What's Next: Our AI Roadmap for Year Two

A year ago, we started building AI tools for our dental practice. Not buying off-the-shelf software. Not subscribing to another SaaS platform. Actually building — custom tools designed around the way we work, running on our own hardware, solving problems we faced every day.

If you've been following along, you've seen the results. We wrote about every tool we built in year one and what each one actually does. The short version: an AI phone receptionist, automated email management, push-to-talk dictation, document scanning, a memory system, and more — all running locally, all integrated, all saving real time and real money.

Year one taught us three things.

First, every tool started as "can we even do this?" and ended as "how did we ever live without this?" The gap between scepticism and dependence is surprisingly short when the tool genuinely works.

Second, the biggest value isn't any single tool. It's the integrated system. When the phone agent, the email classifier, the dictation tool, and the memory system all talk to each other, the whole becomes dramatically more useful than the parts. Automation compounds.

Third, and most importantly — building for ourselves proved the concept. We know these tools work because we've used them every day for a year. We've fixed the bugs, smoothed the rough edges, and watched staff go from cautious to enthusiastic. That's not theory. That's evidence.

So now the question is: what do we do with everything we've learned?

Year two: from one practice to many businesses

Year one was about us. Year two is about you.

Everything we built was designed for a dental practice — our dental practice, specifically. It works brilliantly for us, but it's not something we could hand to a plumber or a physiotherapist and say "here, use this." The workflows are different. The terminology is different. The integrations are different.

Year two is about taking what we've built and making it adaptable. Same core technology, same architecture, same local-first approach — but configurable for different industries, different team sizes, and different ways of working.

Here's what we're focused on.

1. Packaging our tools for other businesses

Our AI phone receptionist knows dental appointment types, understands patient queries, and books into a dental practice management system. That's great for us, but useless for a restaurant.

We're rebuilding the configuration layer so that the same voice agent can be set up for any small business. A plumber's receptionist needs to handle emergency callouts and quote requests. A physio's receptionist needs to manage treatment plans and rebooking. A restaurant's needs to handle reservations and dietary questions. The underlying AI is the same — fast speech recognition, natural conversation, intelligent routing — but the knowledge, vocabulary, and integrations are different.

The same applies to our email automation, our dictation tool, and our document handling. We're separating what's universal from what's industry-specific, so the universal parts can be deployed quickly and the specific parts can be configured without starting from scratch.

2. Multi-location support

A single-location business is straightforward. But a surprising number of small businesses operate across two to five locations — dental groups, physio chains, trade companies with multiple depots, accounting firms with a city and suburban office.

Multi-location introduces problems that single-site doesn't have. Shared staff rosters across locations. Location-specific knowledge — "the Palmerston clinic does implants, the city one doesn't." Centralised reporting that shows what's happening everywhere. Call routing that sends callers to the right branch.

We're building for this because it's a gap nobody else is filling well. The big enterprise platforms handle it, but they cost a fortune and they're overkill for a business with three locations. The small business tools ignore it entirely. There's a sweet spot in the middle, and that's where we're aiming.

3. Industry-specific starter templates

Starting from a blank slate is slow and expensive. If a physiotherapy clinic wants an AI phone agent, they shouldn't have to describe every appointment type, every common question, and every workflow from scratch.

We're building starter templates for the industries where we see the most demand: dental, trades, allied health, and professional services. Each template comes pre-loaded with industry-standard appointment types, common caller questions, typical email categories, and standard workflows. You start from a solid foundation and customise from there.

Think of it like a website template. Nobody builds a website from an empty HTML file anymore. You pick a template that fits your industry, swap in your branding and content, and adjust from there. We want AI setup to work the same way — 80% done on day one, then fine-tuned to your specific business over the first few weeks.

4. Better voice capabilities

Voice AI is improving every month, and we're investing heavily in keeping our systems at the leading edge.

Three areas matter most. First, more natural conversations — fewer awkward pauses, better handling of interruptions, more human-like pacing. The technology for this is advancing rapidly, and each improvement makes the difference between a caller thinking "that's pretty good for a robot" and "wait, was that a person?"

Second, better accent handling. This is critical in Australia. Our cities are genuinely multicultural, and a voice agent that only understands standard Australian English is going to frustrate a lot of callers. We're working on accent robustness across the range of English accents you hear in any Australian city — South Asian, East Asian, Middle Eastern, European, and everything in between.

Third, faster response times. Every millisecond of delay in a voice conversation erodes trust. People are remarkably sensitive to conversational timing — even a half-second pause that wouldn't bother you in a text chat feels wrong on the phone. We're optimising every part of the pipeline: speech recognition, language model inference, and speech synthesis. Our local-first architecture gives us a fundamental advantage here, because we're not waiting for data to make a round trip to a cloud server.

5. Self-service monitoring dashboard

Business owners want to know what their AI is doing. Not the technical details — they don't care about model weights or inference latency. They want to know: how many calls were handled? How many emails were processed? How much time did we save? How much did we avoid spending on staff or outsourcing?

We're building a simple, clear dashboard that answers those questions. Calls handled this week, with outcomes. Emails classified and routed. Documents processed. Time saved, estimated in hours. Cost avoided, estimated in dollars. Trends over time, so you can see the value compounding month over month.

No vanity metrics. No complicated analytics that require a data scientist to interpret. Just the numbers that matter to someone running a business: is this working, and is it worth what I'm paying for it?

6. Community and education

We've learned a lot in the past year, and we want to share more of it. More blog posts like this one. Case studies from real businesses. Possibly workshops — small, practical sessions where business owners can see AI tools in action and ask questions without the sales pressure.

The AI conversation in Australia is dominated by two extremes: breathless hype about how AI will change everything, and fearful warnings about how it'll take everyone's jobs. Neither is helpful. The reality is more nuanced, more practical, and frankly more interesting. We want to help Australian small business owners understand what AI can actually do for them today — not in some hypothetical future, but right now, with tools that exist, at a price they can afford.

If you want a starting point, read our post on what AI can actually do for a small business. It's the most honest overview we've written, and it's a good foundation for understanding what's realistic.

What we're deliberately NOT doing

Roadmaps are as much about what you say no to as what you say yes to. Three things we're not pursuing.

Enterprise sales. We're not building for corporations with 500 employees and a dedicated IT department. We're building for the owner-operator with 5 to 50 staff who doesn't have a CTO and shouldn't need one. Small business, always.

Generic AI chatbots. We've written about why most chatbots are useless, and we stand by it. We're not going to slap a chat widget on your website and call it AI. Every tool we build has a specific job, measurable outcomes, and a reason to exist beyond "it's AI."

Anything that forces your data offshore. Our local-first architecture means your data stays on your hardware, in your building, under your control. If you explicitly choose a cloud option for convenience or scale, that's your call. But we'll never make it the only option, and we'll never require your data to leave Australia.

What would save you the most time?

We have strong opinions about what to build, but we'd rather build for real problems than imagined ones. If you're running a small business and there's a task that eats your time, frustrates your staff, or costs more than it should — we want to hear about it.

Not "I think AI could maybe help with..." We mean the specific thing you'd automate tomorrow if you could. The phone calls that nobody answers. The emails that pile up. The paperwork that takes hours. The scheduling that drives everyone mad. Tell us what it is, and we'll tell you whether we can solve it.

Year one proved that custom AI works for a real business. Year two is about proving it works for yours. We're building fast, we're building in the open, and we're building for Australian businesses that are tired of paying too much for software that does too little.

If you want to be part of what we're building — or just want to have an honest conversation about what AI could do for your business — get in touch. We'd love to hear from you.