Overcoming Strategic and Management Challenges: The Hidden Costs of Embracing Transparent AI
This blog post was created with the valuable insights and contributions from our esteemed collaborators. Special thanks to Fernando Mourao, Tao Zhang, Sue Ann Chen, and Saumya Pandey for their expertise and support in developing content around AI transparency and legacy system challenges.
In the rapidly evolving landscape of AI, the imperative for transparency is reshaping how companies approach their strategic and management decisions regarding AI. However, this journey is frequently hindered by legacy AI systems and the challenges posed by the hidden costs of paradigm shifts to more transparent AI systems.
In our previous blog post, we debunked the misconception surrounding the perceived trade-off between AI transparency and business interests.
As decision-makers in the organization, it’s important to ensure alignment between the efforts invested into AI transparency and the organizational objectives while effectively managing associated risks and costs.
In this article, we delve deeper into the second of the six main hurdles defined in our inaugural post — the strategic and management challenges linked with embracing transparent AI. We will explore the investments and strategic planning necessary to navigate these hurdles successfully and transition toward a more transparent AI framework.
Explore the Costs of the Legacy AI System
Legacy AI systems are ubiquitous in many organizations, constituting a significant portion of their operational landscape. According to a survey conducted by Forrester Research across 3,700 companies, an estimated 72% of budgets were allocated to maintaining and patching these legacy software systems [1]. Beyond the financial burden, legacy AI systems present a myriad of challenges that impede innovation, compromise service quality, slow delivery speed, diminish customer satisfaction, and hinder enterprise readiness for adopting new technologies.
Updating or replacing legacy AI systems with newer and more transparent AI solutions poses substantial financial and logistical challenges. However, the opacity surrounding these challenges makes it difficult for decision-makers to prioritize the update or retirement of legacy systems effectively. Outlined below are some of the key challenges decision-makers must consider when contemplating the update or replacement of legacy systems:
- Integration Complexity: Transitioning legacy systems to accommodate transparent AI functionalities requires significant investment in system integration, data migration, and interoperability.
- Technical Debt: Legacy systems often carry accumulated technical debt, resulting in complexity, increased maintenance costs, and compatibility issues with modern AI technologies. Addressing technical debt, such as code refactoring, atop other architectural and technological changes, is time-consuming and prone to errors.
- Organizational Resistance to Change: Employees and stakeholders entrenched in legacy systems may resist or feel apprehensive about changes to established workflows, business processes, or technologies. The longer the legacy system exists within the business, the deeper the entrenchment becomes. Overcoming this resistance demands effective communication, training, and engagement strategies to foster buy-in and support for the update process.
- Skill and Knowledge Gaps: The specialized skills and expertise required to support and update legacy systems may be lacking within the organization. Hiring personnel with specialized knowledge in legacy systems may be necessary to design robust solutions for transitioning to modern technologies, further adding to the overall cost and complexity of the update process.
The challenges above can significantly increase the human cost and ultimately the financial cost associated with the implementation of any future innovation or change programs.
Explore the Hidden Cost of Paradigm Shift
Shifting towards transparent AI involves rethinking the entire design and execution process. In addition to the cost associated with the legacy systems, some costs are often overlooked as part of this transformation process.
Types of Common Hidden Costs:
- Data Governance: The costs to ensure data integrity, privacy, and accessibility demands are met under robust data governance frameworks, necessitating investments in data management technologies and talents.
- Transparency Infrastructure: Implementing transparent AI solutions often requires the development of infrastructure for explainability and interpretability. For example, the cost of updating AI products’ user interface, the time investment in unpacking the model behaviour, and the building of the dashboards or visualizations designed to assist human interpretation of the model predictions.
- Culture and People: Resources and training are required to equip the people with the knowledge around AI transparency e.g., core principles, best practices, and guidelines for ethical AI development and deployment. Shifting the company’s culture towards transparency also requires investments in change management initiatives and leadership alignment.
- Regulatory Compliance: Adapting to a framework of AI transparency creates the need to invest in legal consultations, compliance audits, and adjustments to existing processes and governance systems.
