When you buy through affiliate links in our content, we may earn a commission at no extra cost to you. Learn 
Every once in a while, a technology comes along that shows such great promise that businesses deploy it without giving much thought to how or where it can be most useful.
When we look back at the cloud, virtualization or the PC itself, fear of missing out (FOMO) often overruled careful consideration, and many enterprises ended up squandering substantial time and money undoing deployments that were either unproductive or detrimental to their operating models.
We can see the same thing today with artificial intelligence (AI). The current narrative is that AI will remake the entire world, and any organization that is not at the forefront of this revolution will wind up in the dustbin of history. It doesn’t matter what it’s used for or whether it gets good results, as long as executive management can report its existence to investors right now, all the rest will supposedly fall into place.
Planning for Artificial Intelligence
While AI is not necessarily doomed to fail, of course, it can lead to complications down the road if it is not implemented properly. Once AI takes hold of a certain process, it is hard to undo it. This is why a little planning is in order if the goal is to use AI as a valuable tool and not just technological window-dressing.
At the moment, the call center is one place where AI is proving to be highly beneficial. Its capabilities in functions like speech analysis and determining customer intent make it highly valuable as a sales and customer support asset.
Cobus Greyling, chief evangelist at data productivity platform developer HumanFirst, notes that AI can contribute to all four elements of a modern call center environment: connection, orchestration, resource management, and knowledge and insights. But some of its specific applications are easier to implement and provide greater business value than others.
Analyzing speech patterns and learning to gauge what the customer needs are highly feasible and offer good returns, while things like fully conversational self-service assistance and real-time agent coaching offer medium value and are more difficult to develop.
Building knowledge graphs to optimize conversational capabilities or creating the tools to enable intelligent contact routing are challenging projects to undertake at the moment, and they offer limited productivity.
Analytics with a Purpose
The field of business analytics also has a wide-ranging set of operations, some of which are more amenable to intelligent automation than others. Ivy Liu, CEO of ecommerce consultancy Ivy Insights, notes that lead scoring can benefit tremendously from faster, more accurate analysis of performance metrics, which in turn allows companies to rework or abandon underperforming initiatives while doubling down on the high-performers.
In today’s fast-paced digital economy, where margins are becoming increasingly tighter, this is likely to become a key differentiator between successful enterprises and failures.
AI essentially provides the tools to support real-time performance monitoring to deliver accurate pictures of what is happening now and in the future – and they can be applied across a range of processes, such as sales, marketing, finance and both mid- and long-term strategic development.
We can also take a look at the still-emerging field of DevOps to see how AI can be leveraged for maximum benefit. For starters, says tech writer Binod Anand, AI makes it easier to manage the input at each stage of the development process by gathering data from internal and external sources and analyzing it for accuracy, relevancy, and bias. It also enhances the effectiveness of the test cycle to weed out errors and boost overall productivity while also speeding up the execution of security checks.
Despite these advantages, however, over-reliance on AI can create unresolved errors in the DevOps lifecycle, which could lead to performance lags or outright outages. Equally troubling is the possibility of it producing unethical or disruptive consequences to people’s lives, particularly when applied to critical applications like healthcare, personal finance, and government services. Too much AI can also become quite costly, since it needs substantial compute and storage resources both for training and operations and to vet the huge volumes of data needed to drive appropriate outcomes. And once an AI model starts to rely on the output of other AI models, the risk of widespread disruption increases exponentially.
The Right Tool for the Job
According to Gartner, AI is most effective when applied to three general uses cases:
- Process Automation
- Personalization
- Enhanced Workforce Productivity
At the moment, most of these benefits come from applying AI to one-off, point-to-point solutions. High-scale solutions may create more value, but these can be difficult to achieve without making significant changes to established business processes and the interactions between teams within the organization.
One thing the enterprise should consider, particularly when it comes to generative AI like ChatGPT, is to keep abreast of emerging regulatory frameworks, particularly those related to copyright infringement and legal liability.
As well, all forms of AI can run afoul of the law by producing false outputs due to algorithmic instability, bad data, lack of human skills and training, and a host of other factors. The last thing any organization wants from its AI investment is a hefty fine, or even criminal charges.
The Bottom Line
Nobody ever said AI was a panacea for all that ails the enterprise (although some may have implied it). A proper strategic vision should target the low-hanging fruit first; that is, the easiest deployments with the highest return.
That at least gets the buy-in from the user community that the technology can contribute to their lives in meaningful ways. Once that step has been taken, further implementation will benefit from the in-house experience that has accrued to this point and perhaps the AI-driven insights gained from the first round of deployments.
As with all business initiatives, AI should start with a goal and a plan to get there. From there, it should work its way into the business model naturally, not by force.
