The sovereignty of your firm

There’s a simple way I think about my role as a CEO. I have a fiduciary obligation to investors and shareholders to increase the top line and reduce the bottom line of our business. Revenue should go up, costs should come down. Most decisions can be framed through that lens.

In our business, that plays out quite cleanly. Software engineering is fee-earning. When we build new features, we sell them. It directly drives revenue. Legal, on the other hand, is fee-burning. It’s a cost. So the rational approach is to protect and invest in engineering, and to try to get legal as efficiently, and as cheaply, as possible.

If you just look at the company in those terms, the conclusion is obvious.

The complication is that one of our core markets is selling the product of our fee earning (software features) to private practice and in-house lawyers. So the thing I experience internally as a cost is, for our customers, their "fee earning" product. Legal isn’t just a function to them, it’s how they make money (either fees as private practice, or by providing the service to a company as in-house).

Therefore, if people feed legal problems, prompts and proprietary data, into systems that learn from them, like the frontier model providers, those systems improve. As they improve, they start doing work that lawyers could otherwise charge for.

That might not show up immediately, but it could compound. More capability may mean less billable work, or at least pressure on pricing. That flows through to how much money law firms and in-house teams have, and therefore how much they can spend.

So actions that look like small efficiency gains internally can actually add up to weakening the market you sell into.

That same logic doesn’t apply everywhere though. In areas like accounting, where I’m only a buyer, I don't sell Jylo into that vertical, there’s no real downside for us. If those costs come down, it’s just a benefit. So the outcome is different.

The problem however, is that every company makes these decisions locally. Reduce what costs you money. Protect what makes you money. But those categories are different for everyone.

The macro point being, if you look at this across all white collar work, this creates a system where most types of work are constantly being pushed towards zero. Not because of one big decision, but because of lots of small ones.

The problem is there’s no boundary. The same process that reduces costs in one area will, over time, affect others.

So the question is where you draw the line, and what you choose not to feed into that system.

With the release of Claude Opus 4.6, the genie seems to be peeking out of the bottle. You can clearly see that the model is starting to be trained on Enterprise data.

The question is whether we keep feeding it, or have the discipline to put the cap back on.

Jylo is one of the few AI companies who are not incentivised to compete with customers, and have rigorous processes in place to ensure customer data is protected from leading frontier labs.

If you've got this far, you might want to book a demo.