Start With the Problem, Not the Platform: an AI guide for business owners
Artificial intelligence seems to be everywhere. Every week, another tool promises to save time, transform your business and possibly make your coffee before your first appointment.
For anyone who works with people’s personal information, adopting AI is not as simple as opening an account and encouraging the team to “have a play.” Used well, AI, automations and bots can reduce repetitive work and free people to focus on higher value activities. Used carelessly, they can create privacy, trust, accuracy and governance risks particularly when client personal information is involved.
The goal should not be to use as much AI as possible. It should be to use the right tools, for clearly defined purposes, within appropriate boundaries.
What does “using AI” actually mean?
For most small and medium businesses, AI does not need to mean automating clinical decisions or replacing professional judgement.
It may simply mean using automations, bots or agents to:
Create a first draft of a routine email
Turn rough notes into a structured administrative summary
Develop templates, checklists or internal procedures
Generate ideas for newsletters, blogs and social media
Summarise meeting notes
Organise schedules or identify diary inefficiencies
Analyse operational data
Draft the structure of a report for review and completion by a professional
The key words are draft, support and review. Different systems can help produce a starting point, but the professional should always remain responsible for checking accuracy, context, tone and suitability.
This matters because generative AI can produce information that sounds convincing but is incomplete, biased or simply wrong. While current research identifies substantial opportunities for the use of AI in areas like healthcare, it also highlights risks involving accuracy, bias, privacy and the potential for fabricated references or information (Sallam,M. 2023).
The compliance question is often skipped
Just because a tool is being marketed to a business it does not automatically make it suitable for use with private or sensitive information. In Australia, certain types of client information are considered sensitive information under the Privacy Act 1988. This can include information like health data, sexual orientation, racial or ethnic origin, criminal records or religious beliefs.
To learn more about your compliance responsibilities you can read my article “Should I put that in ChatGPT?”
Three common AI mistakes
The first is buying too many tools. Subscriptions accumulate quickly, while teams end up with overlapping systems and no consistent process.
The second is putting sensitive information into consumer-grade AI products without examining the privacy settings, contractual terms, data location or intended use of submitted information. Removing a client’s name is usually not enough if the remaining details could reasonably identify them.
The third is chasing every new release. The AI market moves far faster than most practices can safely assess, implement and govern new technology. Constant experimentation can create fragmented processes and confused staff.
Where can AI save real time?
Professionals rarely complain that they have too much time and not enough paperwork!!!! Documentation, email, scheduling, reporting and business administration can consume hours that could otherwise be spent with clients, supporting staff or doing other life stuff!
The use of tools like AI scribes has resulted in significant reductions in task load and professional burnout, and improve perceptions of efficiency and documentation quality (Shan, et al, 2024). This suggests that the dreaded admin is a good place to start looking for opportunities to introduce AI into your business. Start by reflecting on admin tasks that you do every day, I recommend that people keep a note pad next to the computer and just jot down all the admin type things that you do over a couple of days. Then look for a task that is repetitive, low risk and easy to check, like drafting email templates, developing generic resources for clients or creating the minutes for team meetings. This is your starting point. Only when you have a start point should you start looking for a solution.
A simple decision framework
Before adopting an AI tool, ask:
What specific problem are we solving?
Can we name the task, the current time or cost involved, and what success would look like?What information will the tool receive?
Is it public, internal, personal, health-related or identifiable—and does the tool genuinely need that information?Where does the information go?
Check storage location, overseas processing, retention, access, security arrangements and whether the data is used for model training.Who checks the output and remains accountable?
Define the human review process, especially where content could affect clients, clinical documentation, funding or regulatory obligations.Can we trial it safely before committing?
Test one low-risk workflow using fictional, synthetic or appropriately de-identified data, then measure the result before purchasing another shiny subscription.
Final thoughts
AI can create genuine efficiencies in your businesses. But successful adoption is less about becoming an “AI-powered business” and more about making deliberate, defensible decisions.
Choose one real problem. Start with low-risk information, establish clear boundaries, measure whether the tool saves time. Then decide whether it has earned a permanent place in your systems. That is how AI becomes useful rather than simply becoming another expensive tab left open in the browser.