Machine learning: 4 adoption challenges and how to beat them

As you strategize your AI/ML initiative, consider these four common barriers and how to overcome them
2 readers like this.
machine learning explained

In the first quarter of 2022, global funding to artificial intelligence (AI) startups reached $15.1 billion, according to CB Insights’ State of AI report. However, machine learning (ML) algorithms can lead to counterproductive results when deployed without reason.

Here are four common challenges that companies implementing ML-based systems may encounter, along with some expert tips to maximize the impact of algorithms while avoiding missteps.

1. Finding an ML use case

For some companies, the first issues with AI and ML adoption come before starting. Machine learning is a vast, multifaceted discipline pervading most aspects of artificial intelligence. It paves the way for numerous potential applications, from intelligent process automation (IPA) and natural language processing (NLP) to computer vision and advanced data analytics.

Selecting a use case worth investing in is easier said than done. In this regard, O’Reilly’s 2020 AI adoption in the enterprise study ranked use case identification second among the most relevant challenges (mentioned by 20% of respondents).

[ Also read The AI revolution: 4 tips to stay competitive. ]

Beyond the usual recommendations on framing your corporate goals – i.e., what you expect machine learning to do for your business (enhancing operational efficiency, improving your products or services, mitigating risk) – a rule of thumb for choosing a suitable ML use case is “follow the money.”

Target the most strategic business functions and generate the maximum profit for your organization, depending on its size and industry. Examples might include computer vision-guided assembly for manufacturers or data analytics-driven marketing for retailers.

Another selection criterion focuses on addressing your corporate weaknesses, such as process bottlenecks. You can identify them by proper BPM investigations and KPI assessments.

2. Selecting the right data

Data is the fuel of machine learning. ML systems need to process enormous data sets to be adequately trained. The reliability of output depends on the quality of the data sets and the training process itself. Here are some recommendations to consider:

  • Rely on qualified data scientists to select suitable data sources, be they external or collected from corporate systems. Set up effective data management and governance strategies to ensure that data is harvested and stored correctly.
  • Select a subset of core features from your datasets so the training phase can focus on the most relevant variables and ignore redundant metrics, facilitating result interpretation.
  • Train your ML system with multiple subsequent data samples (typically called training, validation, and test sets) to monitor and enhance its performance in different conditions while avoiding overfitting issues, namely when algorithms are “tuned” on specific data sets but perform poorly with others.

3. Complementing ML with human talent

Machine learning algorithms may still behave unpredictably after training to prepare for data analysis.

This lack of clarity might be an issue when leveraging AI in decision-making leads to unexpected outcomes. As the Harvard Business School reported in its 2021 Hidden Workers: Untapped Talent report, ML-based automated hiring software rejected many applicants due to overly rigid selection criteria.

ML-based analysis should always be complemented with ongoing human supervision.

That’s why ML-based analysis should always be complemented with ongoing human supervision. Talented experts should monitor your ML system’s operation on the ground and fine-tune its parameters with additional training datasets that cover emerging trends or scenarios. Decision-making should be ML-driven, not ML-imposed. The system's recommendation must be carefully assessed and not accepted at face value.

[ Read also AI ethics: 5 key pillars ]

Unfortunately, combining algorithms and human expertise remains challenging due to the lack of ML professionals in the job market. The extent of the skill shortage is worrying for decision-makers around the world. Investments in staff training and partnerships with other organizations interested in adopting machine learning can help address this issue.

4. Managing resistance to change

Corporate inertia, resistance to change, and lack of preparedness could be the worst enemy of ML adoption. According to O’Reilly’s study, as mentioned above, corporate culture represents the main barrier to implementing AI-related technologies. It typically involves top management being unwilling to take investment risks and employees’ fear of job disruptions. To ensure stakeholder and staff buy-in, consider implementing the following best practices:

  • Instead of betting on moonshots, start from small-scale ML use cases that require reasonable investments to achieve quick wins and entice executives.
  • Foster innovation and digital literacy via corporate training, workshops, benefits, and other incentives.
  • Establish centers of excellence to supervise ML implementation across your organization, including operational and technological changes required to integrate these tools into your corporate workflow and software ecosystem.

Flying high without getting burned

Machine learning can take businesses to new heights through NLP-based interactive solutions, business intelligence software, and process automation tools. However, adopting this powerful technology within a robust management framework will save companies from numerous challenges down the road.

[ Check out our primer on 10 key artificial intelligence terms for IT and business leaders: Cheat sheet: AI glossary. ]

Andrea Di Stefano is a Technology Research Analyst at Itransition, a Denver-based software development company. He investigates emerging tech trends and their most impactful business applications, focusing on AI, machine learning, analytics, and big data.