Guide · Buying AI help · Melbourne, October 2026
AI consultant vs software: do you need a person, a product, or both?
You bought the AI licences. Forty of them. There was a launch email with a rocket emoji. Twelve weeks later, six people use it, mostly to make emails sound less annoyed. Now someone’s asking whether you should have hired an AI consultant instead, and someone else is asking whether a consultant is just an expensive way to be told to use the software you already bought. Both are fair questions. This is the honest answer for Australian businesses: when software on its own is plenty, when an AI consultant earns the fee, and when you need a bit of both.
Fixed price A$1,950 + GST. If the software you own already does the job, the report will say so.
- Says when you don’t need a consultant
- Australian Government guidance linked
- Every number sourced or labelled
Short answer: In the AI consultant vs software decision, most small businesses should start with software: switch on the AI already inside the tools you pay for and use it for everyday jobs like drafting and summarising. Bring in an AI consultant when the work crosses several systems, involves personal information or decisions about people, or when you’ve bought the licences and nobody uses them. The software is the gym. The consultant is the trainer. You still do the reps.
Side by side
AI consultant vs software at a glance
It’s rarely one or the other. There are three routes, and the middle one is the one most businesses end up on.
| Software, DIY | Software + help | Consultant-led | |
|---|---|---|---|
| What you buy | Licences for AI tools, or the AI features in software you already use | The same licences, plus a fixed-fee review, setup or training | A project: process review, design, build, rollout and handover |
| Who does the setup | You, in the gaps between your real job | Your team, with someone showing them where to start | The consultant, with your team in the room |
| Cost shape | Per user, per month, forever | Licences plus a one-off fee | A project fee, then licences or running costs |
| Time to first result | Days, if anyone gets round to it | A few weeks | Weeks to a few months |
| Who makes people use it | Whoever sent the launch email | A named internal owner, coached | Built into the plan, with a named owner |
| Best for | Everyday jobs inside one tool | One clear workflow, a team willing to learn | Work across several systems, sensitive data, or decisions about people |
| The risk | Seats nobody opens | Help that stops before the habit forms | A dependency, if nothing is handed over |
What you’re actually buying
The gym membership and the personal trainer
A confession before we start. We’re the kind of people who read the user manual. Cover to cover. Including the warranty section, and the page in four languages about not using the toaster in the bath. It’s deeply uncool and we’re not stopping. It’s also why this topic keeps us up at night, because almost nobody else reads the manual for their AI software, and the software knows it.
Here’s the simplest way we’ve found to explain the AI consultant vs software choice. AI software is a gym membership. An AI consultant is a personal trainer. One gives you the equipment. The other makes sure you use it properly, on the right muscles, without hurting yourself.
The gym membership
AI software
- Gives you capability: drafting, summarising, searching, reading documents
- Priced per person, per month, whether they turn up or not
- Improves on the vendor’s schedule, not yours
- Has no idea how your business works, and isn’t going to ask
The personal trainer
An AI consultant
- Starts with how the work actually moves, then picks the tool
- Decides what needs AI, what needs a simple rule and what stays with a person
- Sets it up, trains your people and measures whether it worked
- Should leave, eventually. A good one plans the exit on day one
The bit both sides leave out
Neither option does the work for you. Software vendors imply the tool is the hard part. Some consultants imply the strategy is. In practice the hard part of any AI implementation is the boring middle: deciding which process to fix, tidying the files the AI will read, fixing who can see what, showing people how to use it, and checking a month later that they still do.
The Australian Government’s own advice agrees on the order. The National AI Centre’s guide to choosing an AI solution says that before finding a solution, “it’s important to define what you want to improve”. It then lists four options, from AI built into software you already use through to customised AI, and recommends you “start with the first option and move to later options if you need more control or capability”.
So the real AI consultant vs software question isn’t “which is better?” It’s “who’s going to do that boring middle, and do they have the time?”
The shelfware problem
Why AI tools go unused after you buy them
Software that’s paid for and never used has a name in IT: shelfware. AI tools are especially good at it, for a simple reason. They don’t come with a job. A payroll system knows it does payroll. An AI assistant can do a thousand things, so on a busy Tuesday it does none of them.
