Enterprise software is racing to bolt AI onto every surface it can reach, and Dovetail Software co-founder and CEO Benjamin Humphrey has a name for the pattern that predates the current AI cycle by years.
Inchification Problems Didn’t Start With AI
Humphrey calls it “inchification,” a term he borrows for the slow creep of features that serve the business rather than the customer. “There’s the whole inchification trend, where Samsung puts advertisements on their TVs, for example,” he says. “My family tries to buy things all the time, and I’m like, don’t click that.” The example is a television, not a piece of enterprise software, but the mechanism he’s describing, a feature added because it can be monetized rather than because anyone asked for it, is exactly what critics of the current AI rollout are pointing at when they call an AI chatbot bolted onto a settings menu “AI-washing.”
Humphrey doesn’t treat this as a hypothetical. “You’re always fighting the commercial realities of the business versus the user needs,” he says. “That’s a tension every business faces: do they commercialize, or make the best product for customers? The best companies are the ones that can figure out how to do both. Apple, for example, they have good products, but they’re also very successful and make a lot of money.”
Data Behind the AI-Washing Complaint
That framing gives the current wave of enterprise AI rollouts a specific test: which side of the commercial-pressure line a feature falls on, no matter how much AI sits inside it. Deloitte’s 2026 State of AI in the Enterprise report puts a number on how many companies are failing that test. Thirty-seven percent of organizations are applying AI at a surface level, with little or no change to their underlying processes, and eighty-four percent haven’t redesigned a single job around what AI can now do. A chatbot sitting on top of an unchanged workflow is treated by the customer the same way an ad on a TV screen is: technically new, functionally decorative.
Buyers have started to notice, and the skepticism shows up in how long they now take to say yes. Internal resistance to AI adoption, cited as a concern by enterprise software buyers, climbed from 16% to 29% in a single year, and almost half of buyers report a CFO vetoing an already-approved purchase in the past twelve months. Evaluation cycles are stretching precisely because a feature announced at launch is no longer taken on faith. Buyers are doing the work of separating the Apple approach from the Samsung one themselves, one procurement review at a time, and vendors that once relied on a demo to close a deal are being asked to prove the feature earns its place.
Separating a Real AI Feature From Investor Theater
Humphrey’s own test isn’t complicated: does the feature make the business more money, the product better for the person using it, or both? A feature that clears only the first bar is the AI-era version of a TV ad, defensible on a quarterly earnings call and resented by everyone who has to look at it. A feature that clears the second without the first is a nice idea that won’t survive a board meeting. The companies Humphrey holds up as the standard, Apple among them, are the ones that never had to choose.
None of this argues against building fast or shipping AI features quickly. The case is against treating “we have AI” as the achievement itself, rather than the starting condition for a harder question. What matters is what the feature changes for the customer, not what it changes for the press release. Dovetail’s own bet is that the products built to pass Humphrey’s test, useful before they’re profitable, are the ones still standing once the current round of AI features gets judged the way inchification eventually was: as something users tolerated rather than wanted. The features that survive that judgment tend to be the quiet ones, the kind a customer notices only when trying to imagine the product without them.