A $40k/year retainer buys ~6 hours of senior attention per week. An always-on agent runs 168. Here's the spreadsheet most founders never build.

Most founders compare agencies to agents on a single axis: monthly cost. That's the wrong frame, because it hides the unit you actually buy. An agency sells you hours of human attention. An agent sells you cycles of compute. The two are not interchangeable, and they don't scale the same way.

Run the math honestly. A mid-tier agency at $3,500/month, after account management, status calls, and slide decks, gives you roughly 6 hours of senior strategist time per week. The rest is junior production. That's ~$135 per genuinely strategic hour — and those hours are batched, not on-demand. You wait for the Tuesday standup to ship a Monday idea.

An agent runs 168 hours a week. It doesn't context-switch, doesn't onboard, doesn't sleep through a launch in a different timezone. The marginal cost of one more campaign brief, one more ad variant, one more retargeting segment is effectively zero once the loop is wired up. That changes which experiments are worth running. Suddenly the 12% lift idea is worth chasing, because chasing it costs $0.40 in tokens, not a half-day from a $180/hr planner.

The honest counterpoint: agencies are still better when the work is novel, political, or requires taste only humans have. Brand repositioning, executive comms, a Super Bowl spot — pay for the humans. But research, drafting, ad ops, audience cuts, weekly reporting, creative iteration, post-mortems? That's loop work. Loop work is what agents are for, and paying agency rates for it is a tax on not having modernised yet.

The cleanest setup we see in 2026 is hybrid: one strategist (in-house or fractional) holding the brand and the bets, plus an agent running the loop underneath them. The strategist's leverage goes up 5–10x because they stop doing production. The agent's output gets sharper because a human is curating the inputs. That's the configuration the math actually rewards.