I think quite a few software companies are living off the lag time between what agents can do and what their customers realise they can ask them to do. By agent, I mean an AI assistant that can carry out steps in software, including through a browser.

This especially interests me for people who are opinionated about the result they want. They know their work and can tell when something is wrong. They just weren't considered "technical". I suspect "prompting" sounds like another technical skill they need to learn, when they may already be able to give a useful brief and check whether the result meets it.

Imagine a marketing manager preparing a newsletter. They could brief Claude or ChatGPT much as they'd brief a freelancer: put this copy and these images into our usual template, prepare it for this group of customers, and show me a preview before anything goes out. They can ask for changes in the same ordinary language. They don't need to explain how the software works.

With browser access, the agent could prepare the newsletter in the service they already use. Their templates and subscriber lists stay put, and moving to a new AI marketing platform becomes harder to justify.

That is the lightbulb moment I expect more people to have: they can explain the job and let their agent handle the steps. An AI-native marketing platform might be overkill if their agent prepares the content and a straightforward newsletter builder handles subscribers and sending. Similarly, paying for a website builder (such as Wix, Squarespace, or Webflow) becomes harder to justify if their agent can build the simple site they need and ordinary hosting can put it online.

How much of an (AI) product's value survives when the customer's own agent can operate the tool they already have? If I'm paying for an agent anyway, another monthly subscription needs to save me enough extra effort to earn its place.

If I have to keep correcting the agent, a specialist tool that gets the job done consistently can still be worth paying for. Some of today's gap is reliability, as well as familiarity. As the execution improves (keep in mind that today's performance is as bad as it'll ever get), though, customers not knowing they can ask becomes a weaker reason to expect them to keep paying.

For companies selling that extra convenience, two responses come to mind. Neither looks easy.

If the product still saves enough hassle, one option is to charge a modest amount for that convenience. The bill is small enough that replacing a useful tool never feels worth the effort. The company has to be able to run on those fees, though. Lower prices leave less room to acquire customers or spend time supporting each one. This could require changing quite a bit more than the pricing page.

The other is to offer expertise and advice alongside the product. Adding a call with someone who mostly explains the software won't be enough. It would have to help customers make better decisions about their work, including helping customers decide what is worth doing in the first place. That advice has to be worth paying for even when the customer can ask their agent for advice too.

A company built to sell the same tool to thousands of people would need relevant expertise and time to understand individual customers. It would also need to charge enough for that involvement. That's a substantial change in how the business works.

If you're building a company, ask yourself: are you living off the lag time? As agents improve and your customers get better at using them, they may get the same result with tools they already have, or much simpler ones. Will your offering still have a strong footing? Figure out what they'd still pay you for, and start building towards that today.