THINXSTER
Blog/AI Agency
AI Agency9 min readAugust 4, 2026

AI and Advertising Agencies: What's Really Been Automated, and What Clients Should Demand Now

AI has eaten specific parts of agency work and left others untouched. The real map, the margin problem it created, and four things to renegotiate this quarter.

RK
Ryan Korsz
Founder & CEO, Thinxster

TL;DR

AI has eaten specific parts of agency work and left others untouched. The real map, the margin problem it created, and four things to renegotiate this quarter.

→ See how this applies to your business (free 30-min call)

The conversation about AI and advertising agencies is stuck between two useless positions: agencies are finished, or nothing fundamental has changed. Neither describes what's actually happened, which is more specific and more interesting.

AI has thoroughly automated a set of tasks that agencies used to bill significant hours for. It has barely touched a different set. And the boundary between them explains almost everything about which agencies are growing and which are quietly losing accounts.

What Has Genuinely Been Automated

Media buying execution. The platforms did this to themselves. Automated bidding, broad targeting, and campaign types where the algorithm makes the audience and placement decisions have removed most of the manual optimization work. The old craft of manual bid adjustments and granular audience carving produces worse results now than letting the system optimize with good inputs. That was billable work. It largely isn't anymore.

Creative variation. Producing thirty versions of an ad across formats, aspect ratios, and hook variations used to consume a designer's week. It's now a fraction of that. Note the precision here: *variation* has been automated. The original concept has not.

Reporting. Assembling numbers into a monthly deck was a real cost line at most agencies. Automated dashboards and generated commentary have made it nearly free. Any agency still billing meaningfully for reporting production is billing for a solved problem.

Copy at volume. Ad variants, headline testing, description lines, meta descriptions. Fast and adequate, which is exactly what volume testing needs.

Campaign QA. Checking naming conventions, tracking parameters, budget pacing, and setup errors is rule-based work that machines do more reliably than tired humans on a Friday.

Research synthesis. Competitive scans, review mining, audience research. Hours to minutes.

That's a substantial share of what a mid-tier agency's junior and mid-level staff used to spend their weeks doing.

What Hasn't Moved

The offer. What you sell, to whom, at what price, with what guarantee. This determines more of your campaign performance than any targeting decision, and no model has meaningful access to the judgment involved — it requires knowing your margins, your capacity, your competitors, and your appetite for risk.

Positioning and the original creative idea. Models produce competent variations of concepts that exist. The concept that reframes a category comes from somewhere else. This is the most durable agency skill and the least evenly distributed.

Reading a client. Knowing which stakeholder actually decides, when to push back, when the stated problem isn't the real one. Agency work is substantially relationship management, and that hasn't changed at all.

Accountability. A model can't be responsible for an outcome. Someone has to answer for the number, and that has value independent of who did the work.

Cross-channel judgment under constraint. "We have $40,000 and eight weeks to fill a service calendar in three markets" is a strategic allocation problem where the trade-offs are business trade-offs, not media ones.

AI automated the production of advertising. It didn't automate deciding what to advertise, to whom, or why anyone should care.

The Margin Problem This Created

Here's the structural bind. The traditional agency model recovers costs through hours — billed directly, or bundled into a retainer sized around headcount. When the hours required to deliver the same output collapse, that model punishes the agency for improving.

Three responses, in rough order of how well they're working:

1.

Absorb the gain quietly. Keep the retainer, do the work faster, improve margin. Works until a client asks what they're paying for, or a competitor prices honestly.

2.

Reprice to outcomes. Tie fees to booked appointments, qualified leads, or revenue. Harder to sell, harder to measure, and the only model that survives further automation. Agencies doing this are growing.

3.

Expand scope into what wasn't previously offered. Take the recovered hours and put them into things clients always wanted and never got: response systems, attribution infrastructure, sales process work. This is where the genuinely interesting agencies have gone.

The agencies in trouble are the ones doing the first while telling clients they're doing the third.

Four Things to Renegotiate at Your Next Review

If you're the client, the automation has shifted leverage toward you. Use it specifically.

