Moving Forward with AI
With the launch of ChatGPT, firms find themselves having to respond to rating agencies, investors and employees about their artificial intelligence (AI) strategy. Given the enormous investments going into it, AI continues to develop at a break-neck pace. Many firms have scrambled to get started simply for the sake of not falling behind. But the distance from use cases to scaling to delivering tangible results is not a short one. From my conversations with executives and experts, many agree on these key hurdles to overcome:
Knowledge/Communication Gaps
Clear Business Purpose
Data Challenges
AI Risks and Governance
Human Factors: Fear, Trust, Workforce Impact
What can firms do to overcome these hurdles and move forward with AI adoption?
1. Business and Technology Sharing the Same AI language Is the Critical First Step
Knowledge and communication gaps between business and technology are common in transformation projects, but they are especially pronounced with AI due to its complexity and revolutionary approach to computing.
Traditional computing relies on programmers writing step-by-step code to process input and generate output, strictly following programmed instructions. Any improvements require manual updates by programmers. In contrast, AI computing uses algorithms to train “AI models” on vast datasets through multiple iterations. These “trained” AI models then infer new dataset to produce outputs. AI models can self-refine, automatically in most cases, with more data and improved algorithms, becoming more accurate and efficient over time.
What do these differences all mean? How are AI models trained and improved? What types of AI models are there other than Gen-AI? How do they work?
This hurdle can be relatively easy to overcome through training and open dialogue. Creating a common AI language and a shared understanding of AI technologies, and the organization’s unique infrastructure and capabilities that enable or constrain AI adoption, is the critical first step in harnessing AI’s next-gen computing capabilities to remain competitive and relevant.
2. Business Should Drive AI Initiatives, Not Technology
As you probably start to see now, AI gets technical very quickly, often steering organizations to let IT drive AI initiatives. That would be a mistake. Without a business purpose, IT might pilot AI projects for the sake of technology. Large organizations with ample resources are particularly prone to this "AI trap." However, the real competitive advantage will come from the organization’s ability to truly understand AI’s potentials and limitations and integrate AI as powerful tools within their overarching business strategy and operating models.
Use cases for applying AI to raise productivity, detect fraud, and enhance customer experience thereby producing new growth have been widely established. What’s critical is for business and technology to come together to identify areas where AI technologies can generate the biggest impact. That could mean integrating AI in their operating models. Often, an end-to-end redesign of the process, while more difficult to execute, can produce bigger, more sustainable results at reduced costs in the long run. These are important business decisions.
Clarity on the business purpose for leveraging AI will in turn guide the best approach to acquiring the technology. Given the resource intensity of building and maintaining AI models, firms should carefully strike a balance between building or buying solutions using AI-as-a-Service. Resourceful firms can also rely on strategic partnerships to gain access to the latest technology and expertise to achieve some customization without substantial upfront investment while at the same time upskilling employees. The optimal approach will be an adaptive one that aligns with long-term strategic goals considering commercial viability, capabilities, and customer experience, which is increasingly a competitive differentiator.
What would be a good structure for deploying AI? According to MIT Senior Lecturer and Digital Capability Leader Paul McDonagh-Smith, firms that excel in creating value from AI often adopt what he calls “dual-tracks.” They have a short-term pilot track for business-driven proof of concepts projects of less than six months. Insights from this pilot track then help inform a comprehensive, firm-wide AI strategy looking ahead up to two years. This dual-track approach focuses investments and resources on AI initiatives that produce tangible returns while adapting to AI’s exponential advances and the rapidly changing market conditions.
3. Data Is Key for AI Adoption Success
Remember AI models are trained by data? “Data is the blood of AI infrastructure,” to quote Jae Kang, a risk and data executive, “It’s important to constantly remind ourselves that AI is only as effective as the data it is trained on.”
More specifically, the quality of the training datasets directly impacts the accuracy of predictions and decisions made by the AI model. The availability and quality of the real data in production impacts the reliability of the output which in turn impacts user trust and confidence, which is essential for wider user adoption.
Jamie Cattell, global managing partner of IBM and an early adopter of AI technology, likens data to the oil that powers the “jet engine” of businesses using AI “in the same way that raw data in the enterprise often requires refining to power AI”. He observes that many organizations can execute small AI pilots with limited datasets but struggle to scale due to data challenges.