The research is fairly consistent on where it goes wrong, and it isn’t the technology.
- People aren’t shown how. Microsoft and LinkedIn’s 2024 Work Trend Index found 75% of knowledge workers use AI at work, but only 39% of users had received AI training from their company, and 60% said their company lacked a vision and plan for it.
- The value gets stuck in people and process. In a 2024 survey of 1,000 executives, BCG found 74% of companies had yet to show tangible value from AI, and put about 70% of the challenges down to people and process, against 20% technology and 10% algorithms.
- Plenty of businesses don’t know where to start. The National AI Centre’s adoption tracker found 19% of Australian SMEs “simply don’t know how to use AI in their business”.
[Puts down protein shake.] Two caveats. The Microsoft and BCG numbers are global and mostly about larger companies, so read them as direction, not destiny. And none of this means software fails. It means software on its own is only the membership.
“Seventy-four percent of companies have yet to show tangible value from their use of AI.”
BCG, October 2024
- Users trained by their company
- 39%
- Australian SMEs who don’t know how to use AI
- 19%
The shelfware calculator
What are your unused AI seats costing you?
Open your AI tool’s admin page, find the usage report, and put three numbers in. It takes about a minute, which is roughly a minute longer than most businesses have ever spent looking at it.
How it works: annual spend is seats × price × 12. Unused spend is (seats − active users) × price × 12. Cost per active user is your monthly bill divided by the people using it. It’s an estimate based on what you enter, ex GST, and nothing is sent anywhere. Under 40% in use reads as a rollout problem, 40 to 69% as patchy, and 70% or more as healthy. One quiet week isn’t a verdict, so check a few.
The honest bit
When you don’t need an AI consultant
We’d be bad at our jobs if we pretended everyone needs us. Lots of businesses get real value from AI tools with no outside help at all. You can probably go it alone if most of these are true:
- The AI is already inside software you use, and you just need to switch it on
- The job is the same in every business: drafting, summarising, meeting notes, tidying a spreadsheet
- It happens inside one tool, and nothing has to move between systems
- There’s no personal or sensitive information going into it, or the tool is approved for it
- A person checks every output before it goes anywhere that matters
- Someone on the team is curious and has a couple of hours a week to experiment
That matches the government’s advice almost word for word: “choose the simplest option that meets your needs, and add complexity only where the value justifies it.”
If budget is tight, there’s free and subsidised help too. Small businesses can get one-on-one digital advice through the federal Digital Solutions program, and AI Adopt Centres offer free specialist help, though only to eligible SMEs in National Reconstruction Fund priority sectors. Our guide to AI consultant cost in Melbourne covers who qualifies and what paid help actually costs.
If that’s you, buy the software and close this tab. Bookmark it first, though. You might be back around week ten.
The other honest bit
When an AI consultant earns the fee
Then there’s the other half. An AI consultant is worth paying for when the problem isn’t “which tool?” but “how does this work actually flow, and what should change?” These are the signs:
- The work crosses systems. Email to PDF to spreadsheet to practice software to a manager’s inbox. No single AI tool owns that whole journey.
- It involves personal or sensitive information, or decisions about people such as approvals, eligibility or pricing.
- You already bought the licences and the calculator above made you wince.
- You can’t say which process to fix first, only that “admin takes forever”.
- Nobody inside has the time. The curious person is also the busiest person. That’s not a coincidence.
- A vendor is promising things you can’t test, and the contract is longer than this guide.
What a good AI consultant actually does
Less magic than the title suggests. They watch how the work moves today and time it. They decide what needs AI, what needs plain workflow automation (cheaper, faster and it never makes things up), and what should stay with a person. They set it up, test it on your messiest real examples, train the team and measure the result. Then they hand over the documentation and leave.
A good one will often tell you to keep the software you already have. If the first meeting is a demo of their own platform, you’re talking to a vendor in a consultant’s blazer.
The second price tag
AI consultant vs software: the hidden costs of each route
Both routes have the price on the quote and a quieter price nobody prints. Here are both receipts.
