Nan Li, AI leader and former Nationwide VP of AI/ML and Statistical Practice, speaks with Gavroshe Founder Derek Strauss in a video interview about the human-centric AI framework to guide, develop, and implement AI strategy and how to prioritize use cases.

When asked how a human-centric AI framework can guide, develop, and implement organizational AI strategy, Li states that while AI strategy is important, having a business strategy is critical. Adding on, she simplifies the framework using a few key question words: why, who, what, how, and so what.

Elaborating, Li says, “AI strategy should always start with the business strategy.” A business strategy should clearly define where and how AI will drive growth and innovation. The role of AI strategy is then to support and achieve those goals.

“Strategy is all about validating assumptions. Deciding what not to do and making important trade-off decisions,” says Li. Starting with ‘why,’ she stresses why business exists and what strategic role AI could play in it and then further breaks this down into four categories:

1. AI as a value driver: This is where AI is the competitive advantage. This applies to companies like OpenAI or Anthropic, where AI is the business.

2. AI as a key enabler: It enhances the competitive edge, like image analysis in healthcare or drug discovery in pharma. These companies are not AI companies, but AI is essential to their success.

3. AI as a common utility: She takes the example of route optimization in logistics, which is needed but is not a differentiator.

4. AI as a critical defender: AI helps reduce risk in areas like compliance, which protects rather than drives.

Moving on to the ‘who and what,’ Li states that here the focus should be on people and not just technology. This involves asking critical questions like, “Who are you trying to help? Who needs to be involved?” She lists four distinct groups here:

1.      AI sponsors: These are business people who approve and fund AI initiatives and are concerned about ROI. Finance also plays a key role here.

2.      AI users: The ones using or affected by AI tools. This includes employees and sometimes customers whose experiences or roles might change due to automation.

3.      AI builders: A cross-functional team including data scientists, engineers, designers, UX experts, and infrastructure/security folks.

4.      AI protectors: This comprises legal, compliance, ethics, audit, security, and vendor management. They help avoid risk and should be involved early.

When it comes to ‘how’ to execute AI strategy, Li mentions five key components:

• Data

• Governance

• Design

• Technology

• Model

Highlighting the placement of models as the last component, she insists that having a great model means nothing if one does not know how it will change the processes or fit into the business. She emphasizes understanding the future process before building the model.

Out of the five, organizations should own data, governance, and design, says Li, as these are core to success. Tech and models can increasingly be outsourced. She notes that design is becoming a major focus area, especially when it comes to things like error handling with increasing automation.

“My advice is don’t focus on GenAI only; look at all the AI techniques as solutions and start with the tested, proven solutions first,” says Li. That will save time, money, and energy and reduce risk because tech will always evolve, but it is crucial to focus on core business value.

Finally, arriving at ‘so what,’ she discusses what comes next after having strategy and components, and this is about change management and cultural transformation. Li urges revisiting the four personas of sponsors, users, builders, and protectors and tailoring the change management plan to each group.

Sharing her approach to prioritizing use cases in the framework, Li suggests distilling it into a clear question: “Is this a good idea worth pursuing now, for us, over other options?” She breaks this down into four key dimensions of value, risk, readiness and difficulty, and prioritization.

When it comes to value, it is necessary to assess whether it is a good idea aligned with strategic priorities. Moreover, is it worth pursuing considering the potential risks, whether the organization is ready to carry it out immediately, and whether it is feasible, and ultimately if it is the right idea compared to other initiatives.

Thereafter, Li recommends rating each of these dimensions as high or low (value, risk, readiness); you end up with a simple 2x2x2 matrix, with eight possible combinations.

For example, she says that if something is high value, low risk, and low difficulty, that is a no-brainer, and one must go for it. If it is high value, low risk, but high difficulty, Li calls that a competitive advantage.

Some use cases are high-value, high-risk, and high-difficulty; those are big bets that require strategic decisions and are often pursued by AI-native companies. On the other hand, if a use case is low value and high risk, it’s a waste of time, but if something is low value, low risk, and easy, she advises letting people explore it.

Furthermore, Li urges to focus on the value chain and not just isolated use cases. “Don’t start with what technology could do, but start with what we will do and what we should do and find the best available technology to make it happen,” she adds.

According to Li, too many teams take a scattershot approach, but none of them connect back to real, sustained business value. That is because they are not integrated across the full value chain.

Short-term POCs can show surface-level success, but without end-to-end implementation, they will not deliver long-term value, says Li. For example, in sales and marketing, AI should support the entire journey, from lead generation to nurturing, conversion, and post-sale engagement. Otherwise, optimizing one step just creates a bottleneck in the next.

Eventually, every company will have access to similar AI capabilities. “What separates you from the competition is still what value you can deliver to your customers and the trade-off decision you’re willing to make,” concludes Li.

CDO Magazine appreciates Nan Li for sharing her insights with our global community.