Based on a panel discussion featuring Ekaterina (Katie) Curry, SOM ‘02 (Chief Client Services & Operations at MSI) and Michael Andolina, SOM ’15 (Strategic Advisor, Human Capital, The Josh Bersin Company)
The headlines about AI would lead us to believe that we are all falling behind on adoption and learning, but the reality is more nuanced. Two Yale SOM alumni who work at the operational and organizational front lines of AI shared their candid take on what has changed, what is working and what every manager needs to think about differently.
At the organizational Level: from Experiment to Enterprise
We started the discussion with the marked shift in 2026 from AI literacy to AI fluency: moving from isolated productivity towards more integrated organizational cases. Katie described the transition succinctly: AI is moving from “fanboy/fangirl” enthusiasm to genuine enterprise level adoption where agents are being deployed to automate and augment processes.
Michael discussed organizational experiments that sound cool but lack a real framework while also giving us examples of personal tasks and productivity.
Together, those examples show the gap between scattered individual wins and coordinated enterprise adoption. So, what does this mean organizationally for new roles and for you?
Team Level: New Roles, New Expectations
The organizational shifts are still in initial stages but are starting to show up in team structures and staff. Katie described several emerging roles that she did not anticipate two years ago:
AI Trainers: specialists who work iteratively with AI platforms to build domain-specific knowledge bases.
AI Quality Controllers: professionals who audit outputs for accuracy, compliance and alignment with business norms and principles.
Knowledge Managers: owners of organizational data (standard operating processes, process documentation, and policies).
These roles require a technical grounding, but they lean most heavily on human skills: sound judgment, critical thinking skills and systems thinking. They also require deep domain expertise to be able to know where AI is failing or succeeding. Katie acknowledged that this is where she sees possibilities but didn’t disregard the fear that AI will replace some existing roles.
Michael’s firm evaluates client AI adoption on a 1–4 scale: Levels 1–2 focus on personal productivity (doing tasks faster), Level 3 is process redesign, and Level 4 is autonomous agents. He said most organizations are still at Levels 1 or 2.
Katie and her team built an AI voice agent that handles renters’ insurance sales 24/7, escalating to humans only when specific judgment or licensing is required. The new AI agent has allowed them to reach more customers more quickly. No roles were eliminated yet, but possibilities do exist for less headcount while the remaining roles done by humans require more subject matter expertise and judgment.
Strategic Process Mapping: Katie used a protected and closed enterprise AI platform to analyze dozens of company SOPs, training materials, and “one-pagers”. She tasked the LLM with identifying inefficiencies, distinguishing “quick wins” from long-term projects, and outlining exact technology dependencies and API integrations needed for a six-month plan.
Michael referenced a conversation with a large strategy consulting firm who was boasting about 4,000 GenAI tools deployed across the firm with no real governing framework around them. This is consistent with Level 1–2 adoption: lots of tools, limited integration.
Given that most organizations are still at level 1 or 2, they believe that we still have a six-month to 12-month opportunity to experiment and learn how to use these AI platforms personally and organizationally. The key is to start with sandbox experiments and iterate.
But change must be led by leaders who encourage experimentation and are willing to get into the trenches with their employees.
What good leaders are doing
Both agreed that effective leadership now requires a blend of traditional soft skills and technical intuition.
Katie has shifted the default question from, “what is the right tool for this problem?” to “can I solve this problem with AI? And only then, what else? This is what she also encourages her team to do. Before using the tool, have you taken the time to understand the problem? Or how you framed the problem? Without this initial reflection, the tool, regardless of which one, will only deliver noisemi or inaccurate recommendations.
Michael has created a master prompt that instructs his platform to be direct, skip validation and point to evidence rather than comfort. His broader observation is that the value of AI is highest when it helps you do what you would have done well anyway – faster and with more information – but not to replace your own judgment.
As an example, Michael used the company’s proprietary AI tool to attend his meetings and provide feedback on his interpersonal dynamics. It once noted that he spoke 80% of the time and didn’t let others speak, and it helped him interpret if a teammate felt “shot down” during a difficult call. In another instance, the AI pointed out that the coworker wasn’t rude but was simply being “precise,” helping Michael adjust his perspective.
Katie uses AI daily as a thought partner and Devil’s advocate using it to review her decisions and ask, “Have I missed anything?”. She recommends explicitly prompting the AI to act as a devil’s advocate rather than a “yes voice” that simply validates your current thinking.
Another expectation is that all leaders will have to become “technologists” even in a non-technical role meaning everyone will need to have a foundational knowledge of the LLMs while also understanding APIs and technology dependencies. Think back to the emerging roles referenced above.
Regardless of whether it is at an organizational or personal task level, the human must still play a critical role in overseeing, training and providing the final judgment on the output of AI models.
What does this means for you: Three things to do now
The panel closed with practical recommendations. Three stood out as actionable for alumni at any career stage:
- Understand the model, not just how to use it: Michael recommends reviewing the tool’s stated principles and usage guidance (and, where available, its data/privacy stance) so you understand the defaults it may optimize for. Different tools make different tradeoffs between safety, capability and openness).
- The Master Prompt: As noted earlier, Michael developed a specific “master prompt” for his AI agent to align it with his personality. His prompt instructs the AI to avoid “sugarcoating” or fluff, provide evidence-based directness, and stop providing “therapist-like” validation. Prompting takes patience and a lot of experimentation. Katie mentioned that one prompt took her two weeks of iteration – which is why this is an important skill. But do keep track of the newer version of the models, as this also changes how prompting works.
- Shift the Default Question: Katie advises shifting your mindset from “what tool should I use?” to “can I solve this problem with AI?” If the answer is no, only then should you look for other software solutions. What she is referring to is taking the time to frame the problem you are trying to solve. Without this clarity, you will be using the tool to deliver output that might be solving a different problem.
If you missed our webinar, you can find the full recording: here
Interested in continuing the AI and its impact? Don’t forget the alumni group Yale AI & Impact Makers. Their next meeting is Thursday, May 7th. Register here: Y-AIM: Yale AI & Impact Makers
Upcoming Webinars:
Join the CDO for a conversation on AI with Dignity: Learning, Motivation, and Flourishing. We will explore how employers are adopting AI in ways that grow people—not just output. This panel examines practical strategies to motivate continuous learning, build confidence with new tools, and create the psychological space for upskilling and career mobility.
When: Thursday, May 21st from 12-1
Jonathan Beauford, SOM ’12, Global Talent Leader at Yahoo
Mario Ruiz, SOM ’19, Managing Partner at Infinity Ventures
Blue Steps Webinar
Mastering the Modern Job Search with AI
Wednesday, May 27th from 12-1pm: Register here
Pathway to the Boardroom: Turning Executive Experience into Board Value
Wednesday, June17th from 12-1pm: Register here
