In the wake of the week…

Leadership teams are confronting a fast-changing AI landscape with outdated metrics and mental models. Pennsylvania’s pilot offers a real-world playbook for deploying generative AI at scale with human-centered design. New reasoning models from Google, Microsoft, OpenAI, and others signal a shift in capability that breaks from traditional software cycles, proving that AI is a system that must be managed in motion.

Let’s dive in. 👇

ONE // AI Demands a New Kind of Leadership—and a New Kind of Math

Boards and executive teams are all facing the same conundrum: AI offers undeniable promise, but the returns remain uneven. New reports from Gartner and Deloitte, and an article from AI veteran Meghan Anzelc trace the problem to its roots. The technology is evolving faster than the mental models and value frameworks used to manage it.

What’s Happening

• Gartner: Despite heavy investment, just 34% of generative AI users report strong productivity gains.

• Anzelc: Leaders shape adoption through mindset. “Pilots” embrace AI, while “Passengers” resist, dragging down team engagement.

• Deloitte: Traditional ROI models—focused on efficiency and cost—are no longer sufficient. New tech enables new work, and organizations need new metrics to capture value.

The Disconnect

C-suites tend to view AI through a strategic lens, anticipating new skill needs and value creation. Entry-level workers and line teams often remain skeptical or confused, unsure how AI will affect their jobs. Boards and finance leaders, meanwhile, still lean on outdated assumptions: that AI’s value can be measured in headcount reductions or short-term cost savings.

AI does not merely automate work, it changes the nature of work. Yet many organizations evaluate its impact as if it were just another tool for faster execution. This creates a dangerous mismatch between what the technology enables and what the business case expects.

Leadership Insight

Anzelc urges directors to lead with curiosity and agility, keeping pace with fast-moving capability shifts. Gartner advises CFOs to temper their assumptions and focus on internal redesign. Deloitte goes a step further, calling for an entirely new calculus: one that measures outcomes like workforce resilience, innovation velocity, and long-term adaptability.

Case Studies In Action

• Salesforce: Deployed low-code AI agents not to cut staff but to build agility and improve HR service quality.

• Eaton: Used AI to revamp hiring, increasing candidate velocity and improving recruiter outcomes.

• Grupo Bimbo: Focused on safety, customer satisfaction, and front-line performance, not just efficiency.

The Bottom Line: AI's potential will not be realized through old playbooks. Leaders who apply old metrics to new tech will miss the mark, eroding trust, wasting resources, and exhausting their teams in the process.

Sources: Deloitte [report], Directors & Boards, Gartner [report]

TWO // Pennsylvania’s AI Pilot Demonstrates What Pragmatic AI Deployment Looks Like

The Commonwealth of Pennsylvania has completed its first generative AI pilot across 14 state agencies, offering one of the clearest real-world cases of how AI can improve large-scale operations. Governor Josh Shapiro unveiled the results alongside OpenAI and Carnegie Mellon, positioning the state as a national leader in responsible AI integration.

Driving the News

The pilot program provided ChatGPT Enterprise access to 175 state employees over the past year. Participants reported saving an average of 95 minutes per day through AI-assisted writing, research, summarization, and tech support. The program was built on principles of transparency, human oversight, and employee empowerment.

By the Numbers

• 14 agencies participated across policy, HR, IT, and program management

• 95 minutes saved daily on average per employee

• 85% of users reported a positive experience

• Nearly half had never used ChatGPT prior to the pilot

Employee Impact

Feedback shows employees used AI to generate public health campaigns, translate legal language into plain English, consolidate state policies, and reduce administrative workloads. Training played a critical role in success, with many participants emphasizing the importance of learning how to prompt effectively and review outputs critically.

Leadership Insight

Shapiro framed the initiative as part of a broader shift in how government must evolve with technology. Speaking at Carnegie Mellon, he called AI “one of the most significant technological developments of our time” and committed to future AI use that strengthens, rather than sidelines, public servants.

The Bottom Line: Pennsylvania’s pilot offers a credible roadmap for putting AI to work in complex organizations. With a focus on measurable outcomes, thoughtful design, and employee engagement, it shows how AI can boost productivity and unlock capacity across a wide range of roles and workflows.

