FIELDNOTES · AN OPERATING BRIEF

Where the new marketing jobs come from

Everyone can name the jobs AI takes. Almost no one can name the jobs it makes. I have worked through four supply shocks in twenty-five years of marketing, and each one expanded demand faster than it cut it. This is my read on where the work goes in the fifth.

As production automates, the value moves up to the brand and translation layer. That is where the new jobs are.

The demand-frontier framework is Nathaniel Whittemore's. The marketing translation, the three seats, and the operator read are mine. The source.

01The blind spot

The debate is being told from one side

The displacement list gets budgeted because every item can be named. Copywriters, designers, ad ops, the junior media and analytics bench. The creation list arrives as a shrug, or a title nobody can define yet, so it never makes the plan. The displacement story wins by default, and not because it is the truer one.

What AI takes

Copywriters
Designers
Ad ops
Junior media planners
Performance analysts
SEO specialists

What the debate calls back

“AI strategist”
“Prompt engineer”
“Brand AI lead”
The list runs out here. Too vague to cost, so it never reaches a budget.

The other half has a shape. No one has drawn it yet.

The pattern that always holds

This is the fifth supply shock I have worked through. Each one expanded demand faster than it cut it, and each time the value moved up a layer from where the disruption hit.

The webEARLY 2000s

A new surface created new demand. Value moved up from hand production to digital strategy.

ProgrammaticMID 2000s

Automation displaced manual media buyers. New roles emerged above the disruption, in trading and optimization.

MobileEARLY 2010s

Another surface, more demand, and a set of roles that did not exist the year before.

Social2010s

Community managers, social strategists, influencer leads. None of these were job titles when the shift began.

AINOW

The same pattern is repeating. The question is not whether demand expands. It is where.

Three things the last four shocks did not have

The pattern holds. Three features still set the fifth shock apart from the four before it.

01 / MACHINE TO MACHINE

The work talks to agents

Customer agents will book, transact, and negotiate for people. Brands need their own agents to answer. A brand that cannot speak agent-to-agent is not in the conversation, and that alone is a category that did not exist eighteen months ago.

02 / NOT VERIFIABLE

You cannot test on live customers

You cannot A/B test a hundred thousand bad emails on a real audience. A wrong agent decision is a viral screenshot. Synthetic customers and digital twins become the test bed, and that work needs people to design it and judge it.

03 / NOT ONLY THE TOP

The long tail switches on

The same compression that runs always-on programs for the enterprise makes professional marketing affordable for the dental practice and the regional restaurant group. The activated long tail is the bigger half.

02The demand frontier

Demand stretches six ways

The framework is simple. When a kind of labor gets cheap, demand does not vanish. It stretches in six directions, and two forces do the stretching: affordability brings a service to buyers who were priced out, and possibility makes service models viable that could not run before.

01

Price

New buyers who could not afford the service before can now access it.

02

Access

Overcoming scarcity that was geographic, temporal, or structural.

03

Complexity

Navigating opaque systems that used to require a specialist intermediary.

04

Continuity

Moving from episodic, project-based work to always-on service.

05

Personalization

Customized per individual rather than produced per segment.

06

Value per hour

The human moments rise in price as automation makes the routine ones rare.

Marketing scores five of six

Run marketing through the six and it stretches hard on five, with one moderate exception.

PriceHigh
AccessHigh
ComplexityHigh
ContinuityHigh
PersonalizationHigh
Value per hourMod
5/6
Five of the six directions stretch hard for marketing. The exception is value per hour, where the lift is only moderate. The other five are the engine.
03Three openings

Three places the new demand shows up

Two forces do the stretching. Affordability brings a service to buyers who were priced out, and possibility makes a model viable that could not run before. A third opening is newer still, where the buyer becomes a machine. Each one expands demand in a different direction.

Affordability

The SMB long tail

The cost floor drops, and thirty million businesses that always had the demand can finally buy. The compression at the top end is real and overstated. The activated long tail is the bigger story.

Out of reach
  • $500K agency engagements
  • $5K projects
  • Brand strategy, media, CRO, lifecycle, and measurement, all walled off
Now affordable
  • $50K AI-augmented engagements
  • $500 projects
  • The same services, within reach of the tail
30M
US SMBs, never agency clients
$500K → $50K
engagement, compressed
Service-as-software
replaces billable hours

Federal policy is pulling it forward: the AI for Main Street Act routes SBA financing toward these tools. New roles: AI marketing coordinators serving fifty to two hundred small clients each, embedded with SBDC and SCORE chapters.

Possibility

Continuous brand programs

The informational layer around marketing falls to near-zero cost, and a model that was always too expensive becomes viable. Same brand, different operating rhythm.

Episodic
  • Quarterly briefs and annual cycles
  • Seasonal creative refreshes
  • Human review at every step
Continuous
  • Real-time brand-health signals
  • Variants generated and tested hourly
  • Humans own the high-stakes calls
Hours
not quarters
1:1
at population scale
Conversational
two-way, always on

New roles: brand continuity leads, conversational program leads, lifecycle outcomes architects.

A new buyer

Agent-mediated commerce

Buyers hand discovery and purchase to their own agents, and demand expands into a dimension that did not exist: marketing to machines. The job becomes being found and trusted by other software.

