Why Have We Not Seen More Disruption of the Veterinary Experience From AI...Yet?
Despite the excitement surrounding AI in the veterinary world, its daily impact on the lives of vets and pet owners remains limited—for now. Just like in other industries, the change is coming, but not as quickly or smoothly as many had hoped.
When ChatGPT launched in November 2022, it marked a step change in AI’s capabilities. It took the world by storm, becoming life-changing in some niche applications. However, its transformative power has, for the most part, been confined to direct use cases—specifically within ChatGPT. The futuristic ideas that fuel AI enthusiasm require integration and customisation, and that takes time. After all, everyone started on the same timeline in November 2022.
AI in Veterinary
Take a veterinary example from my own experience: how can AI, like ChatGPT, help streamline communication with pet owners? The reality is that, right now, it can't—not without laborious human steps like copying and pasting text, or somehow recording calls and having ChatGPT respond via phone.
To be truly useful, AI must integrate seamlessly with Practice Management Systems (PMS), messaging platforms like WhatsApp and SMS, and then deliver the right prompt at just the right time. This level of integration is no small feat. But once it's achieved, AI will be able to access the necessary information and communicate effectively with clients.
Which LLM to Use?
ChatGPT is just one of many Large Language Models (LLMs). Since its launch, numerous models have emerged, each offering unique features, quirks, and costs. They all share the uncanny ability to hold human-like conversations, but now AI teams face a critical decision: which platform to choose? In a fast-moving world, teams often need to switch models, which means starting from scratch in terms of learning and training. That’s where tools like Langchain come in—it creates a flexible layer that makes it easier to shift AI development from one model to another. Even these tools, however, are in their infancy; Langchain, for example, only launched in October 2022.
Training your AI - Prompt Engineering
Developing the correct system prompt is a fun part of the process of teaching the AI your rules on tone of voice, etiquette and importantly, when to stop speaking! The prompts tend to grow longer as developers try to cover all bases and set up guardrails. For more control over how an LLM behaves, teams often train an existing model, feeding it trusted data and tailoring it to respond in ways that align with specific needs.
Big Limitations
Money
The next major hurdle for widespread LLM adoption is money. When ChatGPT first launched, it appeared to be free, but serious use of these models is anything but. The compute power required to run LLMs is immense, requiring powerful servers with high energy demands. While costs are gradually coming down, they’re still significant enough to make businesses think twice before fully embracing AI.
Data Security
Another roadblock is data security. AIs must not be fed personal data that could unintentionally resurface in conversations, making data cleaning essential. There’s also a pervasive mistrust around AI's use of data, especially when it’s difficult to guarantee that sensitive information won’t be used to improve the platform as a whole.
Accuracy
Then there’s the issue of accuracy, which I discussed in my last blog. We tackled this head-on when re-building our symptom checker. It’s functional, integrated with its first PMS, and accurate—but it’s still too expensive per use to offer to the public, given the high traffic on our website.
The Future of AI in Veterinary
Do these challenges mean AI won’t change the world? Absolutely not. I believe we’ll see a complete transformation within five years. But these obstacles explain the current lull in large-scale changes as corporations work to perfect their programs and startups take the time to secure the necessary integrations and exposure.
Across industries, teams are tackling the same issues: integrating AI, training models, and deciding which platforms to invest in. I often picture thousands of teams, just like ours at Digital Practice, wrestling with these challenges. It’s an exciting time to be in the AI space, especially for those of us already working with PMS integrations and communication systems. The list of veterinary pain points that AI can solve is long, and as these tools evolve, I’m confident we’re on the verge of seeing real, meaningful changes in how we care for patients and communicate with pet owners.
