Should I hire an AI-native marketer or a separate automation consultant?

Hire the person who owns the bet

Hire an AI-native marketer who can also ship one bounded workflow, unless you can already name a broken system: the reports disagree with the CRM, a migration is underway, or a sync nobody on the team can repair. A separate automation consultant is a fixed project for that break. The consultant does not choose the buyer, the message, the channel, or the date you stop.

What that person does is the judgment plus the build, not a prompt library. Using a chat window for research or copy is not this seat. The seat connects AI to the tools you already run, such as the CRM, with a human still in the loop.

The consultant title is four different buys

Automation consultant is not one job. Pond describes four shapes of the title: platform onboarding, a migration of contacts and history, ongoing operations, and a scoped project build. The first three are judgment across the engagement. The project build ends in a deliverable you can test. Pond says its marketplace is the wrong route for ongoing operations, for a strategy you cannot yet write down, and for a build that needs live customer data. A certificate is an exam, not proof the last build worked. If you can already describe the workflow, do not hire a strategist to invent it. The marketer ships that one path. A separate build is only for a seam you can name.

Tactical Marketing separates a specialist, who builds, deploys, and maintains, from a consultant who owns the lifecycle model, the platform-to-CRM boundary, what not to automate, and the names and change rules. Hire the consultant when reports disagree with the CRM, the builder has left, or a migration or CRM change is coming. More campaigns means the specialist. The firm sells the architect, and those signals assume a live program a startup may not have. See when marketing automation makes sense for a small company and why two separate hires come later.

If the strategy is not written down, do not start with a builder: the pipeline gets faster while the message stays wrong.

Task surveys do not pick the seat

Career surveys split tasks inside one seat. They are not a seed hiring rule. The AMA’s 2026 career report asked 1,412 practitioners in December 2025 and January 2026. On Stanford’s Human Agency Scale, email, SEO, paid media, performance analytics, copywriting, lead generation, market research, and graphic design are the most disrupted work. Content marketing, data literacy and storytelling, customer experience, consumer behavior, and influencer marketing are an equal partnership. Strategy, brand management, collaboration, creativity, critical thinking, leadership, emotional intelligence, ethical decision-making, and adaptability stay human-led. An AI workflow designer is a newer role.

Gartner’s 11 May 2026 briefing asked 402 CMOs from August through October 2025. They expect AI-automated marketing work to move from 16 percent in 2026 to 36 percent by 2028. That is an expectation, not a measurement, and not a startup sample. The stages are curious pilots, competent teams that scale use cases and risk conformity, and confident teams that join judgment to a changed operating model. Gartner says most of those CMOs are stalled in a competency trap, where early success limits the next step. Eighty percent of those CMOs said staff fear and anxiety is a barrier. Automating an unchosen motion scales that trap.

Ascend2’s February 2024 survey asked 387 decision-makers on a panel of marketing professionals and research subscribers, in a series run with research partners. Quality data ranked at 52 percent and overall strategy at 47 percent, ahead of journeys and integration at 40 percent each. The top goal was identifying ideal customers or best prospects (47 percent), ahead of integrating technologies (27 percent). Forty-four percent worked at firms under 50 people. A missing buyer is the marketer’s problem. A disagreeing sync is the consultant’s.

Choose from the artifact you cannot show

Match the hire to the artifact you cannot point at.

What you can point at	Who you hire
No buyer or offer sentence	AI-native marketer. Do not automate yet.
A motion already run by hand	Same marketer. Put that path in the brief.
A scoring model or nurture you can already specify, including who must never enter	A build, not a strategy. The marketer owns the rule. A builder only if the marketer will not ship that path.
No agreed definition of a good lead	The marketer writes it. A cleanup can be a later build.
Reports disagree with the CRM, or the builder left	Automation consultant, fixed scope.
Migration, new CRM, or a sync into finance or product	Fixed-scope builder. The marketer keeps the stop rule.
A blank offer and a broken sync	Marketer now. A second contract only for the seam you can name.

Run the motion by hand before you automate it. That is the step from a chat habit into a workflow.

