At first glance, artificial intelligence and the IT department may seem a bit like the popular kid and the geek, an unlikely alliance in the enterprise. But just as the new technology is proving its value in streamlining IT operations (AIOps), it has similar emerging potential in IT asset management. Soon, AI can help companies save money while improving service quality and availability.
While these tools haven’t yet arrived in the marketplace, companies can start now to evaluate their potential and understand what’s required to put them to work. CIOs also need to know how the technology works to solve specific ITAM challenges. Here are four use cases where I think AI and ITAM will become a powerful tandem.
Predicting hardware retirement
Companies can use machine learning (ML) applications to more accurately plan the retirement of crucial hardware assets. Instead of calculating projections based on averages, ML techniques can better predict individual hardware retirement dates. And most IT asset and configuration management tools (or all-in-one ITSM platforms) already store the data required to train the models.
Here’s how it all works: ML algorithms examine the technical characteristics of your retired assets and group them logically. Then, they analyze how each asset’s maintenance and associated downtime costs changed over time and at what point they were retired. This helps to uncover common behavior patterns in each asset group. These patterns later form prediction models that identify pre-retirement asset behavior.
As soon as the models are tested, they turn to the not-yet-retired assets. When they see that an asset’s trend of maintenance and downtime costs matches a retirement model for its group, they predict further maintenance costs for this asset and recommend a proper retirement date. If a new asset doesn’t fall into any existing asset group, the models can find a technically similar group for it and use its historical data to predict the right retirement date.
This helps to cut costs by reducing the need of IT teams to support older assets that are past their prime.
Forecasting asset demand
Due to the uncertain nature of demand, managing asset inventory through traditional means, by setting stock rule thresholds, doesn’t always work in enterprise-class organizations. To approach the problem more systematically, IT managers can use machine learning techniques to make monthly demand predictions for each asset model.
To do that, machine-learning models must be trained on two types of data that factor most heavily into demand forecasting: data generated from HR processes and asset retirement data. On the HR side, that includes information on job openings, new hires, position changes, and terminations. With asset management, the models need a range of asset characteristics (including equipment models, types, and retirement dates), as well as stockroom data.
The training data helps the models spot tendencies and see how they are interrelated historically with asset demand. Then they produce demand forecasting models that continually “learn” from ongoing HR and asset retirement trends. If current stockroom data predicts you won’t cover demand in a given quarter, managers can use the tools to automatically create purchase orders to ship needed assets. (Read on to see how AI can manage overall procurement.)
With the ability to better predict asset demand, managers can slash downtime costs by improving asset provision, which in turn helps boost service quality.
Reforming software license distribution
As corporate spending on business software rises, so does the need for companies to better manage license distribution and vendor audits. Here, too, machine-learning tools can make a difference. They can more accurately recognize software usage patterns to optimize license allocation.
To make it happen, companies must compile the right training data sets. Those include the total list of apps, all user entitlements and roles, time spent in apps, features used, and other data points. Data engineers then help the models learn by identifying which software usage pattern they deem optimal. For example, they might look at time spent in the app or the number of features, but also take into account possible exclusions from the rule: when people are on vacation, sick leave, etc.
The ML models can then segment users according to their roles in various applications to adjust optimal software usage to the specifics of each group. Then, after being fed new incoming data, the models can spot behavior deviations and make corresponding suggestions.
For example, if a particular app isn’t mission-critical and is rarely used, AI can recommend handing off the license to someone else. Or, if some users aren’t taking advantage of core features, AI can recommend swapping their license for a more appropriate one.
Ultimately, the tools can help reduce IT license costs by making more efficient use of software investments.
Along with predicting asset demand, machine learning tools can manage asset procurement with equal efficiency, adding another layer of cost savings. For hardware assets, ML tools organize purchased assets by class, model category and model.
They look at each group’s purchase-order details to identify orders that had the most favorable price and delivery costs. Then they search for dependencies in the data to understand which specific variables —vendor discounts, order amounts, or shipment locations— were most influential. Once those are determined, the models recommend the ideal purchasing specs for each asset group. They can also merge multiple purchase orders or recommend switching vendors when they see a cost-saving opportunity in new data.
Ultimately, these tools reduce hardware purchasing costs by enabling a smarter and more transparent buying process.
As demands and workloads for ITAM teams increase, one thing seems obvious: humans can’t do this job efficiently enough on their own. IT teams can leverage the power of machine learning and AI to manage the ever-increasing complexity of IT asset management.