Comprehensive Guide to Developing & Executing a Winning AI Corporate Program

The hype surrounding artificial intelligence (AI) has reached unprecedented levels—and for good reason. OpenAI's ChatGPT Plus service amassed over 1 million subscribers within mere months of its late 2022 debut. Large Language Models (LLMs) represent a quantum leap in natural language processing, enabling more human-like interactions and unlocking many applications, from revolutionizing customer service to transforming content creation.

According to IBM's Global AI Adoption Index, as of November 2023, most enterprises actively explore or deploy AI and have some form of AI strategy. 27% report that their company has an AI strategy for limited/specific use cases, and about a third (32%) state that their organization already has a holistic strategy. 32% are in the process of developing an AI strategy. This trend underscores the importance of developing a robust AI strategy to stay competitive.

But amidst this AI gold rush, a crucial question emerges: How can companies harness AI's benefits faster than their competitors and achieve a higher return on investment than their cost of capital?

The answer lies in a formula not dissimilar to customer acquisition strategies. Your organization's AI IQ will hinge on its ability to identify and optimize two key metrics across multiple use cases and workflows:

The AI Value Equation

Much like customer acquisition strategies, an organization's AI proficiency hinges on two key metrics across multiple use cases and workflows:

1. Lifetime value of AI investment
2. Cost of acquiring AI capabilities

This article will delve into both sides of this equation, providing a comprehensive AI approach for the Enterprise. The higher this ratio compared to your competition, the more urgently you need to scale these capabilities.

As Mohamad Ali mentioned, this year’s focus at IBM Consulting is on creating business value from #AI. Our teams have led numerous clients through their AI journey, helping them navigate the complexities of AI adoption and maximize their return on investment. Drawing from this wealth of experience, we've developed a proven framework for AI success.

Key Requirements for Success

To forge ahead in the AI race, enterprises must focus on several critical areas:

1. Finding High-Value AI Opportunities

Start with Business Pain Points and Opportunities:

Begin by conducting thorough interviews across your organization to identify pain points and opportunities within key business workflows. A prioritization exercise focused on top workflows is crucial—it will not only concentrate your AI program on the most impactful areas but also challenge your team to ideate and creatively leverage AI for value-driving processes.

At this stage, business leads may have preliminary thoughts on how AI could enhance workflows. While capturing these ideas is valuable, identifying concrete AI opportunities should come later, informed by subject matter experts.

Determine the KPIs:

Pinpoint the key performance indicators (KPIs) that will improve if your company successfully addresses pain points and captures opportunities. This will include existing metrics and important new KPIs that need to be measured. Develop a value tree to drive the selected KPI and identify the leading indicators constraining the chosen metric. This analysis will help determine where to focus when building use cases.

Identify AI Use Cases:

Map potential AI use cases to the workflows where they could significantly impact the prioritized KPIs. You may generate dozens of AI ideas for each functional area. A value driver tree centered around the KPI will be instrumental, with AI use cases tied directly to the drivers.

Prioritize Workflows and AI Use Cases:

First, prioritize the workflows with KPIs with the highest potential to add value. Then, select the AI use cases that would most significantly impact these KPIs. This order is crucial—if you prioritize use cases first, you'll likely gravitate towards the most marketed options, which may not be differentiating for your business. By prioritizing workflows and then ideating on use cases, you'll foster creativity and uncover unique applications with the most substantial impact.

After completing this prioritization, your business leads and technical teams should collaborate in an agile fashion to build a tangible plan around the top 2-3 use cases for each prioritized workflow.

2. Moving from Prioritization to Quick Proofs of Concept (PoCs)

With 1-3 AI ideas identified for each prioritized workflow, it's time to craft an execution plan. Leveraging agile methodologies minimizes risk, reduces investment, and allows for the development of AI solutions through rapid feedback gathering and iterations. PoCs serve as your first quick demo, painting a vivid picture of what the final solution will resemble.

PoCs are crucial for jumpstarting your AI journey—action creates momentum. Without them, funding decisions for MVPs or full products could stretch for months, potentially killing excitement and momentum.

A well-designed PoC:

- Uses dummy data that mimics actual company data

- Showcases the solution for a limited scope (e.g., one user persona, one channel, one task within the workflow)

- Should be scoped to be built in 2-4 weeks, resulting in a clear go/no-go decision

- Can use "synthetic data" and simple interfaces like Stream for easy demonstration

If a PoC fails to excite your organization or generate ideas for improvement, it may not be worth pursuing an MVP for that use case at this time.

The benefits of quick PoCs include:

- Increased confidence in the MVP before committing to a longer timeline

- Gathering key MVP requirements based on user feedback

- Enhanced stakeholder engagement, helping to secure funding for the MVP

Companies aiming to move fast and leapfrog competition should target completing 5-10 PoCs in high-value workflow areas identified earlier. With a team of 4-6 AI strategists, data scientists, and developers, it's feasible to complete these PoCs in an 8-week project. GenAI-only PoCs typically take less than three weeks, leveraging LLMs, while PoCs with traditional AI components may require slightly more time.

