Advancing Generative AI entails substantial effort, but can have a transformative impact on business outcomes

Introduction

We've witnessed it time and again – a new "magic wand" under various guises (Outsourcing, Cloud, Agile, Big Data, Cognitive, Connected, Software Defined, and more), promising to resolve issues, boost revenue, elevate profits, enrich client experiences, and foster innovation.

Separating Hype from Reality

In truth, generative AI is no quick fix. It won't miraculously mend flawed processes, dismantle organizational barriers, or instill a culture of change and innovation. Achieving enterprise-wide transformation through Generative AI is a challenging endeavor, demanding a solid groundwork of Information Architecture, Digitalization, Business Process Optimization, Organizational Change Management, Security, Tools, and Governance. Attempting AI without these essential foundations risks relegating it to the realm of failed proof-of-concept projects (POC Hell)

And then, there is the issue of AI washing, with every firm touting the "Gen AI" offerings and AI investment dollars. Phil Fersht Lasse Rindom recently published an article which summarizes how AI washing has consequences beyond trends and its adverse impact on our generation. As rightly pointed out by the HFS Research team, AI has been around for around 50 years and "not all algorithms are AI –and that the generative AI we are currently enthusiastic about is still very much an algorithm". Generative AI differs from traditional AI in its ability to create "new/ original" content based on data it ingests and gets trained on. Be on the lookout for firms that package and position traditional AI capabilities (patterns, trends, analytics, visualization etc.) as Generative AI.

Indeed, it's important to clarify that not all AI endeavors are merely marketing tactics. AI holds immense potential for various industries when implemented correctly. It represents both a remarkable opportunity and a formidable challenge simultaneously. The key lies in adopting an inside-out approach, where you begin by pinpointing a specific business problem to solve, rather than relying on AI's superlatives to transform your company into an AI-enabled enterprise automatically

Making #GenerativeAI (GAI) work for you

Valuable lessons can be drawn from leaders such as Jason Wight, who spearheaded innovation at Ontario Power Generation through a focus on behavioral change. Jason advocates for a "broad stroke" approach to AI implementation, rooted in empowerment and accessibility for all employees, where AI becomes an integral part of daily work. This vision became reality with ChatOPG, an AI-powered digital assistant adept at addressing a wide range of queries, from equipment maintenance to technology

This advice is invaluable because many initiatives falter due to insufficient stakeholder support and limited enterprise adoption, and we certainly aim to prevent such setbacks with AI. To tackle this challenge effectively, it begins with a fundamental question: "What specific business challenges does this AI solution address?"

Avoid tackling this in isolation; instead, foster engagement throughout the organization, involving all stakeholders. Facilitate collaboration through design thinking workshops or create a portal to invite team contributions. To streamline the process, offer a use case summary for key functional areas, and be sure to include your Security team. (Sample use case summary for specific industries is provided in a later section.).

Getting started

Explore corporate use cases that are adaptable across organizations of any size and industry type. These encompass traditional functions such as IT Operations, Application Development, Marketing, Customer Service, Cybersecurity, and related areas. These versatile use cases serve as a foundation for future, industry-specific initiatives.

Developer teams may consider AI for enhancing code documentation, development, and refactoring, while they might not focus on integrating it for complex business logic and tasks. Conversely, a Business Strategist may place less emphasis on development and more on how AI can aid in identifying optimal product features and roadmaps, including user story curation and prioritization, to ensure future relevance. Furthermore, for Customer Service and IT Operations, the introduction of multi-modal Generative AI extends beyond AIOps, providing support through images and voice in addition to text and chat.

To secure the buy-in of cybersecurity and compliance teams, it's essential to approach them as partners, not obstacles. Collaborate with them to help them understand how AI can bring substantial value to the enterprise, particularly in enhancing security. This involves showcasing high-impact use cases in areas like Fraud Detection, Surveillance, Threat Detection, and SIEM (Security Information and Event Management).

Moving up the GAI value chain

As organizations mature in the AI journey, moving beyond corporate user cases, consider embedding GAI within your specific business processes and sustainability initiatives.

