Scalewell / Blog

How to build a B2B marketing team in an AI year: which paid media roles to hire, retrain or stop hiring

For every open seat on a B2B paid team, decide whether to hire, retrain or stop hiring, using how much of the seat is judgment and how much is production AI now does.

A marketing leader sorting role cards into three trays while an AI assistant processes reports behind them.

Building a B2B marketing team in 2026 starts with each open seat, not the org chart: for every role, decide whether to hire, retrain someone into it, or stop hiring for it because AI and the ad platforms now do most of its work. The test is how much of the seat is judgment and how much is repeatable production.

The decision usually arrives as a requisition. Someone has resigned, or the plan says the team grows by two, and HR sends over last year's job description for a PPC specialist. Before you post it, finance asks the question that's become standard in 2026: couldn't AI do some of this? You need an answer that's better than yes or no.

This guide is for the CMO, VP of Marketing or head of demand gen who signs off headcount. It covers what the hiring data says, a hire, retrain or stop test for each seat, the roles worth adding, a filled-in hypothetical for a six-person team, how to test AI skills in an interview, and what goes wrong. Who owns which AI job is covered in the AI marketing team, and how to train the people you keep is in AI training for marketing teams; this article is the hiring decision itself.

Marketing leaders are spending more on people, and still short of skills

Headcount isn't shrinking the way the AI headlines suggest. In Gartner's 2026 CMO Spend Survey of 401 marketing leaders, labor's share of the marketing budget rose from 21.9% to 24.5%, while 57% agreed their department lacked the talent to execute its 2026 strategy, as Chief Marketer reported on June 24, 2026. The same survey found a lack of internal talent was the most-cited barrier to AI-driven marketing efficiency.

So the money is there, and the gap is skills. That changes what a requisition should ask for. A team that hires the 2022 version of a role and hopes the person picks up AI later is paying 2026 salaries for a job description that's half automated. Getting the repeatable production out of each seat is where AI hands a paid team its hours back before anyone hires, and Scalewell partners with paid and demand gen teams to make that shift, from strategy and planning through implementation and operations.

Employers have already shifted what they screen for. In Microsoft and LinkedIn's 2024 Work Trend Index, 66% of leaders said they wouldn't hire someone without AI skills, and 71% said they'd rather hire a less experienced candidate with AI skills than a more experienced one without them (Microsoft WorkLab). We'd be careful with that second finding in paid media, for reasons the interview section explains.

Every open seat gets a hire, retrain or stop decision

The hire, retrain or stop test sorts each seat by what its work has become. Ask four questions about the role as it runs today, using the last month of the person's actual week rather than the job description.

  1. What share of the seat's hours is repeatable production? Building reports, pulling exports, drafting ad variants, checking pacing, uploading lists. AI and the platforms now handle much of this.
  2. What share is judgment that changes spend or pipeline? Choosing what to test, deciding where budget goes, reading results against the CRM, signing off claims.
  3. Could someone on the team learn the new version of the seat in about six weeks? If yes, retraining is usually faster and cheaper than hiring.
  4. Does the seat own a recurring AI job or the data under one? If it does, the seat is more important than before, not less.

Then apply the rule, counting a seat as mostly one kind of work when more than half of its logged hours fall there:

What you findDecisionWhat changes
Mostly judgment, owns an AI job or its dataHire, with AI skills in the requirementsRewrite the job description around decisions and the AI job it owns
Mixed, and someone on the team could grow into itRetrainFund about six weeks of role-based training and backfill the production work with AI
Mostly production that AI or the platforms now doStop hiring for itMove the work into AI jobs with a named owner, and put the budget toward a seat that needs judgment
An empty chair at the foot of a branching path that splits into three routes: a new hire's desk, a retraining desk and an automated workflow.
Every open requisition ends in one of three places: hire, retrain or stop.

The rule is deliberately about seats, not people. Stopping hiring for a seat usually means not backfilling it when someone leaves, or moving a person into a different seat with training. It's rarely a reason to remove someone doing good work today.

The roles worth adding are builders, measurers and strategists

The seats that grow in an AI year are the ones that build, check and decide. Three matter most in a B2B paid team, though the titles vary.

