This article was written by Mark Palmer, host of Executive Programs for Dataiku. Mark is a data and AI industry analyst for Warburg Pincus and a board member for six AI, data management, and data science companies. Time Magazine named him “A Tech Pioneer Who Will Change Your Life.” Mark is a LinkedIn Top Voice in Data Analytics.
Ten months ago, GenAI was a question mark. Now, it's an exclamation point on financial reports. A recent McKinsey & Company AI leadership survey found that a “GenAI high performer” cohort can now connect as much as 10% of EBITDA to the use of GenAI and 20% to conventional AI. So, GenAI has started to move from lab to ledger.
Unfortunately, at the same time, 70% of leaders say their teams aren’t prepared for AI. Gaps include understanding AI bias, data privacy, data quality, the cost of scoring models, and hallucination. So, while AI use skyrockets and business impact begins to be proven, most are unprepared.
Leadership must fill this gap, but how? Research from McKinsey & Company, MIT, Databricks, and Dataiku gives us some clues and reveals the best practices of GenAI high performers.
1. They View AI as Extended Intelligence, Not Artificial Intelligence
MIT showed that GenAI helps humans make complex decisions 44% faster and 20% better. That’s the good news. The bad news is that the MIT study revealed a problem: 68% of participants used AI’s output without modification. This shows a naive approach to using GenAI and points to where leadership is needed to treat AI as an extension of human capabilities, not a replacement.
The Endeavor AI system at the National University Health System (NUHS) in Singapore illustrates how extended intelligence works. As patients enter health clinics in Singapore, frontline medical staff take digital notes, which are transmitted securely to an AI agent that compares data with a data set of hundreds of thousands of patients to predict the likelihood of over a dozen ailments (cancer, heart disease, appendicitis).
AI predictions (e.g., “There’s a 92% chance this patient has appendicitis”) are returned by the agent to apps used by frontline staff. These AI predictions extend the insights of medical teams, who combine AI predictions with their experience, instinct, and context to decide the best next action.
Endeavor is a master class on how to balance AI predictions and human judgment. AI doesn’t replace human empathy, judgment, and patient collaboration; it augments frontline staff, identifies blind spots they may have missed, and quickly leverages the history and broader perspective on patient data in a way humans cannot.
GenAI high performers know this and “endeavor” to reimagine the balance of decision-making in this way.
2. They Use GenAI Regularly, Personally
Savvy leaders get their hands dirty with GenAI.
McKinsey's study found that executives at GenAI high performers are twice as likely (15% versus 8%) to use GenAI at work every day. They know that to lead effectively with GenAI, you must use GenAI. So, if you’re a C-level leader and do not use GenAI several times a day, get started!
3. They Know That AI Isn't About Algorithms
Laggard organizations get bogged down in AI technology. Leaders view AI as an agent for behavioral change instead. For example, Target’s Store Companion changed the balance of decision-making for store staff by helping them answer customers' questions, remember how to apply for a loyalty card, or even reboot cash registers after a power outage. By incorporating AI into their in-store operational experience and upskilling staff, they’ve changed how guest interaction works and used AI to improve the efficacy of in-store staff. These applications of GenAI improved how employees do their job.
To facilitate these changes in behavior, they don’t always focus on the latest and greatest algorithm. According to Dataiku’s “AI, Today” research, they focus on providing easy access to data and efficiently operationalizing models more quickly to teams.
4. They Employ an LLM Mesh
GenAI leaders are twice as likely (42% versus 19%) to use advanced techniques to separate the AI wheat from the chaff.
That is, the challenge enterprises face isn’t the lack of availability of models, it’s that there are too many models. In June 2024, the AI model repository Hugging Face contained 700,000 models, with more being released every day. This overwhelming number of choices is incredibly confusing for data scientists, let alone less technical users who use AI. For example, tools such as PowerBI, Tableau, and Spotfire give business intelligence analysts or technical researchers the ability to browse and select predictive models to include in business dashboards with a few clicks.
The magic comes from knowing which AI model to choose.
GenAI high performers employ the idea of an “LLM Mesh” that can automate, augment, and simplify the selection of the right LLM services for the job at hand. An LLM Mesh provides an abstraction layer for resources to provide federated control and analysis of resource options and maintain a central discovery and documentation repository for LLM-related objects.
