AI is everywhere. Are you ready?

“We’re not ready for AI, but we’re doing it anyway.”

So said the majority of respondents to the 2024 CIO Edge Survey, with just 9% of Australian organisations claiming they are at least somewhat ready for AI. Meanwhile, a whopping 30% are fully unprepared for AI’s impact. Our initial reaction to this?

AI offers a lot of potential to change many things about how we work and live. You may be feeling a lot of pressure to do something about it. Our best advice? Don’t get carried away and definitely don’t panic. Instead, spend time now laying the foundations for a productive AI experience in the years to come.

Here are a few suggestions, taken from customer anecdotes and our own experience.

Think beyond generative AI models.

Even though generative AI is in the news of late, organisations can find value in both discriminative AI (used to classify or predict data) as well as generative AI (create new data). Your data infrastructure should include support for data analytics, data science, discriminative AI as well as generative AI.

Have realistic expectations of what AI can do for you.

Quantify expectations about what AI can deliver for your organisation. Most experts would agree that while AI is a useful technology, it won’t solve every challenge you’re currently facing. A better approach is to develop a small handful of practical, value-based use cases that will address some key pain points; rather than starting with the belief that the technology will transform every aspect of your operations.

An example from the Macquarie Cloud Services world: our managed services are powered by intelligent automation, or code-driven deliverables, 24×7. Our AIops engines help us identify, prioritise and direct actions into the right queues. Our return on investment for this use case is simply time saved and improved service quality as we can diagnose, categorise and remediate job errors faster. For our customers, the benefits include a significant reduction in noise and timely action on true positives.

Look backwards to move forwards.

IDC Nov 23 Infobrief revealed that security breaches and exposed customer data are the top concerns of companies implementing AI, as did the ADAPT cybersecurity survey. (“We’re deploying AI much faster than we can manage the risks”!)

There is a lot to be said for implementing established, tried and tested frameworks into evolving areas such as AI. A few examples of Zero Trust principles:

1. Explicitly validate user permission to APIs and data using OAuth. The application identity should not be used to access segregated grounding data found in vector databases, APIs, files, etc.
2. Limit access with request filtering, which allows requests that originate only from specified IP addresses, IP ranges, or from a list of subnets. Verify integrity of all data through detailed logging, auditing trails, verification frameworks, and oversight procedures. Consider data classification and catalog tools to help ensure that data is used by intended users for intended purposes.
3. Continuously monitor for anomalies, including inputs and outputs, not just traditional signals of user behaviour and data leak prevention policies. Content filter systems run both the prompt and completion through an ensemble of classification models aimed at detecting and preventing the output of harmful content.

Data security posture management is a key element of our managed services, and we are keeping an eye on the rapid advances in AI security posture management as well.

Build on the right foundation.

The 2024 CIO Edge Survey, revealed organisations with a stronger information and data architecture are, on average, eight times more prepared than those with a weak data architecture.

Many of our customers are considering a lakehouse architecture, which combines the structure of a traditional data warehouse with the scale and flexibility of a data lake. Data engineers, business analysts and data scientists can share a single copy of the data, stored in an open table format.

We see the lakehouse as the foundation for all AI and ML workloads, where data is collected, stored, processed and transformed, to be used by discriminative AI and large language models (LLMs) alike.

Microsoft has made some significant data and AI-powered updates over the past year. Ask us about Microsoft Fabric – we think having one copy of data for use with multiple analytical engines can contribute to payback investments.

Garbage in, garbage out.

A November 23 IDC InfoBrief quoted a key challenge – “We needed to clean and provide data that the models find useful.” Data hygiene and transformation exercises aren’t the most exciting projects in the IT landscape, but they’re critical for reliable and insightful predictions or outputs.

Innovate in Azure with a safe pair of hands.

If you’re on Microsoft Azure and have AI plans, we’re here to help. Our managed Azure solution provides a safe pair of hands while you build, test and learn, thanks to the insights and capability of our technology and team of 120-plus cloud experts.

We’ll provide the support to make sure your Azure estate is secured, managed and governed in accordance with the Well-Architected Framework principles. Find out why Microsoft recognises us as an Azure Expert Managed Service Provider. Reach out to us today at 1800 004 943 or drop us an email at enquiries@macquariecloudservices.com to explore how we can help you.