A quote can look tidy while hiding important questions. What work will AI prepare? Which tools will it connect to? Who reviews mistakes?
If those answers are unclear, use this checklist before accepting an AI automation quote, fixed-price build, proposal, or support offer.
A useful quote explains the operating workflow, not just the technology. It should show what is included, what the business must provide, and what still needs human judgement.
The quote should name the workflow in plain English. “AI sales automation” is too broad. “Prepare a weekly queue of stale CRM opportunities with suggested next actions for sales review” is specific enough to quote, test, and manage.
Good scope includes the trigger, inputs, output, reviewer, timing, and exception path. If the provider cannot describe those pieces, you may be buying a demo rather than an operating workflow.
The quote should state which systems AI will read or update: CRM, inbox, calendar, website forms, documents, spreadsheets, job-management tools, accounting exports, or knowledge bases. If the proposed build starts with incoming leads, compare it with the AI Enquiry Triage NZ workflow so the quote separates intake, triage, drafting, approval, and follow-up.
It should also say whether access is read-only, draft-only, update-with-approval, or fully automated. Start narrow. A review queue is easier to trust than a workflow that sends messages, changes records, or publishes content immediately.
Every quote should define what stays human-approved. Customer messages, quotes, discounts, pricing, CRM stage changes, finance actions, public website updates, and sensitive decisions should not move automatically unless the business has deliberately approved that authority.
Use the AI Approval Gates for Business Automation guide to decide where the stop signs belong.
Ask what the provider will test, how exceptions will be handled, and what handover notes you will receive. The quote should cover documentation, ownership, logs, maintenance, and what happens when the source process changes.
An automation build is not finished when the first happy-path example works. It is useful when your team can run it, review it, pause it, and improve it.
A lower price is not automatically wrong. A higher price is not automatically safer. AI automation quotes often include different work under the same label.
One provider may quote only the initial build. Another may include assessment, setup, testing, documentation, monitoring, and monthly improvement. Those are different offers.
Before comparing price, ask who checks logs, fixes broken prompts, updates instructions, reviews edge cases, and decides whether to expand the workflow after launch.
Some quotes include software subscriptions, model usage, workflow tools, or CRM add-ons. Others assume you will pay those separately. The quote should separate tool cost from implementation labour and ongoing management.
For budget context, read How Much Does an AI Agent Cost for a New Zealand Business?.
A prototype proves possibility. A managed workflow supports recurring business work. If a quote only promises a prototype, do not treat it as a production operating system.
A production-ready workflow needs source control, permission rules, test cases, approval points, fallback steps, and quality ownership.
Use these questions in a sales call, proposal review, or internal buying meeting.
Ask the provider to describe the workflow in one sentence and list the steps. If several workflows are bundled together, split them. AI usually works better when the first build is narrow, frequent, valuable, and reviewable.
Do not accept “we connect to your tools” as enough detail. A quote should say whether the AI reads data, drafts work, recommends an action, updates a record, sends a message, books a meeting, or publishes public content.
Each permission level changes risk, testing, and review requirements. The AI Procurement Checklist NZ gives a wider buying checklist for tools, agents, and consultants.
Every workflow needs an owner. Ask who checks output, how rejections are recorded, and when the workflow escalates to a person.
A good quote makes review practical. If the human review step is vague, the business may save little time and gain a new quality-control burden.
Ask about handover, maintenance, monitoring, and improvement. AI workflows drift when templates change, staff change, data changes, product lines change, or customers ask unusual questions.
The quote should explain how the workflow is paused, repaired, reviewed, or retired.
Sometimes you are ready for a build quote. Sometimes you need an assessment first.
If the business can name ten possible AI use cases but cannot choose the first one, paying for implementation may be premature. An assessment compares opportunities by value, readiness, source data, risk, and review effort.
Sales, admin, operations, finance, and marketing may all want AI support. Assessment helps prioritise the build sequence before spend is committed.
If AI may touch customer records, proposals, pricing, invoices, public content, reviews, bookings, or sensitive documents, define approval gates before build.
The $1,000 AI Agent Assessment maps one practical workflow, the tools involved, the approval model, and the recommended next step: build now, prepare first, use simpler automation, or avoid it.
Before accepting an AI automation quote, confirm these checkpoints:
If several boxes are missing, ask for tighter scope before you pay. If the missing pieces are strategic rather than technical, use an AI Automation Assessment NZ or AI Agent Assessment before accepting the build.
Ask what workflow will be automated, what systems AI will access, what it may read or change, what stays human-approved, how testing works, what handover you receive, and who supports the workflow after launch.
AI automation cost depends on workflow complexity, tool subscriptions, data readiness, integrations, permission level, testing, handover, and ongoing support. Do not compare quotes until you know whether they include assessment, build, and management or only setup.
A fixed-price build can be enough for a narrow, low-risk workflow with clear source data and review rules. It may not be enough when the workflow needs discovery, permissions, testing, documentation, monitoring, or monthly improvement.
Get an assessment first when the workflow is unclear, several teams want AI at once, data access is sensitive, approval gates are undefined, or you are unsure whether the right answer is build, prepare, wait, or avoid.
Keep customer messages, pricing, quotes, discounts, finance actions, sensitive CRM changes, booking commitments, public website publishing, legal wording, and unusual exceptions human-approved unless the business has deliberately assessed and approved that authority.
Do not accept an AI automation quote just because the demo looks useful. First, make sure the quote names the workflow, data, permissions, review rules, testing, handover, and support model.
Before you pay for a workflow build, book the $1,000 AI Agent Assessment to map the first workflow, define the approval gates, compare build options, and decide whether to build now or prepare first.
