Put AI in your CRM, not just your browser tab

SmartCompany recently published an opinion piece arguing that 2026 is a make-or-break year for Australian SMEs. The author, Ryan Williams from the Australian Centre for Business Growth, lands on a point we have been making to clients all year:

“Stop relying on openly available LLMs and deploy models trained on your own data and integrate AI directly into your CRM and operational systems.”

In other words, having ChatGPT open in a browser tab is not an AI strategy. It is a very capable intern who knows nothing about your business, forgets everything between shifts, and never touches your actual systems.

The businesses getting real value from AI in 2026 are not prompting harder. They are wiring AI into the places where work already happens: the CRM, the inbox, the quoting process, the job pipeline.

The difference between using AI and embedding AI

Using AI looks like this: a staff member copies an enquiry email into ChatGPT, asks for a reply, pastes it back into the inbox, then manually updates the CRM. If they remember. If there is a CRM.

Embedding AI looks like this: the enquiry arrives, gets classified and summarised automatically, a draft reply appears for review, and the CRM record is created with the right tags before anyone has touched a keyboard.

Same technology. Completely different outcome. The first version saves a few minutes when someone thinks to use it. The second version saves those minutes on every single enquiry, forever, without depending on anyone’s memory or motivation.

Williams makes the same distinction: treat AI as a practical tool “for reducing work process friction and manual effort by employees”, not as a novelty. The friction is in the workflows, so that is where the AI needs to live.

Three automations a Melbourne service business can ship this quarter

These are not moonshots. Each one uses tools that already exist (Make, Zapier, n8n, or a direct API integration), connects to a mainstream CRM, and can go live in weeks, not quarters.

1. Enquiry triage and first response

Every enquiry from your website form, email, or Google Business Profile gets automatically classified: what service, how urgent, roughly what size. The AI drafts a tailored first reply, creates the CRM contact with the right pipeline stage, and flags anything unusual for a human.

For a trades, legal, accounting, or consulting business, this is usually the highest-value starting point. Speed to first response is one of the strongest predictors of winning the job, and “we got back to them in four minutes instead of four hours” is a competitive advantage that costs almost nothing to run.

2. Quote follow-up that actually happens

Most service businesses are excellent at sending quotes and terrible at following them up. The quote goes out, the pipeline entry goes quiet, and the lead buys from whoever called them back.

The automation: when a quote sits unanswered for a set number of days, AI drafts a follow-up that references the actual quote details, the original enquiry, and any previous conversation. A human approves it with one click. The CRM stage updates automatically based on the response.

This is the kind of revenue that is already sitting in the pipeline. No new marketing spend, no new leads required. Just fewer quotes dying of neglect.

3. Post-job reviews and client reactivation

When a job is marked complete or an invoice is paid, the system waits a sensible interval, then sends a personalised review request mentioning the actual work done. Months later, dormant clients get a relevant check-in: the accountant’s clients hear from them before EOFY, the landscaper’s clients before spring.

Reviews compound your local search visibility. Reactivation campaigns convert far better than cold outreach because the trust already exists. Both run quietly in the background once they are built.

Why a human stays at the helm

None of this means letting AI talk to your customers unsupervised. Williams is direct about it: “A human at the helm is critical in most scenarios to manage the risk.”

The pattern that works is draft-and-approve. AI does the assembly: reading the context, pulling the CRM history, writing the first version. A human does the judgement: approving, editing, or overriding. That keeps quality high while still removing most of the manual effort, which was the entire point.

It also keeps the failure modes boring. A bad AI draft that a human catches is a non-event. A bad AI email sent automatically to your best client is a Tuesday you will remember.

Start with the friction, not the technology

The wrong way to start is “we should do something with AI”. The right way is “where does our team lose the most hours to repetitive work, and which of those workflows touch the CRM?”

For most service businesses the answer is somewhere in enquiry handling, quoting, scheduling, or follow-up. Pick one, ship it, measure the hours saved, then move to the next. That is how the gap Williams describes gets closed: not with a transformation program, but with one boring, reliable automation at a time.

If you are not sure where your biggest friction is, that is exactly what a digital systems audit is for. We map how enquiries, quotes, and jobs actually move through your business, find the manual steps that should not exist, and recommend what to automate first. Get in touch if you want a clear-eyed look at yours before the quarter gets away from you.