Earlier this month, Anthropic, one of the best-known rivals to ChatGPT maker OpenAI, released a set of plugins for Claude, its flagship language model. The plugins shipped alongside Opus 4.6, the latest iteration of Anthropic’s most capable AI.
At least since the release of DeepSeek-R1, investors have mostly shrugged at new model launches. This time was different. Anthropic’s new plugins more overtly target white-collar knowledge work, including finance, law, and product management, rather than just software development. Until now, developers were the group that most visibly felt the heat. Software stocks shed $800 billion in value over a few days, with software-as-a-service (SaaS) companies hit hardest.
I think the decline is an overreaction.
Software companies make money by providing a useful service, not by writing code. If anything, I expect many software companies to become more profitable over time as the cost of iteration and experimentation drops. The same logic applies to a lot of knowledge work. Productivity gains usually create more winners than losers, even if there is real short-term dislocation along the way.
So what do Claude’s new plugins, and the market’s reaction, mean for small and mid-sized businesses?
First, a caveat: it’s easier to measure AI’s impact on my own profession (software) because the quality of output is easier to verify. Either the code compiles, runs, and passes tests, or it doesn’t. In many other professions, the feedback loop is slower, and “looks right” can masquerade as “is right.”
With that said, the impact of agentic products is likely to be massive outside of engineering too. The Claude plugins are essentially a user-friendly wrapper around tool integrations that have existed for Claude Code users for a while, but marketed to non-developers. It is a clever step by Anthropic, which has been on a bit of a tear when it comes to dominating the application-layer of the AI industry.
Now, the jury is still out on just how large the productivity gains from coding agents really are. Some research suggests the real gains are smaller than people feel they are. But even if the gains were somehow zero (they aren’t) while quality stayed the same, the job satisfaction improvement alone would be enough to justify widespread adoption.
That’s why CEOs and business leaders are right to encourage or require AI adoption in daily work, including tools like Anthropic’s plugins. In my experience (both in development and consulting), many employees are reluctant, not because they’re irrational, but because their mental model of AI is stuck in 2023. In the AI world, that might as well have been a decade ago.
The catch: blanket mandates with no direction won’t solve this.
Instead, executives need to do two things:
Get current on capabilities. Not by reading hype, but by seeing what the tools can do on real tasks inside your business. If leadership won’t do it, appoint someone who will.
Identify where value actually lives. Where are the high-frequency, high-friction workflows? Where are the handoffs? Where do documents pile up? Where do projects stall because someone can’t find an answer, reconcile data, or draft the first version?
That evaluation can be difficult because it cuts across systems, teams, and incentives. Sometimes outside help is worth it, not because you can’t learn, but because you don’t have time to run dozens of experiments, build a safe rollout plan, and translate capabilities into value.
In that light, Anthropic’s plugins are also an attempt to do part of the work for companies by packaging an agentic workflow into something non-technical teams can use. Early adopters of Claude Code discovered that agents are useful well beyond writing code. Anything that happens on a computer, touches multiple tools, and involves coordination across web services is a candidate. Managing a Trello board. Basic system administration. Digging through logs to find obscure issues. Pulling information from across documents and turning it into an action plan.
The opportunity for SMBs is straightforward: if you can reduce the cost of coordination, drafting, searching, and execution, even by 10 to 20 percent, you can compete like a much larger organization. The risk is equally straightforward: if you roll these tools out haphazardly, you’ll create quiet failure modes (hallucinated facts, inconsistent outputs, brittle workflows, compliance problems) that don’t show up until they hurt you.
The businesses that win won’t be the ones that “use AI.” They’ll be the ones that operationalize it: clear use cases, guardrails, training, measurement, and iteration.
If you made it this far, thanks for reading. If you’re a business leader looking to apply some of the principles from this newsletter, click here. I’d love to hear how I can help.
If you found this newsletter useful, forward it to someone who would too. Click here to learn how Applied.AI can help your business make AI work for you. I use AI to aid my own research and editing, but not to form my opinions.