The marketing technologist or AI workflow builder. A marketing technologist is someone who understands both the marketing work and the systems, and can build and maintain the automations, connectors and AI jobs around it. Scott Brinker has written about the role on chiefmartec since 2008, and on October 29, 2026 he noted that marketing technology is now mainstream enough that you can major in it. In a paid team, this person builds the weekly reporting job, the conversion uploads and the search term review, and fixes them when a platform changes.

The measurement and data owner. This seat owns the join between ad spend and the CRM, the conversion definitions, and the tests that prove what's working. It was a nice-to-have when reporting meant platform dashboards. Once AI reads that data to bid, score and report, a wrong definition costs real money, so the seat becomes essential.

The paid media strategist. The PPC specialist who spent half the week on bids and reports becomes a strategist who spends most of it on budget allocation, test design and reading results against pipeline. This is usually a retrain, not a hire, and it's the seat where experience matters most.

Seats that shrink tend to be the ones built around production at volume: junior roles that mostly assemble reports, ad copy roles that mostly write variants, and bid management as a standalone job. The work doesn't disappear. It moves into AI jobs that someone still owns, reviews and improves.

A filled-in hiring plan for a six-person paid team

The quickest way to see the test work is on a real-looking team. The table below is a hypothetical: a B2B SaaS company with a six-person paid team, one resignation and two new seats in next year's plan. Salary bands are left out on purpose, because they vary too much by market to be useful here.

Seat (hypothetical)Last month's weekDecisionWhy
Head of paid mediaMostly budget, tests and stakeholder workKeepAll judgment; owns sign-off on AI jobs that touch spend
Senior PPC specialistAbout half bids and reports, half test designRetrain into paid media strategistStrong judgment; production work moves to AI jobs
PPC specialist (resigned)Mostly reports, pacing checks and search term reviewsStop hiring; backfill the work with AI jobsThree recurring jobs can run weekly with a person reviewing
Paid social specialistCampaign builds, creative variants, LinkedIn reportingRetrainKeeps campaign judgment; variant production moves to AI with review
Marketing ops analystCRM hygiene, list uploads, conversion checksRetrain into measurement and data ownerAlready owns the data AI depends on
New seat 1 in the planNot yet definedHire a marketing technologistNo one on the team can build and maintain the AI jobs
New seat 2 in the planNot yet definedHold for two quartersRevisit once the AI jobs are running and the hours saved are measured

In this example the team ends up the same size with different seats. The resigned specialist's budget funds the marketing technologist, two people retrain into seats with more judgment, and the second new seat waits for evidence. That last call ties future headcount to measured results, which is the argument finance tends to accept. Our guide to the AI marketing team argues that in a team of five or six, owning the AI workflows is usually part of an existing job rather than a hire. That holds when someone on the team can already build; this plan hires because nobody can.

If you're weighing whether the builder seat should be in-house at all, our comparison of in-house marketing vs agency for AI in paid media sets out cost, speed and control for each route.

Test AI skills with a work sample, not a question

The fastest way to tell whether a candidate can work with AI is to watch them do it. Ask "how do you use AI?" and everyone gives the same answer. Give them a real-looking task and an AI assistant, and the difference shows in one short exercise.

A work sample we'd use for a paid media role:

  1. Give the candidate an anonymized search term report with spend and conversions, and an AI assistant on a business plan. Plant one row where the obvious reading is wrong, such as a term with cheap conversions that are all junk form fills sales rejected.
  2. Ask them to propose negative keywords and two budget moves, using the assistant however they like, and to explain what they'd check before acting.
  3. Watch for three things: whether they check the model's output against the data, whether they catch the one row where the model is confidently wrong, and whether their budget moves reference outcomes beyond the platform's conversions.

The second point is where the Work Trend Index finding needs care. Less experienced candidates often use AI tools fluently, but in paid media the expensive mistakes come from not knowing what to doubt. A candidate who uses AI well and checks it against the CRM is worth more than one who uses it quickly. Experience is often what makes the checking possible.

For the marketing technologist seat, swap the task for a build: a simple workflow that pulls a weekly export, summarizes it and flags anomalies, with the candidate explaining how they'd know when it breaks.

Write the job description around decisions, not tools

A job description written for an AI year names the decisions the seat owns, the AI jobs it runs or reviews, and the outcomes it's measured on. It doesn't list tools, because tools change faster than hiring cycles, and a list of six AI products tells candidates you're hiring for software rather than judgment.

