Who Will Know How Your Marketing Works in 2030? Change Futurist | September 29, 2026

Who in your organization will know how your marketing works in 2030?

Put a name on it. Then name the work that person does this year that will teach them.

Four pieces published on September 28 and 29 answer that question from four directions. Kristina LaRocca-Cerrone at Gartner predicts that most high-performing marketing teams will eliminate the bottom rungs of the career ladder by 2030. Mukul Saha, also at Gartner, predicts that 70% of enterprises will abandon agentic AI built by vendor engineers by 2028. Emily Manock at Marketing Week reported World Federation of Advertisers research showing hourly agency billing falling from 54% of brands in 2011 to 19% today. Adi Ignatius at HBR interviewed Chewy CEO Sumit Singh about running customer obsession through talent decisions and operating metrics.

Read them together and you can see three separate places where marketers used to build judgment, all moving out of the building at once. The junior role goes to automation. The agent build goes to the vendor. The production method goes to the agency. Chewy supplies the counterexample.

On Thursday I asked who owns the decision to stop producing. On Saturday I asked who pays when a model's recommendation turns out wrong. Both questions assume someone inside the organization has the judgment to make the call. This edition tests that assumption.

Rebuild the bottom rung before you remove it

LaRocca-Cerrone, VP Analyst in the Gartner Marketing practice, published the prediction on September 29. Gartner surveyed 1,303 senior leaders between January and April 2026. Eighteen percent of marketing leaders have already eliminated certain functional roles because of automation, 16% have created new roles, and nearly one-third have redesigned roles.

LaRocca-Cerrone names the risk directly. Leaders can confuse the automation of junior tasks with the end of any need for junior people, and the confusion shows up years later as a leadership pipeline with nobody in it. Her prescription has three parts. Hire for the ability to evaluate and improve AI-enabled work. Move new hires into auditing AI outputs, generating insights, and connecting recommendations to business outcomes from the start. Use coaching and mentorship so early-career marketers watch experienced colleagues weigh trade-offs.

The entry-level job did two things for you. It produced output, and it produced your next manager. When you automate the role, you remove the first at a saving your CFO will see in one quarter. You remove the second at a cost nobody budgets until you post a director role in 2029 and interview six people from outside who have never seen your exception logic.

It reminds me of what Sue Keith, Corporate Vice President at Landrum Talent Solutions, said when I interviewed her on The Agile Brand podcast: "How do you become a seasoned, strategic marketer if you never do the entry-level stuff?"

LaRocca-Cerrone's onboarding list reads close to the hybrid role I describe in Marketing Operations 3.0: inspect the reasoning path, spot the weak retrieval, catch the output that sounds plausible and makes no commercial sense. Her version hands that work to someone in their first year. The assignment works only when a senior person grades the audit. An associate who reviews AI output with no feedback learns one skill, and it is approving AI output.

In the framework I am building on top of Coherence at Velocity, humans and agents sit on the same team and humans stay accountable for direction. Accountability requires judgment, and judgment comes from repetitions someone else checked. Remove the repetitions and you keep the accountability on the org chart with nobody qualified to hold it.

Commentary: LaRocca-Cerrone gives CMOs a survey number for a trend most of them have already acted on, and her three prescriptions convert the entry-level role from production work into supervised evaluation work, which only holds if managers have time budgeted to supervise it.

Write the exit into the agent contract before you sign it

Saha, Sr Director Analyst at Gartner, published a prediction the same day: by 2028, 70% of enterprises will abandon agentic AI built through vendor forward-deployed engineering, the model where vendor engineers work inside the customer to build and deploy. His diagnosis is that these engagements fail structurally before they fail technically. Costs climb, and the customer cannot evolve the system alone.

He lays out three phases. Before signing, use vendor engineers only for problems that need deep product expertise or tight integration, name an executive sponsor accountable for business outcomes, and write deliverables, knowledge transfer, IP rights, and exit terms into the contract. During delivery, embed the vendor team with your own domain experts and end users, and use each increment to settle decision rights and the balance between autonomy and human oversight. At the end, execute the exit plan on schedule and refuse the extension someone proposes because your team is not ready.

