What is managed AI? The AI department model explained

What is managed AI?

Managed AI is a monthly service in which a provider is responsible for your AI systems working, not just for building them once. It covers building new agents, hosting and running them, monitoring what they do, fixing them when something they depend on changes, and reporting in plain English on what ran. You get the result of an AI team without hiring one.

An “agent” here means a piece of AI software that does a job without being asked each time: answering after-hours enquiries, categorising Xero transactions overnight, checking stock levels before the team starts, or qualifying new leads and drafting replies. If agents are new to you, our guide to AI agents for small business explains what they are and when they are worth building.

Why isn’t a one-off AI build enough?

A one-off build works on the day it is handed over, then slowly drifts. The booking system updates its software, your prices change, a new staff member never learns the agent exists, and nobody is responsible for noticing. Managed AI exists to stop that decay by making one provider accountable for every agent, every month.

This is the same reason most small businesses outsource IT rather than employing a systems administrator. You pay a monthly fee and someone else makes sure the computers, network and backups keep working. A managed AI department does that for the AI layer of your business:

• Build new agents on a cadence, aimed at whatever is costing you the most time this quarter.
• Run them, with hosting and uptime as the provider’s problem.
• Monitor what they do, catching drift and mistakes before customers do.
• Improve them as your tools, prices and services change.
• Report monthly in plain English: what ran, what it handled, what is next.

How much does managed AI cost in Australia?

On Autopilot’s managed AI costs $1,500 a month for up to two live agents (Essential), $2,000 a month for up to four with a new agent or material improvement each month (Growth), and $4,000 a month for a roadmapped build-out across the business (Scale). Prices are AUD, ex GST, month to month, with AI usage included.

Tier	Price (AUD, ex GST)	What you get
Essential	$1,500/month	Up to 2 live agents run and maintained, monitoring, monthly briefing, quarterly tune-up
Growth	$2,000/month	Up to 4 live agents, one new agent or material improvement shipped each month, priority queue
Scale	$4,000/month	Live agents within scope, roadmapped build-out across the business, optional DotVA virtual assistant

The Claude API costs behind typical agents run about $3–15 a month each and are absorbed in the fee. Full details are on the managed AI page and the pricing page.

Is managed AI better than hiring an in-house AI person or doing it yourself?

It depends on how much AI work you have. Hiring suits businesses with enough work to keep a specialist busy full time. Doing it yourself suits experiments and one-person businesses. Managed AI suits the large middle: businesses with several workflows worth automating and no one whose job it is to own them.

Path	Cost shape	Strength	Weakness
Hire an in-house AI specialist	Roughly $170,000 a year: a data scientist averages about $132,500 in Melbourne on SEEK (October 2026), plus 12% super and on-costs	Dedicated, inside the business	One person’s skills; a gap if they leave
Buy one-off builds	Fixed price per project	Good way to prove value on one problem	Nobody owns it after handover
Do it yourself	Your time plus tool subscriptions	Cheap, good for learning	Stalls at integrations; runs on nights and weekends
Managed AI department	Fixed monthly fee	A team’s range plus one accountable owner	Overkill for a single small automation

The salary is SEEK’s average for a data scientist in Melbourne (SEEK, checked October 2026). The on-costs follow our AI Cost Index model (base salary times 1.28, covering super, workers’ compensation, equipment, recruitment and management overhead), and the superannuation guarantee rate is 12% of ordinary time earnings from 1 July 2025, per the ATO. See managed AI versus an in-house AI hire for the full comparison.

When does managed AI make sense, and when doesn’t it?

Managed AI makes sense when two things are true: you have more than one workflow worth automating, and nobody in the business has the time or skill to own the AI. It does not make sense when you have one narrow task in mind and nothing after it. Then a single fixed-price build is the right starting point.

Good signs you are ready:

• Enquiries, bookings or follow-ups are being missed because everyone is busy doing the actual work.
• You already have one automation running and want two or three more.
• You are paying staff to do repetitive admin in Xero, your CRM or your inbox.
• Something you built or bought last year has quietly stopped working and nobody noticed for weeks.

Not ready yet: you have not identified a specific job for AI to do, or the only job is small enough for a $497 Quick Start build. Start there, run it, and move to a retainer when you want the next things built.

What should you insist on from a managed AI provider?

Insist on ownership, exit terms and a straight answer about data. You should own the code, prompts and settings from day one and have a written plan for moving hosting and the AI account into your name if you leave; the retainer should be month to month; and the provider should tell you which AI service processes your customers’ information, where, and under which plan. In licensed industries the AI must stay on the admin side of the line.

1. Ownership from day one. The code, prompts, settings and documentation are yours. With On Autopilot, if you stop, we move each agent to hosting and an Anthropic account in your name so it keeps running.
2. Month-to-month terms. If the service stops being worth the fee, you should be able to leave without losing the agents.
3. A data answer. Your obligations under the Privacy Act stay with you whoever runs the AI. Our AI privacy guide explains which AI plans suit customer data.
4. The regulated line. In health, legal, financial and tax work, the AI handles bookings, reminders, drafts and admin. Registered and licensed people do the regulated work.
5. A named person. Someone who knows your business, not a ticket queue.

Our guide to choosing an AI consultant has the full list of questions to ask.

What does managed AI look like in practice?

In practice it is a handful of agents doing specific jobs on tools you already use, plus a monthly check-in. We run this model inside our own businesses, which is where our proof comes from. These are our own systems, not client work.

• Hey Bud Skincare (our Shopify brand): an inventory system syncs warehouse stock and Shopify sales at 6am every day, raises stock and reorder alerts, drafts purchase orders for a person to approve and puts a Claude-written daily brief on the dashboard. Ollie Watts, who is not a developer, built it with Claude Code.
• DotVA (our virtual assistant agency): the website lead pipeline runs every enquiry through a Cloudflare Worker that scores spam and job applicants, stores leads in a database and feeds a dashboard and the internal pipeline.
• DotVA HQ: the internal operations platform for invoicing, payroll and client onboarding.
• Lead Gen Empire: a 20-site comparison network built and maintained with Claude Code.

Illustration only: a three-practitioner physiotherapy clinic on the Growth tier might run an AI front desk on its website and SMS, automated recall reminders from Cliniko, a weekly digest of cancellations and gaps, and a Xero categorisation helper for its bookkeeper, adding one improvement each month.

How On Autopilot can help

On Autopilot is Australia’s outsourced AI department for small and medium businesses: we build, run and keep improving the AI that answers enquiries, chases leads and handles admin. Managed AI starts at $1,500 a month (Essential), with Growth at $2,000 and Scale at $4,000, all month to month, and you own the code, prompts and settings. Most businesses begin with one fixed-price build, such as the AI Front Desk at $1,500 setup plus $199 a month, and move to a retainer once it has proved itself. Book a free 30-minute AI audit with Jenn Yang to find out which agents are worth building first.