A short example for the paid media strategist seat:

  • Owns budget allocation across Google Ads, LinkedIn and Microsoft Advertising, reviewed monthly against cost per opportunity in the CRM.
  • Designs and reads at least two tests a quarter, sized so they can reach a verdict.
  • Reviews the weekly AI-drafted performance report and search term recommendations before anything changes in the account.
  • Measured on pipeline from paid channels, cost per opportunity and tests concluded.

Our guide to the AI marketing team covers how to write the access and sign-off lines into each description, so the permissions in the ad accounts match the roles.

Hiring plans for AI years go wrong in four ways

The most common mistake is cutting headcount before the AI jobs exist. A team that doesn't backfill a specialist on the promise that AI will cover the work ends up with the work undone, or spread across people who were already full. Build and measure the AI jobs first, starting with the paid media tasks to automate first, then decide on the seat.

Hiring for tools. A requisition that asks for experience with three named AI products filters for people who used those products, not for judgment. Ask for the decisions and test with a work sample.

Treating AI fluency as a substitute for experience. Fast tool use without the experience to doubt the output is how confident mistakes reach the account. Test for checking as well as speed.

Retraining without changing the job. Sending a specialist on a course and leaving their week unchanged produces a certificate and the same week. Retraining works when the production work actually moves to AI jobs and the person's objectives change with it.

How Scalewell turns hiring questions into a list of AI jobs

Scalewell helps a marketing leader take a requisition to finance with evidence instead of a guess. After a free 30-minute intro call, we audit one real week of the team's work, seat by seat, and score which recurring jobs AI should take on first. The ranking shows which seats grow, which change and which work can move before anyone hires.

That ranked roadmap sets the build sprint, which typically runs 8 to 12 weeks inside the ad platforms, CRM and reporting tools you run today. By the end, the people in each seat are trained to run and change those jobs without Scalewell, and the hiring plan rests on what the jobs took off the week.

Where to start

Take the next requisition on your desk and have the person closest to the seat log one real week of its work in a shared sheet: each task, the time it took, the tool used, and whether it was a decision or production. Run the four questions against that log before the job description goes out. If you'd like someone to read that log with you before the requisition goes out, Scalewell helps B2B paid teams turn it into AI jobs and a hiring plan finance will back.

Frequently asked questions

Is it cheaper to retrain a marketer than hire a new one?

Often, because salary is only part of what a seat costs. The Bureau of Labor Statistics found wages were 70.0% of private employers' total compensation costs in June 2026 (released September 9, 2026), with the other 30% in benefits and payroll costs. A new hire also has to learn your accounts, CRM and conversion definitions, which a retrained specialist already knows. Retraining wins when the person has the judgment the new seat needs. Hiring wins when the skill, such as building AI workflows, isn't on the team at all.

What should a marketing technologist job description include?

Name the recurring AI jobs the seat builds and maintains, such as the weekly report, conversion uploads and search term reviews. Say who signs off changes that touch spend, what access level the seat holds in each ad account, and how it's measured: jobs running on schedule, hours returned to the team, and time to fix a job when a platform changes. Leave out lists of named AI products, because they date faster than the hire.

Can an agency or contractor fill the marketing technologist seat?

It can build the first jobs, but someone in-house still has to own them. AI jobs break when a platform changes a report or an API field, and a contractor who's gone by then leaves the team with a workflow nobody can fix. If an outside party builds, write the handover into the contract: documentation, a named in-house owner, and training so that owner can change the job alone.

Should we stop hiring junior marketers because of AI?

Not across the board. Stop hiring for seats that are mostly production AI now handles, such as report assembly, and keep hiring juniors into seats that build judgment, with AI skills in the requirements. Junior hires who learn to check AI output against real outcomes become the strategists a team needs later.

How long should you wait before deciding whether AI has replaced a seat's work?

Long enough for the AI jobs to run through a full reporting cycle with a person reviewing them, which for most paid teams means at least a quarter, so monthly and quarterly reports have both run. Log the hours the seat used to take, then log them again with the jobs running. If the reviewer still spends most of the old time fixing output, the work hasn't moved yet, and the seat shouldn't be cut.