Write a brief that can fail a build

Write a brief a build can fail: one outcome you already know how to count, one workflow sentence or a line that the motion stays manual, who approves anything a prospect sees, a maximum number of hours, and an end date.

1. A post count, a conversation quota, or a cut in manual work is not that number. Use a count you already trust.
2. Name who approves anything a prospect sees, and how fast. Keep accounts, keys, prompts, and notes in the company name.
3. The build arrives paused. You preview one record per branch, then you switch it on.

HubSpot’s workflow guide, updated August 2026, tests enrollment rules and simulates which branch a record takes. No actual actions run. A send preview goes to the tester, not the contact. Those tests are on the Professional and Enterprise tiers of Marketing Hub, Sales Hub, Service Hub, Data Hub, Smart CRM, and Revenue Hub, not on Free or Starter. Use whatever preview the lower tier has.

If the flow sends mail, Gmail counts a bulk sender as close to 5,000 or more messages to personal Gmail accounts in 24 hours. The same primary domain counts together, the label does not expire, and the rules cover personal Gmail, not Google Workspace inboxes. Meet SPF and DKIM, publish DMARC, and keep user-reported spam under 0.1 percent and never at 0.3 percent. One-click means a List-Unsubscribe header on promotional mail, not on transactional mail. A link in the body does not meet it. Honor an unsubscribe within two days. That threshold is Gmail’s own, so do not treat 5,000 as the line for every inbox.

If UK people are in the file, the ICO applies Article 22 of the UK GDPR to a solely automated decision with a legal or similarly significant effect, such as an online credit refusal or recruiting with no human. A task for a person is not that case. Marketing profiling still needs a lawful basis and a way to object.

If people in the EEA are in the file, Article 22 of the AVG (the Dutch name for the GDPR) is the same limit. The Dutch AP names a refused loan or no interview. Commercial email needs prior consent. The ACM allows an exception for existing customers on products related to an earlier purchase, requires an easy way to stop, and puts the duty on the company that orders the send and on the person who presses send. Agree in writing who the controller is before anyone builds. The company checks Article 22. Building the flow does not make the freelancer the controller.

Open a live path, not a badge

Ask for a channel decision they owned, including a test they stopped, and for a workflow still running, including the failure path. That is the bar for a fractional marketer who also builds automations.

• What did you refuse to automate? Tactical Marketing cites a scoring model with 40 inputs nobody can explain, a nurture for a segment of 60 contacts, and a sync that creates next year’s mess.
• Where is the human check, who runs it after you leave, and which step stays manual on purpose?
• A rerun while the API or the model is down must not duplicate a lead or resend the same mail.

Bring in a second contract for one named break

Add the consultant only after the motion is real, for a break you can write in one sentence. The marketer writes who enters and what a good lead means. The builder ships that seam paused, with notes, then leaves. Tactical Marketing says an audit-shaped project generally runs six to ten weeks. Do not buy that length to find the buyer. Both on day one only if a sync is already breaking the sales handoff and someone already owns the offer: two scopes, and you in the one weekly review.

Do not open either search yet

Open neither search when there is no offer sentence, when the workflow is only a notification, when nobody can approve a customer-facing message each week, when nobody on your side can direct the builder in the one weekly review, or when you want a consultant to invent the strategy. Also wait if the only job is installing a tool.

I would keep this to one seat

I would keep a freshly funded startup on one seat for marketing, AI, and automation, fully remote, from the first message to the first customers. A second contract starts only when a named seam is breaking.

I am one person, AI-native, trading as Poldermarketing. I work remote, in Dutch and English, freelance or for a few days a week. I build and run the work in your accounts: the buyer, the message, the channel bet, and one CRM path, without a separate specialist for every part. A person still approves anything a prospect sees. I am strong in AI, content, automation, and workflows. Google Ads and Meta Ads are newer for me: I set them up and review them, and I am the wrong person when the job is scaling a large paid program.

If the offer is still hard to say, routing can wait. The free growth scan shows how the site reads now. How I work is the engagement. Positioning and messaging is where to start when the words are the constraint.