Friday, November 20, 2026

Top 5 This Week

Related Posts

Five Things Everyone Gets Wrong About Agencies Run by AI Agents

Most people picture an agency running on AI agents as the old agency with fewer people in it. The reality looks close to the opposite. The desks that disappear aren't the ones you'd guess. Billing changes before the org chart does. And the account manager, far from being automated away, ends up with a new job: keeping a roomful of software honest.

The reorganization is already underway at shops that bet early, and it has produced a tidy set of misconceptions worth taking apart one by one.

The Account Manager Didn't Get Easier. They Got a Pack of Agents to Supervise

The tempting read is that when software drafts the status update, pulls the numbers, and writes the first version of the recommendation, the account manager's day gets shorter. At the shops actually doing this, the day gets denser. The role shifts into checking work instead of producing it, and deciding which of the deck's claims are safe to send instead of writing the deck.

That pattern lines up with how agent orchestration is being described in field research across large companies: software handles the task, humans set guardrails, hold context, and make the final call. Adweek's contributor coverage of what the agent-driven agency model looks like in practice describes the same pattern from the agency side: direction and judgment stay human, execution does not.

Flat Fees and Outcomes Replace the Hour Before the Software Is Fully Deployed

A common misread is that pricing changes last, after the workflow is rebuilt. It changes first, because the hour stops making sense the moment a reporting deck that used to take five hours takes five minutes. Billing a client for ninety minutes of work on something the agent finished during the standup is a conversation no account lead wants to have twice.

What replaces the hour varies by workstream, and the shops furthest along have stopped pretending one model fits everything.

  • Flat monthly fees. Used for recurring outputs like reporting, always-on content, and paid media management, where volume is predictable and the agent carries most of the load.
  • Project pricing. Used for scoped builds like a campaign launch or a site refresh, where the deliverable is defined and the agent speeds parts of it but humans own the shape.
  • Outcome-based pricing. Used where the result is measurable — qualified leads, pipeline, bookings — and the agency is willing to tie revenue to the number rather than the hours behind it.

Not Every Workflow Is a Good First Candidate

Handing the agents whatever is most annoying is how shops get burned. The right first candidates share three traits: the work is repetitive, the output is checkable, and a mistake is embarrassing rather than catastrophic.

Reporting fits. Content production at scale fits. Paid media operations and outreach fit, with guardrails.

Client conversations don't fit. Disclosure decisions don't fit. Creative choice and brand taste don't fit either, because the cost of a confident wrong answer is a client who stops trusting the output. Those stay with the humans, and the account manager's calendar fills up accordingly.

Calling an Agent a Junior Teammate Is a Governance Mistake

The language around all this has drifted toward "your new teammate," and it causes real problems. An agent is a tool with a long leash. The person holding the leash is the one who signs.

That matters most at the approval gate. Three gates show up in almost every working setup:

  • The factual gate. Someone checks that the numbers, names, and claims in the agent's output match the underlying data before anything leaves the building.
  • The brand gate. A senior reviewer checks tone, positioning, and anything a client would recognize as voice — the places a confident wrong answer does the most damage.
  • The disclosure gate. A named human decides what the client is told about how the work was made, and signs off on the record that says so.

Headcount Does Shrink, and the Senior Bench Gets Deeper

The last misconception is the loudest: that the agency of the near future is a skeleton crew of generalists steering a rack of software. The shape is different. Entry-level production work thins out because the agents are good at it. Senior judgment — strategy, creative direction, client trust, the call on what to publish — becomes more valuable, not less, because the volume of decisions going through the gate goes up.

The account manager who adapts to this isn't managing fewer things. They're running a mixed team of people and software, chairing the approval gate, and spending the hours they got back from the five-hour deck on the conversations the agents can't have.

That's the job now. The work hasn't gotten lighter; it's moved further up the stack.

Popular Articles