A version of this ran in TechRadar Pro this week.

Most companies have run an AI pilot by now, bew have made it stick.
Boston Consulting Group puts the share of AI initiatives that scale across a business at roughly 15 percent. McKinsey finds a similar pattern: almost every company is experimenting with generative AI, but only a small share sees any real financial impact from it.

I don't think the technology is the problem. I think it's what happens after a company buys it.

General models are good at understanding language and information. They struggle to connect that to workflows, regulations, and the operational detail that runs an industry, whether that's accounting, law, healthcare, or manufacturing.

A model can read a tax rule, but it has no idea which step in a firm's year-end process exists for a good reason, and which step could actually better be skipped.

The gap between technology as a functional platform and technology as a game-changer for your business has created a new kind of job. Palantir calls its version the forward deployed engineer. Law firms have started hiring legal engineers. In accounting, the title we use is Accountancy AI Engineer: someone who knows the work well enough to sit next to a firm's staff and build the automation with them, instead of handing over a manual and walking away.

Being connected to the outcome, changes what we expect from a vendor. A product used to be judged mainly on whether it worked. Now we're expected to stay involved after the sale: training staff, designing workflows, helping firms avoid what's called "shadow AI," where employees happily use their own AI accounts, expose company data without meaning to, and build siloes of context and knowledge that don't help the employee nor the company.

There's also no clean finish line. Models improve every few months. A rollout treated like a one-time software install goes stale fast, and the vendors who keep their customers are the ones who keep showing up.

Every company will eventually have access to the same AI. What will set vendors apart is whether they give their customers more than access to an AI-enabled platform, and go the extra mile to provide a business with the knowledge, guidance and attention that is needed to make this wonderful new technology deliver.

It's not hard to guess where I'm betting on with Ravical.