Saha adds a second forecast worth reading twice. Through 2028, fewer than 20% of these engagements will turn recurring customer needs into capabilities in the vendor's core product. He calls the resulting risk "FDE washing," consulting services marketed under a newer label.

Saha wrote for software engineering leaders. Marketing leaders are buying the same model right now for journey orchestration agents, content operations agents, and service deflection. I have watched the martech version of this for years: the integrator stays through a third renewal because nobody inside can change the platform without them. The agentic version costs more, because the vendor team encodes your exception rules, your eligibility logic, and your brand constraints into the agent's context. In my AI capability framework, that work lives in the Memory/Context and Identity/Permissions layers, the conditions every pillar depends on. When the vendor engineers leave, those layers sit in code your team did not write.

Marketing leaders used to settle the build vs. buy decision at signature. Saha moves it to the exit date.

It reminds me of what Stephanie Trunzo, CEO of MERGE, said when I interviewed her on The Agile Brand podcast: "Never outsource your creativity, your ideas, or your critical judgment."

A vendor engineer can write the agent. Your people have to own the judgment the agent encodes, and they only own it if they sat in the room while the vendor encoded it.

Commentary: Saha converts a vague lock-in worry into a contract checklist, and the item marketing procurement most often skips is the named internal counterpart who pairs with the vendor team from week one and can run the agent on the day the engagement ends.

Price what your agency learns on your account

Manock reported WFA research on September 28 tracking how brands pay their agencies. Labour-based models, billed by the hour or the day, fell from 54% of brands in 2011 to 33% in 2022 to 19% now. Fixed-fee and output models rose from 20% to 33% over the same fifteen years, and labour-plus-performance models more than doubled, from 9% to 23%. Looking ahead, 63% of brands expect to increase performance-based fees, 46% expect more value-based models, and 36% expect more fixed-fee or output arrangements.

The shift makes commercial sense. Paying for hours rewarded the agency for taking longer, and paying for outputs rewards the agency for getting faster, which in 2026 means more AI in the production method.

Follow the learning, though. Under hourly billing you saw the hours, including the junior hours where the agency's people learned your category, your customers, and your approval quirks. Under output pricing the agency owns the method, and the agency's own entry-level layer faces the same math LaRocca-Cerrone describes. So the brand buys outcomes from a process it cannot inspect, staffed by a pipeline neither side is training, and evaluates the result with an internal team whose own junior layer the company just automated.

Manock's roundup also includes a consumer figure that raises the stakes on evaluation. In Kinsta's survey, 58% of consumers trust content less once they suspect AI generated it.

I cover the mechanism in the wiki entry on absorptive capacity. An organization's ability to recognize the value of new information and apply it depends on prior related knowledge. A team that has never built a campaign cannot tell whether an output-priced campaign is good. It can only tell whether it arrived on time.

Commentary: Manock's roundup documents a fifteen-year procurement trend that moves agency accountability toward results, and the clause to add at the next renewal is a method-transparency requirement that shows your team how the agency produced the output.

Treat customer obsession as a staffing decision

Ignatius hosted an HBR Executive Live conversation with Chewy CEO Sumit Singh, published September 29. Singh describes customer obsession as a bet that trust and loyalty drive repeat business and advocacy. Chewy runs that bet through leadership principles, talent decisions, operating metrics, and a willingness to fund investments whose full return shows up late. The conversation also covers Chewy's move from online retailer toward an ecosystem spanning commerce, health, services, and technology, and asks how the company adapts as AI mediates more of the customer experience.

Chewy built its reputation on hand-painted pet portraits and handwritten condolence cards. Each one started as a judgment call by an employee Chewy had hired and trained to make it.

That is the capability at stake when an AI layer takes over the interaction. In Coherence at Velocity I describe the moment an enterprise introduces agents as the moment it discovers how much of its coherence depends on memory: the experienced operator who knows which exception to grant, which field to distrust, which customer needs a different answer than the workflow suggests. If nobody codified that judgment, the automated system becomes literally wrong at scale. If people who understood the judgment codified it, the system extends it.

