We are living in amazing times.
The capabilities of online AI services such as ChatGPT, Co-Pilot and Claude are expanding at a dizzying pace, helping us draft ever-more-sophisticated documents, or to interact naturally through text and voice with software solutions that can serve up powerful analytical results and recommendations to improve business outcomes.
For enterprise leaders, the downside of the frenetic pace of these advances is that it is easy to feel that we are losing control. As our dependency on the online AI services grows, we can lose control of budgets, and we are at risk of losing our grip on security.
Few executives are comfortable with the idea that sensitive organisational information, or worse yet, the private data of customers, being entered into AI systems with only hypothetical promises that the data is “safe.”
Do we have an alternative?
In many cases, the answer is “Yes! An in-house AI strategy that you can be developed.”
There is a growing number of powerful miniaturised versions of the huge Large Language Models (LLMs) that lies at the heart of the online AI services. These models, often referred to as Small Language Models (SLMs) can run on the private infrastructure of any organisation.
Amazingly, as I write, I am running two powerful little AI services entirely within my laptop that can do multiple tasks for me, including searching internal corporate documents, or allowing me get insights from huge, dense data sets using my voice alone. Of course, none of these SLMs is individually as broadly capable as an online AI service, but they all have strengths for different tasks.
This approach of applying smaller models to specific jobs is a crucial component of the “Agentic AI” that you are probably hearing a lot about lately. I will write more about Agents in a separate article.
To use this approach, your infrastructure will require your team or your solution vendor to set up different internal AI models for different tasks, and you will want the ability to upgrade to newer, more capable models when they become available. This is not too different from the care and feeding of other internal software tools
At Oii.ai we take a hybrid approach to AI, giving our customers the flexibility and power of online AI services when appropriate, for example to map out a problem-solving strategy, and our own locally-deployed in-house developed AI models to generate analytical results on actual customer data. The combination is cost-effective, secure, and powerful.
Bottom line, organisations of any size can now get the enormous benefits of modern AI without losing control of our budgets or our data.
That’s great news.