1. Reporting should be free and continuous. You should have a live dashboard, not a monthly deck. If a meaningful share of your retainer covers report production, that's a line item to remove. Redirect it toward testing volume.

2. Creative testing volume should have increased substantially. Variation production got roughly an order of magnitude cheaper. If your agency is running the same number of creative tests as three years ago, you haven't received any of that gain. Ask directly: how many distinct creative concepts and how many variants ran last quarter, versus the quarter before?

3. Response time should now be part of scope. This is the biggest one, and the one most agencies still treat as your problem. They generate the lead; it sits in your inbox for four hours; conversion craters. The technology to close that gap is available and affordable, and an agency that generates leads without owning what happens in the first five minutes is delivering half a service.

4. Attribution should be non-negotiable. Ask which channel produced your best customers last quarter. If the answer is platform-reported conversions with no closed-loop data from your CRM, you're being reported to, not measured.

What Agency Roles Look Like Now

If you want to see whether an agency has adapted, look at who's on the account rather than what's in the proposal. The composition has changed in specific ways:

  • The media buyer's job moved from execution to inputs. Feeding the algorithm good conversion signal, clean audience seeds, and a steady supply of creative — rather than manual bid and placement management. An agency still describing granular manual optimization as their edge is describing a skill the platforms deprecated.
  • Creative direction split from creative production. Production compressed; direction didn't. The valuable person is the one deciding which three concepts to test, not the one producing thirty variants.
  • Analytics became a senior role. Closed-loop attribution — connecting platform data to CRM outcomes — is now harder and more valuable than campaign management. It's also where most agencies are thinnest.
  • A new role appeared: whoever owns the response layer. Automation architect, lifecycle lead, whatever they call it. The person responsible for what happens between a lead arriving and a human speaking to them. Five years ago this role didn't exist in agencies. It's now the difference between generating leads and generating revenue.
  • Account management got harder, not easier. With execution automated, the conversation with the client is about strategy and results rather than status, and that requires more senior people, not fewer.
  • If the team on your account is three juniors executing and one senior selling, you're buying the pre-automation model at post-automation prices.

    The Question That Sorts Agencies

    One question, asked plainly: "What happens to an inbound lead in the first five minutes?"

    An agency that has adapted describes a specific sequence with timings — automated contact, qualification, booking, handoff, and what happens if the lead doesn't answer. An agency that hasn't describes where the lead is stored.

    That single answer correlates with everything else more reliably than any case study, because it reveals whether they think their job ends at lead generation or continues to booked revenue.

    90s
    how fast our AI callers reach every inbound lead, day or night

    What We Do With the Recovered Hours

    We're an agency by category and structurally not much like one. The work isn't running campaigns and reporting on them — it's building the system that converts spend into booked jobs: AI caller agents engaging every inbound lead within 90 seconds, running a real qualifying conversation, booking the qualified ones, and writing everything into a GoHighLevel pipeline where the chain from ad dollar to closed job is visible.

    9.2×
    peak ROAS — what happens when the response layer stops leaking

    The reason we can do that isn't superior media buying. It's that the automation freed up the capacity to build the layer everyone was ignoring, and that layer turned out to be where most of the money was leaking.

    $102M+
    client revenue generated on systems built this way

    The Honest Forecast

    Agencies aren't going away. The ones selling execution hours are, because execution is the part that got automated. What replaces them is a smaller number of firms selling judgment and systems — and clients who have finally stopped paying for decks.

    If you want a straight assessment of what your current agency is actually delivering versus what the automation should have already bought you, [book a free strategy call](/book). We'll tell you honestly whether the fix is switching or just renegotiating.

    Free Weekly Briefing

    One AI Marketing Tactic.
    Every Tuesday. Free.

    What's actually working across our client accounts right now — ROAS moves, follow-up sequences, creative angles. The stuff that isn't in any blog post yet.

    No spam. Unsubscribe anytime. 1,200+ business owners already in.

    Ready to Deploy

    SEE THIS IN
    YOUR BUSINESS.

    30 minutes. We scope the exact systems that apply to your situation and give you a plan.

    ★★★★★ Trusted by 47+ local service businesses

    BOOK A STRATEGY CALL →