DIY with AI software
Tax invoice · monthly · forever
- Seats × everyone, on day one
- per user
- Seats for people who tried it once
- still billed
- Your weekends, learning it
- unpaid
- The workflow only Dave understands
- until Dave leaves
- Files the AI found that it shouldn’t have
- awkward
- The vendor moving the button
- quarterly
- Total
- more than the sticker
Thank you for subscribing. Please enjoy all 400 features.
Hiring an AI consultant
Tax invoice · one-off, hopefully
- The fee
- as quoted
- Your team’s time in workshops
- hours
- The strategy deck nobody opens
- if nobody owns it
- A tool you didn’t need
- if they also sell one
- Ringing them for every change
- if nothing’s handed over
- Software licences
- still needed
- Total
- depends who you hire
Thank you. We’ll be in touch. About phase two.
The DIY receipt is mostly time and risk. The consultant receipt is mostly about who you hire. Both are fixable. For DIY, start small and measure before buying more seats. For a consultant, insist on a fixed price, a written handover and the right to be told “you don’t need us for that part”. If you want the actual dollar ranges, the AI consultant cost guide has Melbourne prices, and our piece on the ROI of AI automation shows how to check whether either route pays for itself.
Worked example
One accounting practice, three routes
An illustrative example, not a client story. Picture a 25-person accounting practice in Melbourne. Every tax season the team drowns in client documents: PDFs, phone photos of receipts, bank statements, a shoebox scanned sideways. The partners decide it’s time to “do something with AI”.
Buy 25 AI seats
Everyone gets an AI assistant at the default A$31.40 a month. That’s A$9,420 a year plus GST. By month three, six people use it, mostly for emails and the odd summary. The calculator says about A$7,160 a year is going on seats nobody opens.
The document pile is untouched, because the assistant was never connected to the practice’s systems.
Fits when: the goal really was better emails.
Review first, then buy
A half-day walkthrough times the work and finds that client document intake eats the most hours at tax time. They keep AI seats for the eight people who use them, and switch on the document features already inside their practice software.
Someone owns it, there’s a 30-day check, and the licence bill drops.
Fits when: one workflow is clearly the expensive one.
Build the intake workflow
A workflow reads incoming documents, matches each one to the right client, checks what’s missing and sends a reminder. Anything doubtful goes to a person to review. It’s more work up front.
It’s worth it when the volume is high and the same mistakes keep costing rework.
Fits when: the volume is high and the data is sensitive.
For this practice we’d start in the middle. The licences aren’t wrong. They’re just aimed at the wrong job. A short review points the spend at the work that’s actually expensive, and only then decides whether anything needs building. If intake volume justifies it later, that’s when you look at the consultant-led route, and our document intelligence page and document AI vs OCR guide cover how that side works.
Accounting is one of the five industries we work in. The accounting automation page has more on where the hours go in a practice.
Going it alone?
A DIY AI rollout plan that actually sticks
If you’ve decided software on its own is enough, good. Run your AI implementation in this order and you’ll avoid most of the shelfware. (Yes, we wrote a manual. Of course we did.)
Pick one workflow and time it
Not “use AI more”. One job, like the weekly client report or the inbox triage. Write down how long it takes today, so you can tell later whether anything improved.
Check the AI you already own
Your email, accounting, CRM and document tools have probably grown AI features already. Switch those on before buying anything new.
Read the data terms
Check where your data is stored and whether it’s used to train the vendor’s models. The OAIC recommends not putting personal information, especially sensitive information, into publicly available AI chatbots.
Pilot with the people who have the problem
Three to five people, not the whole company. Seats for everyone on day one is how you get the report in the hero.
Train, and write down what works
Run a proper session. Then keep a shared page of the prompts and steps that work, so knowledge doesn’t live in one person’s head.
Measure after 30 days, then decide
Time the job again. If it’s faster, add seats. If it isn’t, find out why before spending more. This is the moment some help is worth paying for.