Financial service firms, being data-intensive, have been trying for years to turn data into assets. Progress has been slow – it’s hard work with benefits not immediately visible. AI now empowers data executives to accelerate good data management practices. Recognizing the critical role data plays in realizing the potential of AI technologies, a Fortune 100 global firm has created a new “Chief Data Administration Officer” role.
Strong data capabilities translate directly to cost-effective adoption of AI technology in the long run. For example, the more an organization can integrate different types of data (text, images, videos; structured and unstructured) from various internal and external sources and can provide timely, reliable access to data streams, the more it can be creative with AI technologies to produce substantial value. All these are no easy tasks for organizations with siloed systems and limited resources and expertise. In such cases, data solution provider like illumex may be worth investigating.
Finally, sensitive or personal data will need to be protected to meet privacy and security mandates by regulators, customers, and the public at every stage of AI deployment. This typically challenges existing risk management and governance frameworks, the next key hurdle.
4. Strong AI Governance Is the Competitive Advantage
The popularity of Gen-AI has magnified the risks and challenges surrounding AI. Firms quickly find their existing enterprise risk management and governance frameworks inadequate. For example, protecting privacy and information security now also requires ensuring fairness, transparency and explainability of AI models, as increasingly demanded by regulators and the public.
AI risks extend beyond regulatory compliance. In his article, How to de-risk your AI strategy?, Simon Torrance, founder of AI Risk, discusses enterprise AI risks in four areas: Strategic Risk (losing customers, market share) as well as the conventional Financial Risk (investment vs. return), Operational Risk (new forms of cyber-attack, talent challenges) and Compliance Risk (standards, regulations). More than two-thirds of CEOs in his survey believe the productivity gains from AI are so significant that they are prepared to accept more risk to gain an advantage.
To address these AI risks and ensure responsible AI practices, firms need to proactively incorporate AI governance in their policies, processes, and controls. In response to these increasing needs, forward-thinking AI solution providers like Mind Foundry and Lumenova AI have built model monitoring, explainability, bias detection and auditability capabilities into their respective, yet unique, products. Firms that get strong AI governance firmly in place will have a competitive advantage in the AI race.
5. AI-Ready Workforce Through Training and Open Communication
AI is essentially a continuous improvement process, affording tremendous advantage to early adopters willing to experiment and fine tune throughout the entire organization. However, as with most transformation projects, the people aspect is harder to get right than technology and process. Fear of being replaced by AI, anxiety about the unknown consequences of deploying AI, and its potential to turbocharge harmful abuse are the elephant in the room that needs to be addressed head-on with trust.
Equally challenging is the cultivating of a forward-thinking culture of innovation and flexibility through the training and upskilling of employees. As the pace of AI adoption accelerates, so does the need for an evolving training program to support the increasingly influential interplay between AI and humans.
To build trust in AI within an organization, transparency and sustained, intentional engagement is key. Management should double down on communication. Openly talk about AI strategy and lessons learned. Demonstrate a commitment to using AI as tools to strengthen their workforce to compete in the new economy rather than eliminate jobs. Encourage collaboration across functions and disciplines in developing AI solutions. Incentivize ideas unique to their organization where humans and AI leverage their respective strengths to solve problems and create value that not long ago were thought impossible. Firms that develop an AI-ready workforce will do best.
Conclusion
Don’t let compliance and governance concerns, ignorance, fear of change, or complacence get in the way of moving forward, John Teddy of Lazarus AI says with the graph above.
Rapidly advancing AI capabilities and the increasing integration into business practices present significant challenges and opportunities. It’s crucial for management to take measured and swift action to address knowledge gaps, define clear business purposes, overcome data challenges, effectively manage AI risks, and build trust with the workforce as well as with regulators, clients, investors, and the public. By tackling these hurdles head-on, organizations can harness AI's full potential and secure a competitive edge.
How to get started? Have you done the following?
Create a common AI knowledge base and taxonomy bridging business and technology
Understand the AI-readiness unique to your tech stack and data capabilities
Establish an AI taskforce with a clear charter
Integrate AI adoption in your overarching business strategy
Determine AI risks and governance framework
Cultivate internal and external AI partners
Identify early-adoption areas with the largest impact
Communicate your AI strategy, especially with employees, openly and continuously
Build an AI deployment structure that produces tangible results
Maintain open dialogue with regulators and remain transparent where AI is concerned
Has this article been useful as you think through and move forward with AI adoption? Could you use some help navigating these complexities? Get in touch.
* Ichun Lai founded Propel Global Advisory focusing on accelerating the thoughtful and responsible adoption of AI technology in financial services