Can artificial intelligence (AI) help companies improve efficiency? The simple answer is yes. However, incorporating AI into business processes and workflows is not a simple process, though it is both an achievable and in many cases, a necessary one. As a leader of AI teams at Intuit®, I have learned a few lessons about the process, and I’ll share them with you today as we look at AI’s ability to improve efficiency in businesses in every industry.
How artificial intelligence improves efficiency
First, what is AI? Brian Gorbett has a great definition. He writes in Demystifying artificial intelligence and machine learning that AI is “taking data, learning from it, and redeploying outputs that help your customers.”
According to IDC’s 2019 Spending Guide, AI system spending will reach $97.9 billion in 2023. Why? Because businesses are finding that AI technology provides a myriad of benefits for their customers as well as to the business itself, including improving their efficiency by performing data-driven tasks faster and better than humans.
This doesn’t mean the human element is extinguished by AI; in fact, the human element is enhanced and supported it. For example, AI can help accelerate customer support processes by generating automated case note summaries to help agents catch up with previous calls content. It can also help coach agents on providing better support for their customers by leveraging explainable AI tools.
In addition, organizations can reform their products and data security by leveraging AI to detect anomalous behaviors in their systems.
One of the main benefits of leveraging AI for such tasks is the ability to automatically learn and update the models based on changing patterns in the data. With traditional business rules, human interaction is required to modify the logic in order to address changes over time.
Without a doubt, AI is becoming a necessity for businesses wanting to improve their efficiency and remain competitive in a dynamic marketplace. However, it may be intimidating to those who are new to AI, so I have some advice.
How-to advice on using artificial intelligence to improve efficiency
As I mentioned earlier, I lead AI teams at Intuit. We have found that using AI to improve efficiency should first start with the gathering of efficiency problems, and then ranking them by impact.
Once you’ve done that, try to understand whether a simple rule-based solution based on domain expert heuristics could solve this problem. If you find there’s still room for improvement, then you should pair the domain and data experts with an AI expert to see if relevant labeled data could be gathered to solve the problem using AI.
Note that AI is not always the best solution. Sometimes, there is just not enough relevant data to generate an efficient AI solution, and sometimes simple rule-based logic would be enough.
It is important to understand that developing AI models should always start with a problem and a hypothesis that a solution can provide a certain benefit. It is recommended to test the hypothesis with a simple solution first, and then go on with researching AI techniques to solve the problem. This process exemplifies Intuit’s Design for Delight.
Adopting AI into your process and workflows does pose some challenges, including prioritizing the integration by the product developer (PD) teams, which are needed to get AI integrated into existing products along with other business initiatives. Working closely with the PD and project manager (PM) as a mission-based team during the model development process, explaining to them how the AI works, and showing the potential business impact of the service. will help to build trust with the PD teams and accelerate the integration.
I would also recommend leadership invest in AI education for the PD and PM communities. Education, combined with specific goals and metrics around AI adoption, can really help the teams communicate and work better together. For more information on AI, check out How artificial intelligence is redefining apps and Forecasting and predictive modeling for marketing analytics.