An LLM Mesh makes model assets easier to understand, find, and consume, helping teams work quickly and with increased confidence.
5. They Push More Functional Units to Use GenAI
The McKinsey survey found that laggards use GenAI more narrowly, in one or two functions, typically marketing or sales. High performers engage more groups, more broadly, and involve functions like legal and IT operations earlier in the process of developing solutions. For example, GenAI high performers are more likely to use GenAI in marketing, sales, IT, ops, and legal.
So to be a leader with GenAI, remember: the more, the merrier.
6. Their AI L&D Programs Employ Project-Based Learning (PBL)
GenAI high performers promote a curated learning journey more than twice as much as laggards, 43% to 18%. The style of training matters, too.
Research about what style of training works best for GenAI skills shows that project-based learning is 2-3X more effective than conventional, classroom-style learning. This is because you must use GenAI to learn GenAI.
The Effectiveness of the Project-Based Learning (PBL) Approach as a Way to Engage Students in Learning measured four elements of project-based learning that can lead to GenAI upskilling effectiveness:
- Collaborative Learning (CL): Teachers negotiate knowledge in dialog.
- Disciplinary Subject Learning (DSL): Projects are focussed on a specific use case.
- Iterative Learning (IL): A teaching style of repetitive questioning, meaning-making, reflection, and sharing.
- Authentic Learning (AL): Students are encouraged to actively generate new ideas, share ideas, develop one another’s thinking, and assess their and their colleagues’ academic thoughts.
So leaders prioritize L&D higher than laggards, and their programs are collaborative, disciplinary, iterative, and active.
7. They Embrace Data They Didn't Use Last Year
GenAI high performers are twice as likely (43% to 15%) to have a foundational data strategy that enables new types of data that have been ignored until recently, and they propagate these new data sources more quickly than laggards.
For example, until recently, most companies with large call centers ignored most of their customer conversations; even though speech-to-text translation technology is inexpensive, conversational AI, LLMs, and vector databases make it relatively easy to use. Conversation data can help reveal what customers think, find problems in your business, and understand what call center agents say.
Conversation data is just one example of data leaders use more readily than laggards; having a flexible foundation helps them adapt to new data types more quickly.
8. They Utilize the Scientific Method for GenAI Solutions
High-performing GenAI leaders have a startup mentality because its use cases have yet to be proven at scale. Startups experiment, adjust their plans and embrace failure. Research shows that using the scientific method is one of the most fundamental best practices of that startup mentality. A Scientific Approach to Entrepreneurial Decision-Making: Large-Scale Replication and Extension found that entrepreneurs that use the scientific method generate, on average, €492,000 more than companies that don’t.
The scientific method works because it instills an attitude researchers call methodic doubt — an attitude of critical thinking that helps decision-makers be more aware and avoid overconfidence. Methodic doubt helps leaders formulate, test, and tweak hypotheses and develop GenAI use cases that deliver ROI.
9. They Involve Legal and Risk Teams Earlier in AI Solutions (They Shift Left)
GenAI high performers bring lawyers into the fold earlier than laggards. Involving legal teams in the pilot, requirement gathering, and experimentation stages helps ensure important questions about privacy, IP, rights, security, and appropriateness are as early in the process as possible so that business and technical teams consider them part of their work.
10. They Achieve ROI With Conventional AI First, Then Build With GenAI
Forty-two percent of high performers in the McKinsey study say more than 20% of their EBIT is attributable to their use of non-generative, analytical AI. So, while they can connect 10% of EBIT to GenAI, research shows that using traditional AI is an essential first step.
Indeed, most of the ten best practices shared above for GenAI also apply to conventional AI. So, if you feel behind on GenAI, consider starting with conventional AI as a first step.
10 Steps to Move AI From Lab to Ledger
AI is moving from lab to ledger, and these ten research findings show how leaders should adapt to seize the opportunity to decouple decision-making and infuse AI into their organizational culture and operations.
GenAI leaders recognize AI as extended intelligence and not just artificial intelligence. C-level leaders personally engage with AI on a regular basis. Rather than getting bogged down in AI technology, they focus on using AI as a catalyst for behavioral change. They use the scientific method to guide AI initiatives.
By adopting these ten habits of high-performing GenAI leaders, you can successfully navigate the AI age and ensure that your team is prepared to capitalize on AI's transformative possibilities.