What should an AI marketing scope of work include?

01What is an AI marketing scope of work?

An AI marketing scope of work is a written document that defines what a provider will build or run, what the client provides, how results are measured and who owns the output, agreed before work starts. It turns "help us use AI in marketing" into specific jobs with owners, dates and a price both sides can check.

You will see it called a scope of work, a statement of work (SOW) or simply the proposal. A proposal sells the work. The scope describes it precisely enough that anyone reading it later can tell whether it was delivered. A contract or terms of business then covers the legal relationship, and often refers to the scope as a schedule.

AI work needs more in the scope than a traditional marketing retainer, because the provider is building systems that act inside your accounts and touch your data. This guide is general information for New Zealand and Australian businesses, written from the buyer's side. Have a lawyer read any contract you are unsure about.

02Why does an AI marketing project need a written scope?

Delivery and value decide whether a marketing relationship thrives, and both are scope questions. A written scope sets out what will be delivered and how value will be judged before any money changes hands. For AI work it also records what the systems may do on their own, which nobody remembers accurately six months later.

The data on why relationships end points straight at scope. Setup's 2025 Marketing Relationship Survey found dissatisfaction with delivery and with value were each cited by 61% of clients who ended an agency relationship, and 44% said the agency did not understand their business¹. It is a small sample of about 100 responses, mostly from larger US brands, but the pattern will be familiar to any owner who has parted ways with an agency.

AI projects add their own risk. Gartner predicted in June 2025 that over 40% of agentic AI projects will be cancelled by the end of 2027 because of escalating costs, unclear business value or inadequate risk controls². A scope that names the outcome, the measure, the costs outside the fee and the review rules deals with each of those before the build starts.

03What sections should an AI marketing scope of work include?

A complete scope covers ten sections: outcomes, agents and systems, data access, review points, deliverables, measurement, ownership, handover, exit and price including GST. Small projects can cover each in a few lines. If any section is missing, ask for it in writing before you sign, so both sides agree on it from day one.

Outcomes

One or two business outcomes, in plain words: "more enquiries from organic search and AI assistants for our three main services in Auckland", or "a steady outbound lane producing qualified meetings". Activity ("12 blog posts a month") belongs under deliverables. The outcome is the reason you are paying.

Agents and systems

Name each agent or system to be built, what job it does, and what tools it runs on. Replace phrases like "AI-powered marketing" with a sentence such as "a search agent that keeps service pages crawlable and quotable, and reports monthly on AI citations". List the software it depends on and who pays each subscription.

Review points

Say where a person checks the work and who that person is. As a default, anything a customer will see or that spends money waits for approval, and so does any change to the agent's rules. Everything else, such as research, drafts and internal reports, can run on its own. This is human in the loop, written down.

Deliverables and measurement

List each deliverable with a due date and a test for "done". Then set the measures: the baseline taken at the start, the numbers tracked, how often they are reported and in what form. Choose measures close to revenue, such as enquiries, qualified replies or cost per enquiry, over counts of output.

04How should the scope handle data access and privacy?

The scope should list every account and dataset the provider and its agents will touch, the access level for each, and what happens to the data at the end. Customer and prospect data needs extra care: say which AI tools may process it, where they store it, and what is kept out of them entirely.

Regulators on both sides of the Tasman are clear about caution. New Zealand's Office of the Privacy Commissioner expects organisations to complete a privacy impact assessment before using AI tools, and says that if in doubt, do not use AI tools to handle personal information³. In Australia, the OAIC recommends that organisations do not enter personal information, particularly sensitive information, into publicly available generative AI tools⁴.

Lead lists bought or built from other sources raise a newer duty. The Privacy Commissioner's note on the Privacy Amendment Act says that from 1 May 2026, under the new IPP3A, organisations collecting personal information indirectly must take reasonable steps to tell the person it was collected, why, and who will receive it⁵. If an outbound agent enriches prospect data, the scope should say who handles that. The guide to AI marketing and privacy law in NZ and Australia covers the detail.

Grant access by role rather than through shared passwords: "editor access to the website CMS; read access to Google Analytics; no access to the accounting system". At the end of the project, the provider removes its access and confirms in writing that copies of your data are deleted.

05Who owns the work, and how do handover and exit work?

