AI for marketers: what to hand to AI, and what to keep.

01What is AI for marketers, and what is it good at today?

AI for marketers is the use of AI assistants, AI features inside marketing software and AI agents to do the production side of marketing work. Today it is strongest at research summaries, first drafts, repurposing one piece into many, pulling reports and drafting follow-ups. It is weakest where a decision needs taste or accountability.

Most marketers already use it somewhere. HubSpot's 2026 State of Marketing, a survey of more than 1,500 marketers, found 86.4% of marketing teams use AI in at least a few areas, with 42.5% using it extensively for content creation¹. Closer to home, Constant Contact's Small Business Now research (June 2026) found the top AI use among Australian and New Zealand small businesses was writing emails, subject lines and social captions, at 45.1% of AI users².

Use is still shallow, though. The CMO Survey, which asked 308 US marketing leaders in January 2026, put AI in 24.17% of marketing activity now and an expected 55.91% within three years³. The gap between those two numbers is where most in-house marketers sit: they use a chat assistant for drafts, and the rest of the week runs the way it did in 2022. The guide to AI marketing transformation for small business covers what closing that gap looks like across a whole function.

02What should marketers hand to AI, and what should they keep?

Hand AI the work that repeats, can be checked quickly and does no harm as a weak first draft: research, drafting, repurposing, scheduling, tracking and follow-up. Keep the work that sets direction or carries the business's word: positioning, the claim worth making, taste, the decision on what ships, and conversations with buyers. A short review step sits between the two.

The table below splits common marketing jobs at that line. The review step column matters most, because a job is only safe to hand over once someone knows exactly what they check and when.

The sales follow-up row is where AI for marketing and sales overlaps, and for a 5 to 50 person business it often pays back first. Enquiries go quiet because nobody had time to send the second email, and AI is very good at noticing that and writing the draft. For ten more worked cases, one per marketing job, see the AI marketing examples guide.

03Which kinds of AI tools do marketers use?

Marketers use three kinds of AI tools: general chat assistants such as ChatGPT, Claude and Gemini; AI features built into software they already pay for, such as the email platform or CRM; and AI agents that run a multi-step job on a schedule. Each kind suits a different job, so choose by fit.

General assistants are still where most organisations start. Datacom's 2026 State of AI Index found general-purpose assistants were the most used AI technology among New Zealand organisations, at 68%, against 13% using agentic or autonomous systems⁴.

On "best AI for marketers" and "best AI for marketing managers" searches: no independent study ranks these tools for small business marketing, and most "best of" lists are written by a vendor on the list. A practical order works better. Turn on the AI already inside the tools you pay for, add one chat assistant for drafting and research, and look at agents only for a job that repeats every week.

04How is agentic AI for marketing different from a chatbot?

Agentic AI holds a goal and works through several steps on its own, using tools such as a browser, a CMS or a CRM, then checks the result and decides what to do next. A chatbot answers one prompt and waits. For a marketer, the difference shows up as who starts the work each week.

A chatbot helps you write a newsletter when you sit down to write it. An agent notices it is Thursday, pulls the week's published articles, drafts the newsletter in your voice, builds it in the email platform as a draft and tells you it is ready for review. The production is the same. The agent removes the step where a busy person has to remember to begin.

Agents are still early in marketing. Salesforce's State of Marketing report (4,450 marketers including 890 at small and medium businesses, surveyed late 2025) found only 13% of marketers using agentic AI, and reports that agents free up 6 to 7 hours a week for creative and strategic work⁵. The full definition, the loop an agent runs and what it must never do without a person are in the pillar guide, What is an AI marketing agent?

05A first-30-days plan for using AI in marketing

A good first month for using AI in marketing has four steps, one per week: write the inputs, hand over one production job, add a review gate and a second job, then measure and decide. It needs a few hours a week on top of normal work, and it ends with a written decision about what to keep.

