How Australian IT management can use external data to power internal AI tools
Article by ManageEngine & Zoho Labs product manager Ramprakash Ramamoorthy.
Artificial intelligence (AI) tools are only as good as the data that’s used to train them. This makes them a powerful asset for organisations like e-commerce enterprises, and other companies whose operations generate a high volume of accurate, real-time information.
Conversely, for IT managed service providers (MSPs), low data environments with minimal processing capabilities are a less fruitful source for the type and volume of information needed to turn AI tools into seriously useful solutions for operations monitoring, help desk, chatbots, and other automated, IT management functions.
Opting out isn’t an option
At home and abroad, investment in AI systems is booming, with spending in the Asia Pacific region (excluding Japan) expected to reach USD 21.4 billion by 2023, up from USD 6.2 billion in 2019.
While AI adoption is still in its infancy, industries are "spending on this technology at scale" to improve the efficiency of operations and tasks, notes IDC Associate Market Analyst (APAC), Ritika Srivastava.
Seeking data from alternative sources
IT MSPs in Australia frequently cater to small and medium-sized clients that generate minimal data for the MSPs to work with. Moreover, those clients can be a diverse bunch drawn from a wide range of industries, and with few similarities in their operating models and infrastructure. This means developing one-size-fits-all AI solutions is not an option for service providers.
However, that doesn’t make the uptake of AI tools an unrealistic proposition for them, so long as they’re prepared to approach the issue in a different way. In many circumstances, globally available data can be a perfectly acceptable substitute for personally harvested data, particularly if it’s bolstered by historical help desk data.
Once an initial AI model has been created, it can be fine-tuned by opening it up to a small percentage of users for quality assurance testing—the adjustment of algorithm parameters and algorithms themselves, as well as balancing the data set.
Making users part of the process
The reality is that AI systems are not all-knowing, not perfectly generalisable, and in practice their accuracy varies anywhere between 70-80%. Despite less-than-perfect accuracy, the results produced by AI need to be trusted by the end users who will be working with those results.
If MSPs want customers to trust AI-driven management solutions, its recommendations need to be explainable, and come with a quantifiable confidence measure so users can interpret the results and take further actions based on the AI engine's output.
Seeking customers’ feedback throughout the journey, and acting upon it if users believe AI recommendations and reality are out of sync, will ensure the technology becomes a valuable addition to the MSP's arsenal of solutions.
Looking to the future
Bypassing the AI revolution is not a viable option for Australian IT MSPs, especially since experts have warned companies that they may fall behind the competition if they don’t embrace the technology. Seeking innovative ways to enjoy the powerful advantages finely honed AI tools can offer should be an incentive for providers that hope to survive and thrive in the 2020s and beyond.