- Business Integration: The shift towards transparency and explainability also requires corresponding changes in business processes. These include more strategic changes such as updated business value or company policies to more operational changes, such as redefining success metrics to embed new constraints like transparency.
According to “Study to Support an Impact Assessment of Regulatory Requirements for Artificial Intelligence in Europe” [3], the average estimated cost to ensure the AI solution is compliant with the AI regulation in Europe is about EUR 10,977, and if external data and services or additional staff are required, the cost may rise to EUR 29,277. This represents about 17.22% of the value of a reference AI solution.
Therefore, to prepare for the budget required to support this transformation process, leaders should consider costs such as project management and communication support, training, consulting, and technology and set aside some contingent funds that can cover the hidden or unexpected costs that might occur during the transformation process.
Preparing for the Shift to a More Transparent AI Solution
How can companies prepare for the shift towards AI transparency and strategically plan for long-term gains despite short-term costs? Organizations need to understand what is required to achieve transparency, identify current gaps in meeting those requirements, and assess the corresponding costs to bridge those gaps.
Here are some simple steps to guide your organization on this transformative journey:
- Clarify Success Criteria: Define how AI transparency aligns with your business goals to establish a clear and measurable vision of success.
- Adopt a Top-Down Approach: Ensure leadership commitment and involvement in driving transparency initiatives throughout the organization. This is especially important due to the misbelief that might exist around the trade-off between AI transparency and business interests mentioned in the previous blog post.
- Evolve Governance Systems: Update governance frameworks to incorporate transparency principles and ensure accountability at all levels.
- Empower the Workforce: Equip employees with the knowledge and tools necessary to support transparent AI initiatives, fostering a culture of transparency and accountability. There are quite a few online courses available to assist with training, an example is an online program offered by the Tech Stewardship Practice Program (TSPP) [2]
- Prioritise Impactful Transformations: Focus on implementing changes that have the highest potential to enhance transparency and deliver significant business value.
- Monitor Progress: During the transformation, it’s important to set up a regular cadence to review the progress on the key metrics associated with the success criteria defined above, and identify any roadblock or dependency that might delay the progress.
The following section provides practical tips to assist senior stakeholders in prioritizing the impactful changes effectively:
- Conduct Comprehensive Cost Analysis: Evaluate current AI systems and processes to identify areas needing transparency enhancements and estimate associated costs accurately. A good example of how to conduct the cost-benefit analysis can be found here.
- Assess Long-Term ROI: Recognise the enduring benefits of transparent AI, such as increased trust, reduced legal risks, and enhanced stakeholder satisfaction, to justify initial investments.
- Optimise Resource Allocation: Allocate resources strategically, prioritizing investments based on their criticality, feasibility, and potential impact on transparency goals.
The following section provides information and resources that would be helpful to empower the technical workforce to navigate through the challenges that arise as part of the transformation:
- Explore External Partnerships: Seek opportunities for collaboration with external experts and vendors to access specialized skills and cost-effective solutions for transparent AI implementation
- Resources on how to provide transparency and explainability for a complex black-box AI solution
- Resources on how to provide transparency and explainability to end users without disclosing the trade secrets
- Examples of what success looks like for some organizations pursuing AI transparency
- Resources on the existing tool to help with AI transparency.
To wrap up, while the road to AI transparency may seem daunting, the potential rewards far exceed the initial challenges. By prioritizing transparency in AI implementation, companies can build trust, foster accountability, and position themselves as ethical leaders in their fields. Moreover, the long-term benefits, including happier customers, reduced legal risks, and a competitive edge, make the journey toward transparent AI not only worthwhile but essential for sustainable growth and success.
As we prepare to dive deeper into the intricacies of AI transparency in our next posts, it’s crucial to acknowledge the pivotal role of bridging conceptualization and objective specification, as well as the challenge of verifiability. These aspects are fundamental in ensuring that transparency efforts are not just aspirational but also actionable and measurable.
Reference:
[1] What are Legacy Systems and Why Do Companies Still Use Them?
