The Cost of Waiting: What 'We'll Look at AI Next Year' Really Means

Nobody's business dies the day they decide to 'wait and see' on AI. That's what makes waiting feel safe. The damage arrives as a slow leak: a competitor quotes faster because their proposal drafting is automated. Their support answers at 2 a.m.; yours opens at 9. Their three-person back office does what your seven-person one does. None of it makes headlines — all of it compounds.

Harvard's Karim Lakhani put it precisely: AI won't replace humans, but humans with AI will replace humans without. The same is true one level up — companies with AI will replace companies without, one customer decision at a time.

Humans With AI Will Replace Humans Without AI — Karim Lakhani, HBR

Harvard Business School's Karim Lakhani on why the divide isn't humans vs. AI — it's adopters vs. everyone else.

The math of compounding disadvantage

Suppose automation gives an adopter a modest 15% cost advantage and a 2× faster response time. In year one, that's margin they can spend on price, marketing or talent. In year two, the gap funds better tooling and better hires, widening it further. By year three you're not catching up to their software — you're catching up to three years of compounding decisions you didn't get to make.

There's also a talent dimension nobody prices in: skilled people increasingly choose employers whose tooling doesn't waste their time. The firms that automated the drudgery are winning recruits from the firms that didn't — which makes catching up harder still.

Waiting for the technology to 'settle' is backwards

Yes, models improve every quarter — that's an argument for starting now, not later. The durable assets you build are model-agnostic: clean data pipelines, integration points, evaluation sets, a team fluent in working with agents. When better models ship, adopters swap engines and accelerate; non-adopters start from zero against a faster field.

The rational move isn't a giant bet. It's a small, real one: pick a workflow where you bleed hours, automate it properly, measure the result, repeat. Six months from now you'll either have compounding returns — or a very well-rested competitor. The discovery call for figuring out which workflow to start with costs you nothing.