Before you sign anything
Questions to ask a software vendor or an AI consultant
Whichever route you take, the Australian Government’s Guidance for AI Adoption is blunt about where responsibility sits: “your organisation is ultimately accountable for how and where AI is used”. Buying the software or hiring the help doesn’t hand that over. These questions help.
Ask the software vendor
- Can you show proof it’s been properly tested? (The guidance literally says: “Ask for proof.”)
- Where is our data stored, and is it used to train your models?
- How do we get our data and settings out if we leave?
- What will the price per seat be at renewal?
- Which of our systems does it connect to, and which does it only “integrate with” in the brochure?
Ask the AI consultant
- Will you tell us if we don’t need you for part of this?
- Do you start with our process, or with a tool you sell?
- Is the price fixed, and what exactly is included?
- Who owns the code, the configuration and the documentation?
- What stays with our people, and how will a person review the AI’s work?
The Australian rules that apply either way
[Switches to serious face.] These apply whether you buy software, hire a consultant or both.
- Privacy and AI products. The OAIC’s guidance on commercially available AI products recommends due diligence before you adopt a tool, and not entering personal information, particularly sensitive information, into publicly available AI chatbots.
- Accountability across suppliers. The Guidance for AI Adoption asks you to “clarify supply chain accountabilities” when vendors, developers and integrators are involved. Write down who is responsible for what.
- Automated decisions. From 10 December 2026, privacy policies must explain when computer programs make, or substantially inform, significant decisions about people. Our automated decision-making review covers what to check.
The Privacy Act doesn’t cover every small business, so check whether it applies to you. If you handle data for clients who are covered, expect them to ask anyway.
Whichever route you take
What should stay human, with or without a consultant
Buying the software doesn’t buy you out of responsibility, and neither does hiring someone. If the AI gets it wrong, your client rings you. So the same split applies on every route.
Rules do
- Calculations and totals
- Moving data between systems
- Reminders and scheduled jobs
- Routing by fixed criteria
AI does
- Reading messy documents
- Sorting and classifying
- Summaries and first drafts
- Flagging what looks unusual
People do
- Approvals and sign-offs
- Exceptions and judgement calls
- Anything sent to a client
- Choosing vendors and consultants
That review step has a name: human-in-the-loop. Build it in from day one, rather than after the first embarrassing email.
Common mistakes
AI consultant vs software mistakes we see (and one we’ve made)
[Clears throat.] The last one is ours.
- Buying seats for everyone on day one. Start with the team that has the problem. Expand when the 30-day numbers say so.
- Hiring a consultant to “do AI”. That’s not a brief. Hire help to fix a named workflow with a number attached to it.
- Treating the launch email as the AI implementation. The real work is the training, the shared prompts and the check-in a month later.
- Choosing a tool before choosing the problem. Define what you want to improve first. The government says so too.
- Paying for a strategy with no build. A roadmap nobody owns is shelfware in PDF form.
- Skipping the permissions clean-up. AI that searches your files will find the salary spreadsheet shared with “everyone” in 2019.
- Assuming people read the manual. We do, so we assumed everyone else did too. Nobody reads the manual. Plan the training.
Our take
How Lumeio approaches AI consultant vs software
We don’t start with AI. We start with how your business actually works: where the hours go, what gets copied between systems, and which jobs people dread on a Monday. Then we pick the cheapest thing that fixes it properly. Sometimes that’s switching on a feature in software you already pay for. Sometimes it’s a simple rule. Occasionally it’s a custom workflow with AI in the middle and a person checking the output.
Whatever the AI implementation looks like, we leave behind the same things every time: written documentation, checks on the AI’s output, a review step for doubtful cases, and a handover so nothing depends on our phone number. If you’re also weighing up whether to buy or build the AI itself, our build vs buy AI guide covers that decision, and the custom automation vs Zapier, Make and n8n guide covers the no-code tools.
When we’d tell you to skip us
- The job is generic and the AI is already in your software
- Your calculator result was 70% or more and people are happy
- You qualify for free government help and haven’t used it yet
See our solutions, browse real automation use cases, or read about the Pain Point Audit we start every engagement with.
