work better with AI to finding a better, easier-to-use app. And that’s the right course of action.
Significant training is still needed to overcome fears and the usual technology change issues. According to PWC, 41% of executives say that workforce issues, such as training, culture, or change in work, are among the top-five challenges their organizations face in using GenAI.
But AI products and companies have UX challenges and functionality fits that lag behind what customers have come to accept as viable, too. Not all apps are great, even those named Copilot. At least Microsoft is a bit more transparent about its efforts. At its most recent BUILD developer conference, Microsoft highlighted the good and the bad of its Copilot efforts, as noted by Gergely Orosz:
Dogfood that’s not tasty: Copilot Agent fumbles in real world with .NET. Microsoft is experimenting with Copilot on the complex .NET codebase, meaning everyone sees when this agent stumbles – often in comical ways.
Still, I expect the enterprise market, already burned by phase one of the generative AI boom, to be much less tolerant of bad AI apps and other unnecessary copilots. AI apps must become more useful, not just in the now, but in their continual evolution. There will always be a next-gen app around the corner.
Mission Entanglement Is Good… Until It’s Not
The realities of good technology management, whether on a personal level or an enterprise level, require constant evaluation of software packages and pricing. Mission entanglement occurs as an app becomes more intrinsic and valuable to an organization’s efforts, and that’s a win-win.
But once a software platform or application loses its value or becomes too costly, consumers and well-managed small businesses press the cancel the account button. Eroding functionality becomes a challenge for most organizations, though.
Removing a bad CRM or AMS system, for example, can turn into a costly months-long process that involves reporting APIs, records reconciliation, and process disruptions and rebuilds. Many prefer to live with “the devil they know” rather than replace them. They become “locked-in” with an almost impossible platform to replace.
As AI makes things easier, for example, flexible access to data via the MCP protocol instead of traditional API access, organizations will become less encumbered by legacy platforms. AI functionality will make it easier to migrate away from bad enterprise platforms that no longer provide expected results.
While not quite disposable, big tech companies and their foundational platforms will become less entrenched. Lesser apps, well, they should be removed as soon as they lose value. AI makes everything faster, including the lifespan and usefulness of some productivity applications.
What do you think? Should AI apps — and apps in general — be considered disposable?
Geoff, you mention evolution. In my opinion, it will become survival of the fittest, not unlike any other group of tech products. Remember the early search engine days? Excite, Altavista, Dogpile, and so many others are long gone. Until the I/O conference, I was concerned that Google may die on the vine. Not so any longer.
My go-to AI tech stack consists of ChatGPT, Claude, Perplexity, and Gemini—all paid for and, so far, indispensable. But I'm with you. If they cease to provide value, then out with them. (Although I must confess that I don't see that happening soon.)
You also mentioned the PWC study, which found that 41% of executives say workforce issues, such as training, culture, or change in work, are among the top-five challenges their organizations face in using GenAI.
That's precisely why I'm starting the AI Technostress Institute—to ease the pain and ensure people become AI-productive without becoming AI-overwhelmed.