The hype around AI is driving excitement in the business world, with many organizations eager to harness its potential. A 2023 study by OliverWyman Forum notes that 60% of firms have already seen some form of productivity improvement in their workspaces through the implementation of AI. In data science practices, implementing AI can enhance productivity, streamline decision-making, enable automation, and drive innovation. To effectively leverage these technological breakthroughs, stakeholders should consider three key criteria to set clear expectations regarding costs, benefits, risks, and responsibilities when integrating AI into their data science workflows.
Economy: Would it be worth your time and effort?
New AI capabilities require a significant upfront investment of time and effort from data science teams. These costs include the necessary talent, development efforts, and infrastructure to implement and operate these capabilities. Ongoing upkeep, such as monitoring, maintenance, and future enhancements, is also essential. If the overall cost is not feasible, consider a more cost-efficient alternative or focus on areas where productivity gains can be achieved with minimal effort. Leveraging AI tools to handle non-trivial or routine tasks is an excellent starting point. More frequently, a simpler solution can be the solution. Predictive forecasting using linear regression can be just as accurate and beneficial as deep learning models. It’s important to step back and tactically assess your organization’s needs.
Effectiveness: Would it deliver the promised value?
Be cautious of ‘cookie-cutter’ services that may not address your specific needs. Many pre-built solutions can miss the mark, and it can be difficult to interpret their insights and outputs without expert guidance. For instance, a lot of large language models (LLMs) like ChatGPT predict the next word in a sequence, rather than reasoning through why they are generating certain responses or what they are aiming to accomplish (See a theory On Bullsh*t). Even the newer developments like OpenAI’s GPT-o1 may reason yet is not perfect and comes with a steep cost. Consult with data science practitioners to ensure that the solution you intend to implement effectively addresses your unique challenges and questions.
Efficacy: Would it be responsible and stable?
The progression of AI now aims to perform tasks on par with human capabilities. This also means that there should be growing emphasis on responsibility and stability when leveraging the technology.
Ensuring responsibility involves protecting consumer privacy and company data and knowledge (and not violating data privacy and copyright).
Ensuring stability helps the AI-enabled services to provide users with reliable service over time. Setting clear expectations and goals is essential to avoid investing in costly capabilities that may be underutilized and short-lived.
If your intention is to enhance or replace the human workforce, you must also inherit the same level of responsibilities and potential liabilities. Remember that AI is a tool, not a comprehensive solution.
A value-added capability comes from crawl-walk-run steps
In our work with clients, we find the greatest value in first establishing robust data infrastructures and developing a roadmap for data science practices (crawl stage). Through a series of analyses, we deliver actionable insights and recommendations to address their immediate needs (walk stage). This step also helps us evaluate the most efficient AI tools to address clients’ long-term needs. From there, we design and build impactful AI-driven solutions (run stage). One such case involves a web application that generates insights on-demand from predictive ML/AI models for clients looking to optimize their physical retail presence.
Partner with Us
We are committed to guide your AI journey to deliver value and efficiency at every stage. From foundational planning to advanced solutions, we help ensure your data science initiatives drive measurable success.