• Energy & Utilities: From mature use cases around demand and output forecasting to more advanced ones around Grid Management & Infrastructure Planning, GAI has the ability to significantly improve Operational Efficiency for the Energy sector. Energy industry is asset/ capital intensive and GAI can improve the Performance, Longevity, Security and Safety of these assets. GAI can work along side field technicians to detect, probe, resolve, track and document plant maintenance activities
• Manufacturing: GAI for Manufacturing should be approached from the perspective of how it can help address key challenges facing this industry, namely: Overall Equipment Effectiveness (OEE); Supply Chain Disruptions; declining revenues and an aging workforce. Start with ones which require minimal infrastructure investments and can realize immediate value such as Predictive Maintenance planning for factory floor equipment and Simulation and Training for the aging workforce. As the initiatives gain traction, consider the next step: self funded investments that can achieve ROI within the next 12-24 months: this includes use cases for Quality Control and Defect Detection, Supply Chain Optimization and Process Optimization. Finally, go for the more complex ones - Product Design, Prototyping and Customization and Personalization among others.
• Automotive: Software Defined Vehicles and Autonomous Mobility are well researched subjects, and hence not the focus on this article. If we exclude the two, Automotive has many similarities to Manufacturing and Energy and organizations may consider a similar approach for use case adoption and maturity development which is based on Time to Value and ROI. Start with use cases that can lay the foundation for the more complex ones. Add an additional dimension based on the 1% impact by quantifying how a 1% improvement in key processes would impact the business and map this to the expected improvement through GAI for the identified use cases. Expand this approach for each of the mature Automotive Use Cases, quantify what is the 1% impact for : Material Selection; Demand Forecasting; Warranty & Reliability; Toolpath Optimization; Predictive Maintenance; Supply Chain, Auto Parts Development, Plant Scheduling Optimization, Production Planning; Part Nesting etc. A word of caution: Rapid innovation in Automotive is also growing concern for Cybersecurity with many instances of data leaks from leading Automobile firms. GAI requires access to data and API from many players within and outside your firewall and hence a Zero Trust Architecture should be a primary consideration before pervasive GAI use cases are deployed.
• Sustainability: This is a tricky one. GAI advocates promote sustainability solutions leveraging GAI, however most GAI solutions require significant processing capability and hence high energy consumption to generate the expected outputs. For reference, ChatGPT consumes around 1300 Megawatt hours just for training and is definitely not environment friendly. Most known GAI solutions are delivered through Hyperscalers (AWS, GCP, Azure) and efforts are underway to reduce the power requirements per query. Mature use cases for GAI in sustainability initiatives include: Sustainable product design, ESG Roadmaps, Waste Reduction, Circular Economy and Smart Grids.
• The Finance industry has traditionally been big adopters and advocates of consumer grade AI solutions especially in Customer Experience, but with limited traction within the regulated, corporate side of the business. This is slowly changing as a result of financial pressures and increasing competition and a majority of clients are now accelerating adoption and identifying low risk areas to drive cost savings. Mature use cases include: Fraud Detection; Finance Advisory Services; Financial Forecasting; Claims Processing and Loan Underwriting.

Conclusion:

• AI/ GAI should be adopted as a business accelerator and not a "nice to have" technology competency. Even the best capabilities fail unless contextualized for a business need/ requirement. Always ask why use GAI before how and where.
• There isn't a plug and play GAI solution that will work for you from Day 1. No vendor has a magic AI wand.
• AI can be your "right hand man" or fall by the side depending on how you manage its adoption (and not just implementation). Identify champions within your organization and engage them early.
• Learn from your peers, friends, industry, forums. You don't need to take this journey alone. Engage and collaborate to drive success across industries.
• Either do it or don't, avoid POC hell. Very often POCs go nowhere and unfortunately a lot of GAI projects have not gone beyond pilots in many firms. AI/ Gen AI needs to learn from your data to deliver and a lab setup is not the right deployment.
• Clearly define and establish the risk profile and associated guardrails. We really don't want to create a pervasive AI that leaks critical information, accesses confidential data, hallucinates or drives biased decision making.
• and finally, GAI/ AI isn't possible without a strong data foundation. GAI is only as good as the quality of data it has to work with.

Recommended reading

https://www.mckinsey.com/capabilities/people-and-organizational-performance/our-insights/the-organization-of-the-future-enabled-by-gen-ai-driven-by-people

https://www.mckinsey.com/featured-insights/mckinsey-explainers/whats-the-future-of-generative-ai-an-early-view-in-15-charts

https://techmonitor.ai/technology/ai-and-automation/chatgpt-update-openai

https://www.ey.com/en_ca/coo/how-generative-ai-in-supply-chain-can-drive-value

https://www.salesforce.com/news/stories/automotive-industry-security/

https://www.ibm.com/topics/ai-hallucinations

https://www.horsesforsources.com/ai-washing-taking-over-humanity_060223/

DISCLAIMER:

• The ideas, views and opinions expressed in my LinkedIn posts and profiles represent my own views and not those of any of my current or previous employers, Clients, Partners, Vendors or LinkedIn.