HBR's summary lists talent decisions next to operating metrics. I built the same pairing into the Agile Brand principles: focus on the relationship over the transaction, and keep learning. A relationship at scale needs people who learned what the relationship requires, and they learned it by handling the cases.

Commentary: Ignatius and Singh show a retailer treating customer obsession as an operating system with a talent input, which gives CMOs a model for the AI transition: codify the judgment your best people exercise, and keep hiring people who can exercise it.

Run the numbers before you cut the rung

A $900 million B2B software company runs a 55-person marketing organization. Fourteen of those people sit in associate roles: campaign builders, list and segmentation analysts, QA, reporting. Each year about three associates move into manager roles, and the manager bench of twelve loses two or three people to attrition. The company fills most manager openings from inside.

This year the CMO makes three moves at once. She automates the associate layer down to four people, cutting ten roles at roughly $110,000 fully loaded, a saving of $1.1 million a year. She moves the creative agency to output-based fees. She signs a vendor to build a lifecycle orchestration agent with forward-deployed engineers.

Now follow the numbers into 2028.

The vendor engagement reaches its planned exit at month twelve. Nobody inside can modify the agent's eligibility and suppression logic, because the four remaining associates spent the year running campaigns and never sat with the vendor team. The company extends the engagement at $650,000 a year. Over three years, seven or eight manager seats open. The four associates produce three internal candidates. The company hires five managers externally, and each one costs about $120,000 in fees and six months of reduced productivity while they learn the exception logic the departed associates would have known. Spread across three years, that adds roughly $200,000 a year.

Net annual saving from the original $1.1 million: about $250,000.

Run the alternative. The CMO keeps six of the ten roles and redesigns them. Two associates pair full-time with the vendor engineers and own the agent's context layer at exit. Two own the evaluation of agency output against a written quality standard. Two maintain the exception log for lifecycle and service handoffs. Each redesigned role has a manager who reviews their audits weekly.

The saving drops to $440,000 a year. The vendor exits on schedule at month twelve. The manager pipeline refills from inside, and the people moving up in 2029 spent three years learning the logic they will supervise.

The two plans save similar money. Only one of them leaves the company able to maintain its own agent and promote its own managers.

Take these four moves into Q4 planning

Map where your people build judgment today. List the roles where people learn your customers, your exceptions, and your approval logic by doing the work. Check each one against your automation roadmap before the roadmap reaches it.

Rewrite entry-level roles around supervised evaluation. Assign audits of AI output, exception handling, and insight generation, and budget manager time to grade that work weekly. An unreviewed audit trains approval.

Name an internal counterpart on every vendor-built agent. Write knowledge transfer, IP ownership, and an exit date into the contract, and staff a person who pairs with the vendor engineers from week one and owns the agent's context layer at exit.

Add method transparency to output-based agency contracts. Keep paying for results, and require the agency to show your team how it produced the output so your people keep the ability to judge it.

I was a photography major before digital cameras were widespread. Cameras already had automatic exposure. I learned what exposure meant in the darkroom, printing the same negative over and over until I could see why one print worked and the next did not. I never learned that from the camera's automatic settings, and they gave me no way to know when to override them.

Your 2030 marketing leaders are somewhere in your organization right now, or they are not. Find them, give them work that teaches judgment, and have someone check it every week.

Featured this cycle:

1. Gartner Predicts AI Will Allow the Majority of High-Performing Marketing Teams to Eliminate Traditional Bottom Rungs of the Corporate Ladder by 2030. Kristina LaRocca-Cerrone, Gartner, September 29, 2026

2. Gartner Predicts 70% of Enterprises Will Abandon Agentic AI Built by Vendor Forward-Deployed Engineering by 2028. Mukul Saha, Gartner, September 29, 2026

3. Agency pay, AI research, advertising's economic impact: 5 interesting stats to start your week. Emily Manock, Marketing Week, September 28, 2026

4. How Chewy Is Evolving Customer Obsession for the Digital Era. Adi Ignatius with Sumit Singh, HBR Executive Live, September 29, 2026