AI Usage Best Practices for SMEs

By Kelly Eldridge

When you’re running a small business on tight margins, the idea of adding another subscription, especially for AI that might not even work, feels risky. The noise around AI right now is deafening, and it’s hard to separate genuine value from something that’s simply expensive and performative.
So here’s my honest take: don’t invest in AI because you think you should. Invest in it when it solves a problem you actually have.

Start with the friction points

Before you touch a single AI tool, map your actual bottlenecks. Where are you or your team spending hours on tasks that feel soul-crushing? Where are errors happening repeatedly? Where would an extra 10 hours a week genuinely change your capacity?
If you can’t answer those questions clearly, you’re not ready to evaluate AI tools yet. You’re still figuring out what problem needs solving.

The “Must-Have” List (With Human Oversight)

I’m hesitant to give a prescriptive list because every business is different, but here are categories where AI can deliver genuine value for SMEs:

1. Writing and Communication Tools (e.g., ChatGPT, Claude, Grammarly etc)

What they do: Draft emails, refine copy, brainstorm ideas and summarise documents.
Pros: Save hours on routine writing, help non-writers produce clearer communication and speed up content creation.
Considerations: Output requires editing and fact-checking, although they can sound generic without thoughtful prompting and monthly costs can add up.
Reality check: These work if you’re spending significant time writing. If you send three emails a week, don’t bother. If you’re drafting proposals, reports, or client communications daily, the time savings are real.

2. Scheduling and Admin Automation (e.g., Calendly with AI features, motion.io)

What they do: Automate meeting scheduling, prioritise tasks, manage calendars intelligently
Pros: Eliminate calendar tennis, reduce no-shows, free up admin time
Considerations: Require initial setup time, can feel impersonal for relationship-heavy businesses, integration issues with existing systems
Reality check: Worth it if you’re drowning in scheduling back-and-forth. Not worth it if you have three meetings a month.

3. Customer Service Chatbots (e.g., Intercom, Drift)

What they do: Answer common customer questions, route queries, provide 24/7 basic support
Pros: Reduce repetitive support queries, improve response times, capture leads outside business hours
Considerations: Frustrating when they can’t handle nuanced questions, require significant setup and training, risk damaging customer relationships if poorly implemented
Reality check: Only invest if you’re getting the same questions repeatedly and have documented answers. Don’t use AI as a replacement for human connection in relationship-driven businesses.

4. Accounting and Bookkeeping AI (e.g., Xero with AI features, QuickBooks)

What they do: Automate expense categorisation, flag anomalies, predict cash flow
Pros: Reduce manual data entry, catch errors early, provide financial visibility
Considerations: Still require human oversight for accuracy, integration complexity, learning curve
Reality check: Worth it if you’re spending hours reconciling expenses or struggling with cash flow visibility. The time savings compound over months.

5. Design and Visual Content (e.g., Canva AI features, Adobe Firefly)

What they do: Generate images, suggest layouts, remove backgrounds, resize designs
Pros: Dramatically reduce design time for non-designers, maintain brand consistency, create professional-looking content quickly
Considerations: Generic outputs without customisation, IP concerns (more on this below), subscription costs
Reality check: Useful if you’re creating social content, marketing materials, or presentations regularly. Don’t pay for this if you create one flyer a quarter.

6. No-Code/Low-Code Development with AI (e.g., Base 44, Lovable)

What they do: Build functional websites, web apps, or prototypes through natural language prompts, without traditional coding
Pros: Dramatically lower barrier to building digital products, rapid prototyping, significantly cheaper than hiring developers for MVPs or simple tools, iterate quickly based on user feedback
Considerations: Limited customisation for complex features, potential technical debt if you outgrow the platform, you still need to understand what you’re building (AI can’t read your mind), hosting and scaling considerations
Reality check: Game-changing if you need a functional website, internal tool, or MVP but don’t have a developer budget. Not suitable for complex enterprise software or highly customised platforms. Perfect for testing ideas quickly before investing in proper development. Be aware that what AI builds might look functional but have underlying issues you won’t spot without technical knowledge. Critical consideration: these tools may not implement proper security protocols by default. If you’re handling customer data, payments, or sensitive information, you absolutely need a developer to audit what’s been built before going live. Don’t assume AI-generated code follows security best practices. It often doesn’t.

Addressing IP Concerns

This is a legitimate worry, and honestly, the legal landscape is still evolving. Here’s what I’d advise:

Never upload anything proprietary or confidential to free AI tools. Assume anything you put into ChatGPT, free image generators, or similar platforms could potentially be used in training data or seen by others. If it’s exclusive designs, client information, trade secrets, or competitive IP, keep it out.

Read the terms carefully for paid tools. Many paid AI platforms (like Claude for Work, ChatGPT Enterprise, or Adobe Firefly) have specific terms stating they don’t train on your data. But “don’t train on it” doesn’t always mean “don’t store it.” Understand the difference.

For designers specifically: If you’re uploading exclusive designs to AI platforms for editing or generation, you’re taking a risk. The IP ownership around AI-generated content is murky. Some jurisdictions are still figuring out whether AI outputs can even be copyrighted. My advice? Use AI for inspiration, rough drafts, or internal brainstorming but don’t rely on it for work you’re selling as original or exclusive to a client.

Create internal guidelines. Establish clear rules for your team about what can and can’t go into AI tools. Make it explicit: client data, proprietary methods, unreleased designs are all off limits.

The Bottom Line

AI is powerful when it solves real problems. It’s an expensive distraction when it doesn’t.
Before paying for any AI tool, ask yourself:

1. What specific task will this handle that’s currently taking too much time?
2. How will I measure whether it’s actually saving time or creating value?
3. Does this fit into our existing workflows, or will it require everyone to change how they work?
4. What’s the worst-case scenario if this tool fails or if our data is compromised?

If you can’t answer those questions confidently, don’t buy it yet.
At Quickli, we’re intentional about building purposeful AI tools that support real decision-making in the moments that matter for our brokers, not AI layered on for the sake of it. That same principle should guide how SMEs adopt AI: solve real problems, fit existing workflows, create measurable value.
Everything else is just noise.

About the author: Kelly Eldridge is Chief of Staff at Quickli, Australia’s leading technology platform for mortgage brokers This is an opinion column. The views and opinions expressed in this article are those of the authors and do not necessarily reflect the official policy or position of this publication.