3. Moving to MVP Planning

The Minimum Viable Product (MVP) is a critical step when building AI solutions for enterprises. Here's why:

1. Faster time-to-market: Launch a basic version quickly, gaining a competitive edge and early user feedback.

2. Cost-effectiveness: Focus on core features to reduce initial development costs and resource allocation.

3. Validation of concept: Test the viability of your AI solution in a real-world environment.

4. User feedback: Gain valuable insights from early adopters to guide further development.

5. Risk mitigation: Identify and address potential issues before full-scale implementation.

6. Iterative improvement: Enable continuous refinement based on real usage data and user needs.

7. Stakeholder buy-in: Demonstrate a working product to secure support and additional resources.

8. Technical feasibility assessment: Evaluate the practical challenges of implementing AI within your infrastructure.

9. Change management: Introduce AI solutions gradually for smoother adaptation.

10. Data collection: Start gathering valuable data for training and improving AI models from day one.

Your MVP plan should include these core elements:

MVP Functionalities:

Conduct a 2-week discovery phase to prioritize functionalities for the MVP. User feedback from the PoC stage will be invaluable here. Resist the temptation to include too many features—focus on one channel, one user type, and one platform. Simultaneously, assess data requirements and select capabilities that won't demand a heavy lift in data collection and engineering.

MVP KPIs:

Select KPIs that the MVP will directly impact. If you've followed the priority-driven process described earlier, your MVP should target an important KPI. Include leading indicators for the target KPI as well, ensuring baseline measurements are in place.

Financial Analysis:

Develop a robust benefit and cost analysis for your AI initiative and MVP scope. While financial analysis isn't the sole factor in decision-making, it's a core component for securing funding. Estimate costs across key layers: data, AI services, UX design, and technology/GenAI token costs.

Remember, ROI analysis may not fully capture intangibles like brand or customer lifetime value. However, it's crucial to think through financial implications and present your best estimates. The goal of the MVP is to validate assumptions about the AI initiative's value at a lower cost. Once stakeholders validate the opportunity, a bias for action is the best path forward.

4. Determine Resources and Estimated Costs for MVP and Full Product

As excitement builds around your use cases and PoCs, conduct a reality check. With LLMs likely integrated into various workflow areas, proactively assess potential costs, especially variable expenses related to GenAI. This visibility is crucial for effective planning and budgeting.

Remember that AI solutions demand diverse skill sets spanning UX design, AI services, and data management. Ensure your team has the right mix of expertise across these domains:

1. UX/UI for AI Integration:

Focus on creating intuitive interfaces that seamlessly incorporate AI into workflows. Key roles include UX/UI Designers, Interaction Designers, User Researchers, Front-end Developers, and UX Writers.

2. AI/ML Services:

This encompasses the AI and machine learning technologies powering your solutions. Essential roles are Data Scientists, Machine Learning Engineers, AI Researchers, and Domain Experts. Key resources include LLM tokens, cloud computing infrastructure, and GPU/TPU hardware for larger projects.

3. Data Infrastructure:

This area focuses on collecting, processing, and managing data for AI systems. Critical roles comprise Data Engineers, Database Administrators, Data Architects, ETL Developers, and Data Governance Specialists.

5. Implementing an AI Operating Model for Execution

In most cases, AI initiatives need to be part of innovation driven by a separate organizational structure and operating model. The core component of this model is the establishment of an AI Steering Committee.

Key responsibilities of the AI Steering Committee include:

1. Strategic direction and governance: Define the overall AI strategy, establish ethical guidelines, and ensure compliance with regulations.

2. Resource management: Approve budgets and reallocate funding in an agile manner as project needs evolve.

3. Project oversight: Agree on specific AI initiatives and their corresponding KPIs, conducting regular reviews throughout the AI lifecycle.

4. Risk and performance evaluation: Identify and develop mitigation strategies for AI-related risks, review project outcomes and measure ROI.

5. Stakeholder engagement and innovation: Facilitate cross-functional collaboration and promote a culture of innovation and continuous learning.

Potential Roadblocks to AI Change

Resistance within the organization is natural, given that AI implies significant changes in processes, resources, technology, and KPIs. Some key challenges include:

1. Change management: AI adoption requires effective training, communication, and reskilling initiatives.

2. New ways of working: Building differentiated AI solutions demands agile methodologies, which may be unfamiliar to traditional IT or digital teams.

3. Skill gaps: Organizations may struggle to acquire necessary AI skills internally. Partnering with credible external parties can accelerate development and ensure accountability for results.

4. Dependency on a single provider: Being locked into a single vendor's ecosystem can make integrating with other technologies or switching providers difficult. Embrace the importance of interoperability and choice.

Conclusion

To develop and execute a winning AI strategy, follow these key steps:

1. Identify and prioritize key workflows and KPIs

2. Map use case ideas to selected workflows and prioritize 1-3 AI use cases

3. Quickly develop prototypes for selected use cases

4. Develop an MVP plan for multiple use cases and secure funding

5. Establish a CEO-led AI Steering Committee with clear governance

6. Implement a stage-gate plan for reviewing MVPs: stop, revise, or scale based on outcomes

By following this comprehensive approach, enterprises can position themselves at the forefront of AI innovation, driving significant competitive advantages and unlocking new realms of business potential. The AI revolution is here—those who act decisively and strategically will reap the rewards of this transformative technology.

The future of AI is bright, and the time to act is now. Let's embark on this transformative journey together and unlock the full potential of AI for your enterprise. Your competitors are already moving—don't get left behind in the AI revolution.