You’ve decided your practice needs AI. Maybe it’s clinical documentation, maybe it’s patient flow optimisation, maybe it’s analysing the mountain of CPAP telemonitoring data sitting unused on your servers. Whatever the use case, you’ve realised you need outside help to make it happen.

Good. Self-awareness about capability gaps is the first step. The second step is choosing the right partner, and this is where a lot of healthcare organisations get it wrong.

Here’s how to avoid the most common mistakes.

Healthcare Experience Is Non-Negotiable

This is the single most important filter. The company that builds recommendation engines for retail doesn’t understand clinical workflows, regulatory requirements, or healthcare data constraints.

When evaluating potential partners, ask these questions:

Which healthcare organisations have you worked with? Ask for specific examples, not vague references to “health sector clients.” Which hospitals, clinics, or health systems? What did you build? What were the outcomes?

Do you understand health data regulations? In Australia, that’s the Privacy Act and the Australian Privacy Principles. In the US, HIPAA. In the EU, GDPR plus specific health data directives. An AI consultant who can’t speak fluently about these frameworks hasn’t done meaningful work in healthcare.

Can you work with clinical data formats? HL7, FHIR, DICOM, EDF (for sleep study data) — healthcare has its own data standards. If the consultant’s experience is limited to clean, structured datasets from other industries, they’ll underestimate the complexity of healthcare data integration.

Technical Competence vs Technical Theatre

It’s surprisingly easy for consultants to sound technically sophisticated without having the depth to deliver. Here are some red flags and green flags:

Red flags: buzzword overuse without concrete explanations, promising “AI” when simpler automation would suffice, claiming 99% accuracy without discussing context, and no discussion of data quality or validation.

Green flags: honesty about what AI can and can’t do, clear explanations without jargon, asking lots of questions before proposing solutions, and proposing small starts with iteration.

If every problem they see needs a neural network, they’re selling hammers, not solving your problems.

The Discovery Phase Matters

Before any development contract is signed, there should be a paid discovery phase where the consultant examines your data, observes your workflows, and produces a detailed assessment of what’s feasible. Consultants who skip discovery and jump straight to building are either overconfident or desperate for the contract.

During discovery, evaluate how well they listen (the best spend 80% of the time asking questions), whether they observe real workflows, and whether the discovery report contains specific recommendations with cost estimates — not generic observations about “digital transformation.”

References and Track Record

Call their previous clients. Not the ones they suggest — ask for the full list and pick your own. Questions to ask:

• Did the project deliver what was promised, on time and on budget?
• How did the consultant handle problems and setbacks?
• Is the solution still in use, or did it get abandoned?
• Would you hire them again?

That last question is the most revealing. A hesitant “yes, probably” tells you a lot.

Team 400 is an example of a consultancy that specialises in AI for professional services and healthcare — they understand that the technology has to work within clinical realities, not despite them. But whoever you choose, the principle is the same: domain expertise plus technical competence, verified by real references.

Ownership, IP, and Ongoing Support

Clarify intellectual property ownership before work begins. Who owns the trained models? Who owns the code? What happens if the relationship ends — can you take the software and have someone else maintain it? Get these answers in writing before the first line of code is written.

Also consider ongoing support. AI systems need maintenance — models drift, regulations evolve, staff turn over. A brilliant system that nobody maintains becomes a liability within a year.

Start With the Problem, Not the Technology

The best AI consulting engagements start with a clearly defined clinical or operational problem and work backward to the technology. The worst start with “we want AI” and go looking for places to put it.

Before you talk to any consultant, write down: what’s the specific problem you want to solve, who does it affect, what does it cost you (in time, money, or outcomes), and how would you measure success?

If you can’t answer those questions, you’re not ready for a consultant. You’re ready for a strategy session with your own team first.

The right AI partner will make your practice measurably better. The wrong one will take your money and leave you with a PowerPoint. Choose carefully.