Data: The Forgotten Ingredient to Automation and AI Success
In this article
Often when we talk to clients about automation, AI quickly enters the conversation. This doesn't come as a surprise based on how automation and AI are being positioned in the market.
Many automation companies have marketed AI as part of their solution portfolio. Likewise, AI companies have promoted a bevy of ways AI can enable automation.
Unfortunately, the key ingredient to the success of both gets lost in the shuffle. Good automation and AI start with an organization's data.
While there are many AI point solutions, an enterprise-wide AI capability — think bespoke large language models — is simply an evolution of what an organization can do with its data. In turn, this can lead to more sophisticated automations.
To better understand the interplay between automation and AI, we've found it helpful to discuss the two in the context of three general categories of data use cases:
- Visualization and business insights
- Predictions and recommendations
- Triggering actions
Let's explore each in turn.
Data visualization is the representation of data in graphical or visual formats, such as charts, graphs, and maps. Business insights refer to the understanding that results from data analysis, which is used to inform decision-making.
The two are powerful tools for identifying what to automate.
For example, a financial institution that we work with was having trouble tracking the health of its IT assets. We delivered the client a data visualization trends and insights dashboard that tracks infrastructure utilization trends, performance and hardware. With asset health visualized, the client could identify areas ripe for automation — for instance the process of decommissioning devices.
Although the financial institution's data was cleansed and combined into a single source of truth, staff must still recommend what to automate.
In our work with another financial institution, we took similar data related to asset health, curated it into a brokered data set and performed predictive modeling using neural networks — an example of quantitative AI.
The predictive model automatically alerts IT staff to correlations and causations that may lead to problems but that would be difficult to spot manually. Staff can then use this automated, AI-powered information to proactively pinpoint root causes of potential failures versus reacting to them when incidents occur.
In this example, we have automated away the tedious effort of correlating data points manually, but we also benefit from the AI system continuously learning from each incident and resolution, improving accuracy and efficiency over time.
AI-generated information recommends automations that can head off repeat incidents. However, these automations still require manual intervention.
The idea of self-healing often comes up in discussions about automation and AI. In the previous example of the financial institution that leveraged automated predictions and recommendations, we added a self-healing layer that triggers actions.
For example, staff can program the model to trigger a reset if a server's cache is nearing a threshold that is predicted to cause an incident.
Some of our clients are concerned by the idea of AI executing automations without manual intervention. They worry about a future where people are removed from operations completely. But make no mistake, people are still very necessary.
Staff are still part of the automation loop, examining triggered actions to understand their downstream impact on IT systems and processes. Further, staff are needed to determine if an automation can benefit another part of the organization, or if an automation no longer makes sense based on changing business processes or team structures.
Yes, AI accelerates our ability to automate everything, but only staff with a deep understanding of end-to-end business processes and an organization's long-term goals can determine how to automate everything right.
As automation and AI continue to make waves in the market, we advise clients to take a measured approach.
Organizations tend to think of data as one asset and AI as another. This thinking can lead organizations down a path of onboarding a series of point solutions. Instead, remember that AI is simply a maturation of your organization's data practices.
By mapping AI and automation to your data use cases, you can develop scalable automations that can be tailored across your enterprise because they are born from your organization's data.