1. Week 1: write the inputs. A one-page positioning file (what you claim, for whom, with what proof), five to ten voice rules, and the twenty questions customers ask most. Log one week of your own time by marketing job so you know where the hours go.
2. Week 2: hand over one production job. Pick something that repeats weekly and is easy to check, such as turning an article into social posts or turning sales call recordings into notes and follow-up drafts. Give the assistant your Week 1 files every time. Time the job before and after.
3. Week 3: add the review gate and a second job. Write the checklist a draft must pass (below), and name who runs it. Then switch on the AI features inside one tool you already pay for, such as the email platform or CRM, before buying anything new.
4. Week 4: measure and decide. Count hours saved, the share of drafts that shipped with light edits, and errors caught at review. Keep the jobs that saved time without raising errors, drop the rest, and note any job that repeats so reliably it could run on a schedule as an agent.

Flow AI's free agent templates include starting files for the Week 1 inputs. If you want the month to sit inside a wider plan, the AI marketing strategy guide fits all of this on one page, which also covers how to use AI for marketing strategy work without letting it write your choices.

06How do you keep AI-written marketing on brand and accurate?

Keep AI-written marketing on brand and accurate with two things written down: voice rules the AI receives with every brief, and a review gate a named person runs before anything reaches a customer. The rules make first drafts closer to right. The gate catches what the rules miss, especially invented facts and promises the business never made.

Review is where many teams are thin. Australia's National AI Centre SME pulse (December 2025 to February 2026, at least 400 owners and decision-makers a month) found only about half of AI users check AI outputs before they affect customers⁶. In New Zealand, KPMG and the University of Melbourne, surveying over summer 2024-25, found 51% of employees had relied on AI output at work without evaluating its accuracy⁷.

A review gate for a small team can be five questions, answered in under ten minutes per piece:

• Does every number, name and date match a source someone opened?
• Does it make any claim, guarantee or price the business has not approved?
• Does it pass the voice rules, including the banned words list?
• Would a customer recognise this as us, written for them?
• Does it use any personal information, and is that allowed?

The last question matters in New Zealand. The Office of the Privacy Commissioner's AI guidance expects a privacy impact assessment before using AI tools and advises that if in doubt, you do not use AI tools to handle personal information⁸. Customer lists stay out of public chat tools. The glossary entry on human in the loop covers where the person sits when the work is automated.

07Which skills should marketers build to work with AI?

The skills that matter most for marketers working with AI are briefing, editing, fact checking, breaking a job into steps, and reading data well enough to make a call. Each one is a judgment skill. As AI takes more of the drafting, the value of a marketer moves to deciding what good looks like and catching what is wrong.

• Briefing. Writing the context a model needs: audience, claim, proof, examples, what to avoid. Most weak AI output is a weak brief.
• Editing for voice. Cutting a competent draft into something that sounds like your business. HubSpot's 2026 survey found 62.7% of marketers believe they need more unique, human-centred content to compete¹, and editing is how a draft gets there.
• Fact checking. Opening the source behind every number, every time.
• Workflow design. Breaking a job into steps, naming the input and output of each, and marking where a person checks. This is the skill that turns a chat habit into a system an agent can run.
• Reading data. Knowing which number decides whether a job continues, and asking AI the right question of a report.
• Positioning. Deciding what the business claims. AI can research it; only someone accountable can choose it.

08What should you measure when you use AI for marketing?

Measure AI in marketing on two levels: what it does to the work, and what it does to the business. For the work, track hours saved per job, the share of drafts that ship with light edits, and errors caught at review. For the business, track enquiries and cost per enquiry. Hours saved only count if they go somewhere useful.

The last row is the one owners care about. The glossary defines cost per enquiry, and it is worth counting your own time in it at a fair hourly rate, since time is the main thing AI changes.

09Where Flow AI fits

Flow AI is an AI marketing transformation practice: the strategy, the systems and the agents that run a marketing function, engineered end to end. Amber Lan builds and runs these systems herself and runs a whole marketing function on them, with a person reviewing everything before it ships.

The starting point is a written scope, a written plan that names the first system to install and why. The free agent templates are there if you would rather start on your own.

10The short version

AI is now good at the production half of marketing: research, drafts, repurposing, scheduling, reports and follow-up. The marketer keeps the half that decides: the claim, the angle, what ships and the conversation with the buyer. Write the inputs first, hand over one job at a time, and put a five-question review gate in front of every customer.

Start with the AI already inside your tools and one chat assistant, measure hours and errors for a month, and only move a job to an agent once it repeats reliably and has a clear review step. The time you get back is worth most when it goes into positioning and buyer conversations.

To find out which marketing job to hand over first, talk to Amber.