The Hidden Costs of Delegating to AI

UpTrajectory Review

Nitin Nohria's HBR piece uses accounting firms as a live experiment in what happens when a professional service rushes to hand its core work to machines. The teaser promises a counterintuitive finding: faster execution does not lift organizational performance. That framing matters because most small-business operators still hear the same pitch — automate the routine, free your people, grow. Nohria's argument, drawn from firms that delegated heavily to AI, suggests the trade runs deeper than a simple time swap. The piece sits in a growing HBR current skeptical of AI-first management, and it lands while operators face real pressure to adopt tools they barely understand.

For a small-business owner, the stakes are immediate. You do not have a finance department to absorb a botched rollout or a layer of partners to catch errors before clients see them. When a five-person firm delegates bookkeeping, drafting, or client intake to AI, the owner becomes the quality-control department. Nohria's accounting-firm lens fits your world better than a tech-giant case study: these are judgment businesses where trust, accuracy, and relationships carry the revenue. If delegation to AI quietly erodes those, the P&L will not show it until a client leaves.

The genuinely contested point is whether speed itself is the problem. We agree with Nohria that faster output can mask slower thinking — a junior accountant who rubber-stamps AI work learns nothing, and a firm that ships flawed work faster merely fails at scale. But we are skeptical of any argument that treats human review as a free fix. Review takes time, and tired owners often skip it. The under-reported angle is incentive design: firms bill by the hour, so AI-driven speed can cannibalize revenue unless pricing changes. Nohria likely addresses this; the teaser does not.

Second-order effects cut unevenly. Large firms can retrain staff and reprice contracts; a solo operator often absorbs the learning curve alone. Clients may welcome lower fees at first, then punish errors harder because they trusted the machine. Junior staff lose the repetition that builds expertise, which hollows out the talent pipeline your business will need in five years. There is also a competitive squeeze: if rivals adopt AI and you refuse, you look slow; if you adopt carelessly, you inherit risks you cannot price. The hidden cost is rarely the software subscription — it is the rework, the lost client, the skill you stopped building.

Watch how accounting trade bodies and malpractice insurers respond; their guidance will signal where liability actually lands. In the next two quarters, track whether firms that leaned hard into AI start repricing or quietly rehiring reviewers. If you run a small shop, do one thing now: pick a single AI-delegated task, time-box a weekly human audit of its output, and log what the audit catches. That log will tell you faster than any HBR article whether your delegation is a bargain or a slow leak. Do not scale what you have not measured.

“Lessons from accounting firms reveal why faster execution doesn’t automatically improve organizational performance.” — Harvard Business Review

Takeaway: Audit one AI-delegated task weekly and log what human review catches before you scale it.