Sources: pa.gov [report], Govtech.com

THREE // Model Mania Hits a Fever Pitch

Editor’s Note: When I started this newsletter, I said I wouldn’t chase model updates. There’s plenty of that elsewhere. But in a sector that is never short on new tech drops, this week was a doozy. Here’s what you need to know…

A new wave of AI models is prioritizing reasoning, reflection, and problem-solving. That shift is redefining what the most advanced AI systems can do, and how businesses might use them.

Driving the News

Google’s Gemini 2.5 Pro
Google launched its most advanced model yet, boasting deep reasoning skills and top-tier performance in code and math.
It can take in and work with massive amounts of information at once — think dozens of documents or entire websites — making it ideal for high-context tasks like coding, research, or strategic planning.

Microsoft’s Deep Research Tools
Microsoft introduced "Researcher" and "Analyst," two Copilot tools powered by OpenAI’s reasoning models.
These agents can tap both internal enterprise data and the public web to generate strategic analyses and iterate through complex tasks.

OpenAI’s 4o Image Generation
OpenAI rolled out native image generation in GPT-4o with improved fidelity and visual reasoning.
Users can now create polished, coherent, and editable images using conversational prompts — and it’s already live in ChatGPT.

DeepSeek’s V3-0324
The fast-moving Chinese startup, increasingly banned by government agencies and large corporations, released a new reasoning model through Hugging Face with upgraded coding chops.
Early benchmarks show it performing competitively with Western models, at lower compute costs.

Reve Image 1.0
Palo Alto-based Reve AI debuted its first model focused on prompt-aligned image generation and typography.
It leads the current leaderboard for image generation quality and aims to interpret creative intent, not just visual fidelity. [Editor tested. It’s awesome!]

Leadership Insight

This week’s wave of model releases is a reminder that generative AI does not follow traditional software release cycles. We've moved from predictable annual upgrades (on-prem), to quarterly updates (SaaS), to a world where new capabilities can drop anytime, anywhere, without warning. That’s right. Continuous, unscheduled deployments.

For leaders deploying generative AI tools to your workforce, this requires a new level of operational readiness. Communication plans, employee training, policy development, and oversight mechanisms must be prepared for changes that can arrive without notice. This pace of advancement also makes a strong case for building custom deployments within the more stable enterprise environments offered by cloud providers such as AWS, Google, and Microsoft, if that is within your company’s budget and capabilities.

The Bottom Line: The era of surprise upgrades has arrived. AI capabilities can shift overnight, pushing leaders to treat generative AI not as a static tool but as a living system that demands active governance, continuous learning, and thoughtful infrastructure choices.

Sources: TechCrunch, The Verge, Reuters, Google, OpenAI, VentureBeat

Extra Credit

For the overachievers: These are the stories that didn’t crack the top three but are too important to ignore—quick hits on what’s happening and why it matters.

Stanford Economist: Don’t Expect Quick Returns Without Thoughtful Planning

Key Takeaway: Stanford economist Erik Brynjolfsson says AI will reshape the economy as profoundly as the steam engine, but only if businesses rethink how they measure productivity and redesign workflows around new capabilities.

Why It Matters: Too many companies still expect AI to yield quick returns without changing operations. Brynjolfsson argues that the biggest barrier to productivity is no longer the technology; it is leadership stuck in outdated assumptions.

Source: HBS Managing the Future of Work Podcast

The Cybernetic Teammate: AI Joins the Team

Key Takeaway: A landmark study at Procter & Gamble shows AI can replicate the benefits of a human teammate, boosting performance, bridging expertise gaps, and enhancing emotional experience at work.

Why It Matters: This research reframes AI not as a productivity tool but as a collaborator, signaling that organizations must rethink how they structure teams, develop talent, and manage expertise in an AI-integrated workplace.

Source: Harvard Business School [study]

ChatGPT Team Now Connects to Google Drive

Key Takeaway: OpenAI is rolling out internal knowledge integration for ChatGPT Team, starting with Google Drive. This feature enables the model to directly pull in real-time information from your workspace to provide more relevant, personalized answers.

Why It Matters: This update removes friction from daily tasks by grounding responses in your team’s actual documents, acronyms, and workflows. It makes ChatGPT useful for onboarding new hires, prepping for meetings, writing proposals, answering internal FAQs, and summarizing project docs right from within your workspace. It marks a major step toward turning ChatGPT into a practical productivity layer for teams inside the tools they already use.

Source: Nate Gonzalez, OpenAI (via LinkedIn)

Want to chat about AI, work, and where it’s all headed? Let’s connect. Find me on LinkedIn and drop me a message.