Pitfalls of Rushing AI Implementations
April 13, 2024 | Written by Sree Ravela
Image generated by AI
It is real. AI lives with each of us today, be it in our cars, living rooms, bedrooms, and offices in the form of Siri, Alexa, Google Assistant, or even tied into social media apps like Instagram and Facebook (Meta). We regularly find its application on websites such as chatbots or Robo-Advisors. Now, we must choose how we want to be part of this new history.
In the race to weave AI into the tapestry of our everyday lives, we seldom pause to ask: What is the actual cost of speed? When AI solutions are rushed to market for innovation, are there systems that may falter in critical moments, compromise our data, or echo unchecked biases? For better or for worse, as AI becomes an indispensable ally, we’re bound to wonder whether the speed of technological ambitions outpaces the foresight required, sowing seeds of future dangers we are yet to understand completely.
Organizations often feel undue pressure to adopt the latest technologies to stay competitive in this fast-paced digital economy. The fear of falling behind can drive organizations to rush AI solutions, believing that speed captures the market share with innovative services. While this urgency is justified in the context of rapid technological evolution and competitive dynamics, it’s crucial to balance the speed of adoption with careful planning and execution.
This article is intended to highlight key considerations using real-life obstacles encountered during high-profile AI deployments and challenges that companies may encounter. I want to help you ensure AI’s long-term sustainability and continued success within your organization. So consider these pitfalls before rushing into AI implementations.
Unclear Objectives
AI implementation without clear objectives is like setting sail without a destination. Identify and outline clear objectives, whether the solution(s) improve customer service and employee efficiency, enhance security, and/or increase revenues. Without a clear plan, AI initiatives can lead nowhere, consuming available resources and failing projects, causing reputational damage.
Lack of Data Preparation
The “garbage in, garbage out” concept also works for AI. If you rush through the data preparation stages, AI decisions will resemble a guess at best, but most likely, a disaster. A poorly collected and formatted database can lead to many errors, duplicates, and inefficient decisions that require extra human resources to confirm. For example, if the customer contact information in a subscription system or e-commerce transaction is not validated thoroughly, many duplicate records and typos creep into customer data, resulting in multiple cold calls annoying potential customers. Although data preparation is tedious, it is critical so please approach this process carefully.
Ignoring Security, Scalability, and Sustainability
AI initiatives should not be viewed as standalone projects but instead as long-term efforts with long-term consequences. Investing in AI without considering its scalability and sustainability can lead to solutions that are too challenging to maintain, optimize, or scale when organizations need change. In 2016, Microsoft’s chatbot, Tay, was released on Twitter for users to interact with. The idea was to demonstrate and scale up conversational understanding. However, within hours of its release, Tay began to unintentionally send offensive and harmful tweets due to a coordinated attack by a subset of people exploiting a vulnerability in Tay, feeding it inflammatory language. Tay was taken offline within 24 hours due to the controversy, and Microsoft published the lessons learned.
Overlooking Ethical Considerations
AI technologies influence our day-to-day lives, societies, and nations. Ignoring ethical considerations when implementing AI strategies will likely fail to achieve desired outcomes and cause costly, public backlash. From an algorithm’s bias to privacy concerns, several ethical considerations in machine learning should be reviewed, evaluated, and integrated into the solution before it leaves the lab and becomes operational.
In 2018, Amazon decided to stop using an AI-based recruitment tool trained on resumes from the last decade. It turned out that the tool was only copying biases in data from resumes it had received, with the differences particularly impacting women. This case raised a discussion about whether or not AI has valid use cases in Human Resource (HR) management. It highlighted a desperate need for excellent and ethical AI practices to avoid societal damage, public mistrust, and/or legal complications.
Neglecting Change Management
The introduction of AI comes with changes, including within organizations of new roles or processes. It should be embedded into new workflows, requiring careful planning, identification, and communication of these new ways.
Have you heard about IBM Watson’s implementation for Oncology? This high-profile AI use case in the healthcare industry failed due to unrealistic expectations, integration within current systems, training, ongoing support, data privacy, and attention to data quality. Careful planning and execution of Change Management could have avoided this expensive lesson and resulted in a more favorable outcome instead.
Falling for Hype Over Substance
AI is still evolving but is often overhyped in media, making unrealistic promises about its capabilities and what it can achieve. However, not all AI technologies are the same, and not all problems have an AI solution. Aspirational implementations based on unrealistic expectations often lead to disappointment and frustration with AI projects when intended outcomes are not achieved.
For example, the Ed-tech company Knewton has launched an AI platform that claimed to be exclusively implemented using big data and sophisticated algorithms to provide personalized learning. The product was highly advertised, but this led to problems because it was not built on a mature enough pedagogical model—it wasn’t designed for data integration with existing solutions. The hype contributed to a rush-to-market without adequate development.
AI’s promises cannot be neglected, but organizations must clearly state their objectives, prepare data thoroughly, implement Change Management best practices, and consider ethical matters instead of rushing these implementations. Will our legacy be defined by the AI we hurriedly deploy or by our thoughtful stewardship exercised over its potential? The power to shape an AI-enhanced future is in our hands, but will we choose the path of meticulous craftsmanship or the perilous shortcut of haste? Today’s decisions may very well become the benchmarks for tomorrow’s digital society.