Who wrote this
Written in Melbourne by people who read the manual (all of it)
FAQ
Questions about AI consultant vs software
What is the difference between an AI consultant and AI software?
AI software gives you capability: tools that draft, summarise, search or read documents, priced per user per month. An AI consultant works out how to use that capability in your business: which process to fix, what needs AI and what needs a simple rule, how to set it up, how to train people and whether it worked. Software is the gym membership. The consultant is the personal trainer.
Do I need an AI consultant, or can I do AI myself?
You can do it yourself when the job is generic, happens inside one tool you already use, involves no sensitive information, and someone has a few hours a week to learn it. Bring in an AI consultant when the work crosses several systems, involves personal information or decisions about people, or when you’ve bought AI tools and nobody is using them.
Are AI tools like ChatGPT or Copilot enough for a small business?
For everyday jobs like drafting emails, summarising documents and meeting notes, often yes. Check the data terms first: the OAIC recommends not entering personal information, especially sensitive information, into publicly available AI chatbots. AI tools on their own won’t run a workflow that moves information between your systems, checks it against your rules and hands exceptions to a person. That needs some setup around the tool.
When is an AI consultant a waste of money?
When the job is generic and the AI is already inside software you pay for, when the brief is just “do AI” with no named problem, or when you get a strategy document and nobody owns the build. It’s also poor value if the consultant sells their own platform and every answer happens to need it. Ask for a fixed price and a written handover.
What does an AI consultant actually do?
A good one maps how the work moves today and times it, decides what needs AI, what needs ordinary automation and what stays with a person, then sets it up, tests it on your real examples, trains the team, measures the result and hands over documentation. Often the advice includes keeping the software you already own and switching on features you’re already paying for.
How much does an AI consultant cost compared with AI software?
AI software is usually priced per user per month. As of October 2026 Microsoft lists the Copilot Business add-on at AU$31.40 per user per month plus GST, paid yearly. AI consultants usually charge a fixed fee per review or project, or an hourly or day rate. Lumeio’s guide to AI consultant cost in Melbourne lists typical price ranges for each pricing model.
Is there free AI help for Australian small businesses?
Yes, with conditions. The National AI Centre publishes free guidance, including the Guidance for AI Adoption and a guide to choosing an AI solution. AI Adopt Centres offer free specialist help, but only to eligible SMEs in National Reconstruction Fund priority sectors. Small businesses can also get subsidised one-on-one digital advice through the federal Digital Solutions program. Check eligibility on business.gov.au.
What does Lumeio charge to help?
A 30-minute fit call is free. The Pain Point Audit is A$1,950 + GST for a half-day walkthrough of your processes, with a written report within 5 business days covering what to keep, what to cancel, what to build and what stays with your people, plus a fixed price for the first fix. The fee is credited in full against a first build of A$5,000 + GST or more signed within 60 days.
Find out whether you need a person, a product, or both
Bring the usage report, the vendor quote and the process everyone complains about. The Pain Point Audit is a half-day look at how work really moves through your business, with a written report in 5 business days: where the hours go, which jobs need AI and which need a simple rule, what to keep, what to cancel, what to build, and a fixed price for the first fix. If the software you own already does the job, we’ll say so. No gym membership required.
Fixed price A$1,950 + GST. Credited if you go ahead with a build.
Sources
- National AI Centre: Choose a solution (22 April 2026)
- National AI Centre and Department of Industry, Science and Resources: Guidance for AI Adoption, Foundations (first published October 2025)
- National AI Centre: AI adoption insights, December 2025 to February 2026 (May 2026)
- OAIC: guidance on privacy and the use of commercially available AI products (21 October 2024, updated 17 January 2025)
- OAIC: new resources on transparency for use of AI and automated decision-making (30 September 2026)
- business.gov.au: AI Adopt Centres (checked October 2026)
- Microsoft and LinkedIn: 2024 Work Trend Index (8 May 2024)
- BCG: AI Adoption in 2024, 74% of companies struggle to achieve and scale value (24 October 2024)
- Microsoft: Microsoft 365 Copilot pricing, Australia (checked 10 October 2026)