You should own everything built for you: content, prompts, agent configuration, reports and the data produced, with every account in your business's name. The scope should describe the handover documents in detail, the notice period to end the work, and what the provider must hand back when you leave, whatever the reason.

Copyright defaults for commissioned work differ between New Zealand and Australia and by type of work, and neither may match what you assume. Avoid relying on defaults. Write the ownership in: "all deliverables, prompts, configuration files and data created for the client are owned by the client on payment". If the provider keeps the rights to its own reusable tools or skill libraries, that is reasonable, provided the scope grants you a licence to keep using what was installed.

Handover should include written documentation of each agent, the files and settings it uses, the review rules, an account list showing who owns each one, and a session where your team runs a full cycle. The week-by-week guide to the first month of a build shows where handover sits in the timeline.

Exit terms vary widely, so read them closely. Among published NZ price pages, OnlineNZ's packages page states one month's notice⁶, while The SEO Guy lists its entry SEO retainer with a six-month minimum⁷. Neither is wrong. What matters is that you know the term, the notice period and the handover obligations before you sign.

06How should price and GST be written in the scope?

State the price per deliverable or per project, whether each figure includes GST, when invoices are issued, and every cost outside the fee, such as software subscriptions, ad spend and data tools. Also say what triggers a change in price, such as extra agents or extra rounds of revisions, and how changes are approved.

Inland Revenue states that New Zealand GST is charged at 15%⁸, and the Australian Taxation Office describes GST as a broad-based tax of 10% on most goods and services⁹. Business quotes usually exclude it. A review of Australian agency pricing by Digital Nomads HQ found 78% of listed prices exclude GST and only 22% of agencies disclose a minimum ad spend¹⁰, so ask about both.

For market price ranges across agencies, freelancers and AI tools, explore the guide to AI marketing costs in Australia and New Zealand. A clear scope is also what makes two quotes comparable.

07An AI marketing scope of work template you can copy

Use the table below as a skeleton. Copy the section names into your own document, replace the example wording with your project's details, and send it to each provider you are considering so every quote answers the same questions. Leave a section blank only if both sides agree, in writing, that it does not apply.

If you want a starting point for the systems themselves, the free agent templates show how a single agent's instructions and review rules can be written down.

08Red flags in an AI marketing proposal

Walk away, or ask for changes, when a proposal promises guaranteed rankings or AI mentions, keeps accounts or data in the provider's name, describes the work only as "AI-powered", has no review step, or locks you into a long term with no handover. Each one shifts risk from the provider to you.

• Guaranteed positions. Google's Business Profile help states there is no way to request or pay for a better local ranking¹¹, and OpenAI's help centre says placement in ChatGPT search results is not guaranteed¹². A provider promising either is guessing.
• Vague AI. If the proposal cannot name the agents, the tools and what each one does, ask. Gartner's June 2025 release estimated that only about 130 of the thousands of agentic AI vendors are real², and warned about "agent washing", the rebranding of existing products as agents.
• Accounts, ad accounts or domains registered to the provider, or a proprietary platform you cannot export from.
• Customer-facing work with no named approver.
• Prices with no GST statement, or software and ad costs left out.
• A minimum term with no exit clause and no handover obligation.
• No baseline, so nobody can tell at the end whether anything improved.

The guide to choosing an AI marketing provider turns these into questions to ask on the first call, and the guide to what AI marketing services include helps you check the scope covers the right work.

09Where Flow AI fits

Flow AI provides AI marketing services for businesses in New Zealand and Australia. Amber Lan, an AI marketing engineer in Auckland, builds the AI agents that run search, content, outbound and reporting with your team, then hands them over. Every build starts with a written scope like the one above.

Once you agree that scope together, the build is priced per project and ends with a documented handover that leaves the systems in your hands. The page on AI marketing services for NZ and Australian businesses explains how a build runs. Companies hiring a marketing engineer can reach Amber through the hiring page.

10The short version

A good AI marketing scope of work names the outcome and the measure, the agents and their tools, the data they can touch and where a person approves the work. It also lists each deliverable with a date and a test for done.

Then it settles ownership, handover, exit and price with GST stated. Copy the template, send it to every provider you are considering, and compare the answers line by line.

To shape the scope of your AI marketing build together, talk to Amber.