Human discovery
  • Optimize for attention and emotion
  • SEO and human journeys
  • The brand talks to people
Agent discovery
  • Optimize for machine legibility
  • GEO and machine-readable feeds
  • Brand agent transacts with customer agent
90% by 2028
of B2B buying agent-mediated (Gartner)
$750B by 2028
consumer spend via AI search (McKinsey)
Share of Model
joins Share of Voice

New roles: GEO specialists, AI attribution analysts, brand agent designers, agent-to-agent negotiation specialists.

The value moves up a layer

Across all three openings, the same thing happens to the work. The production floor compresses, and the value concentrates one layer up.

Value rises
Strategic outcomesAnchors

The business result the whole system is pointed at. Always human, now exposed.

Brand stewardship & translationRises in value

The layer between business outcomes and the agent stack now running execution. Taste, cultural fluency, accountability, the human signature on the work. This is where the new seats are.

Production: copy, design, ad opsCompresses

The most rules-bound work in the building, and the first thing the agents absorb.

04The new roles

The roles the new layer needs

The new work concentrates in the translation layer. Its sharpest form sits inside the enterprise running continuous programs, where three senior seats appear that were never distinct jobs before. These are the clearest cases. The full set, across all three openings, clusters into the six families further down.

ROLE 01

Brand Continuity Lead

Owns voice, the cultural read, and the moments that matter. The agent layer makes a thousand variants an hour; a human makes the cultural call and approves the high-stakes work. One hallucinated output now reaches every customer at machine speed, so the cost of an unsupervised brand has never been higher.

Human premium: stewardship, taste, accountability
ROLE 02

Marketing Outcomes Architect

Translates a business problem (CAC, LTV, churn, share growth) into programs the agent fleet can run. Marketing is not a verifiable domain, so the fleet runs against synthetic customers and digital twins first. This role designs the simulation, judges the result, and owns the production call.

Human premium: translation, accountability, outcome ownership
ROLE 03

Data & Identity Operations Lead

Owns the data spine the agents act on. Identity is moving from a back-office capability to the real-time context layer for autonomous action. With no human review step in the loop, provable identity becomes the precondition for anything customer-facing.

Human premium: trust, translation, institutional legitimacy

The job math

500K–1M new roles in the continuous-brand opening alone, at maturity. The same order of magnitude as the healthcare case in Whittemore's original.

Read this as a model, not a forecast. Around 150,000 US enterprise and mid-market companies could run continuous, agent-led programs. Put two to five new seats in each, at three adoption rates, and the range falls out.

Conservative
~100K
50K companies × 2 roles
Middle
~300K
100K companies × 3 roles
Aggressive
~750K
150K companies × 5 roles

The ecosystem around these programs adds another 200K to 500K (agency roles, agent operations, GEO and brand-safety specialists). The SMB and agent openings sit on top of this, and are not counted in the figure above.

05What it means

Every new role, six families

The three senior seats are the sharpest cases. Across all three openings and every sub-sector, the full set of new roles clusters into six families. The titles will keep changing. The shapes do not.

1

Marketing navigators

Move clients and customers through complex systems.

GEO advisors · brand agent designers · customer agent navigators
2

Continuous brand workers

Stewardship around AI-monitored brand and customer systems.

brand continuity leads · conversational program leads · lifecycle outcomes architects
3

AI-augmented operators

Cheaper professional services for new market tiers.

SMB AI coordinators · agentic creative directors · agentic media operators
4

Data & operations specialists

Make the models reliable inside enterprise systems.

identity operations leads · brand context engineers · clean room operators
5

Governance, safety, compliance

Keep AI-mediated marketing safe, legal, and on-brand.

brand-safety auditors · agent compliance officers · consent operations
6

Escalation specialists

Handle the hard cases the agent layer routes up.

crisis comms leads · escalation managers · complex-account strategists

Three implications for leaders

1

Headcount is running off the wrong ledger

Most organizations are cutting against production while the translation layer goes unstaffed. Plan both sides now, or hire the new layer cold while competitors are already two years into operating it.

2

The agency is changing shape

The service-bureau model that sold hours for manual execution is ending. The translation between AI capability and business outcomes is the new agency, and the P&Ls built for it will win.

3

Identity is infrastructure, not a tool

The identity graph itself is commoditizing across the big data vendors. The defensible asset is the translation layer between that graph and the CMO's P&L.

The thesis

The production layer leaves. The brand layer becomes the seat.

What leaves

The production layer. Volume copywriting, templated design, rules-based media planning, mechanical optimization. These were always means, not ends.

What stays

Judgment, cultural fluency, strategic translation, accountability, the human signature. These were always the work. Now they are also the whole job.

The displacement story is real, and it is half the picture. The other half is where demand expands, which service models become viable, and which roles persist because the value being paid for never moved to the model.

The map is wide. It is ours to draw.

A personal note. If you are twenty-five in marketing right now, the translation layer is where I would put my money. Not the tools. Not the prompts. The judgment that sits above both.

A note on the numbers. The six-direction read and the job math are my models, not measurements. The market figures, agent-mediated buying and AI search spend, come from the cited sources. The price compressions are illustrative, drawn from current agency and tooling economics.

Framework: Nathaniel Whittemore, “AI Jobs and the Demand Frontier,” The AI Daily Brief, May 2026. Marketing application by Tony Weber. Sources informing the application include McKinsey on agentic marketing workflows and AI discovery, Gartner on agent-mediated B2B buying, Harvard Business Review on redesigning the marketing organization, and reporting from IAB, MarTech, AdExchanger, and Search Engine Land. Listen to the source episode.