In this week’s edition, we feature an excerpt of an essay by May Habib, CEO and co-founder of Radical Ventures portfolio company Writer, on three shifts to the leadership playbook for executives looking to succeed in the age of AI.
The leaders we see actually scaling AI aren’t reaching for the familiar, comfortable playbook that got them the corner office. They’re tearing it apart and radically rebuilding from the inside out. Here are the three shifts that define them.
- They Design for Radical Simplicity
The leaders driving agentic AI forward have shifted into ‘rebuild mode’. That’s a very different exercise. It’s a zero-based design, and it starts with asking:
- What actually matters here? Don’t look at a step on your flowchart and ask, “How can AI make this step faster?” Clear the whiteboard entirely. Take your most critical processes – product launches, customer onboarding, closing the books – and redesign them around the outcome, not the org chart. If you find yourself asking, “How did this survive for 20 years?” – that step goes.
- Is this ‘approval theater’? Look at every approval, review, and sign-off and ask whether it adds real judgment – or just moves information along. If it’s the latter, it’s bureaucratic scar tissue. That step doesn’t belong in a meeting or a manager’s inbox anymore. Cut it.
- Is this pure coordination? Scan for status meetings, Slack threads, and “just-in-case” checks whose only purpose is keeping people aligned. Replace it with an agent that tracks progress, flags issues, and routes decisions automatically. Cancel the recurring meeting.
Only leadership can look at a workflow and say, “this is where we add our genius, that part has to go.” That doesn’t fall on your product or ops team. This is now the most important part of your job.
- They Redefine How People Grow
Forward-thinking leaders aren’t asking less of their people, they’re raising the bar. They’re asking what humans are uniquely capable of – the work that makes someone irreplaceable – and rethinking roles from the ground up:
- Strip and rebuild. Take one role on your team and remove every task an agent can do – data entry, report generation, status updates, first-draft creation. Then rewrite the role around setting direction, making judgment calls, and being accountable for the result.
- Expand scope. Narrow job specs are a relic of the past. Redesign roles to encourage sideways growth and skill stacking, creating true “generalists”. Take your “content strategist” and make them a “growth architect” who orchestrates content, distribution, analytics, and optimization. Sideways growth, skill stacking.
- Redesign growth. Career ladders are dead. The next era of your career won’t be driven by experience, credentials, or hustle, it’ll be driven by AI leverage. Promote people based on how effectively they multiply their impact. Let people follow problems across functions, not titles up a ladder.
The most creative, strategic, and human work of our careers is ahead of us.
- They Expand the Boundaries of Ambition
Start with the excuses you pulled two years ago. Maybe it’s “We can’t scale customer success without hiring 50 more people,“ or “We can’t enter SMB — the economics don’t work.”
Then, ask again:
- What if you could serve everyone like your top 1%?: Instead of asking how to better segment customers, ask how to deliver 1:1 expertise to everyone, not just your high-value accounts. AI makes personalized experiences economically viable at scale.
- Which timelines can we shatter? Research, development, experimentation – work that ran on multi-year cycles – can now ship in weeks. Take one strategic initiative currently on your three-year roadmap. Force it into a 90-day window.
- Which markets are we still treating as “too hard”? Skip the years-long localization plans and use an agent to handle language, support, and compliance. Leaders can now enter new markets simultaneously, not sequentially. What if we could run 1,000 simultaneous pricing experiments instead of one A/B test?
Read May’s full essay featured in Fortune.
AI News This Week
-
These Diseases Were Thought to be Incurable. Now AI is Unlocking New Treatments (BBC)
Researchers are using AI to uncover treatments for drug-resistant bacteria, which now kill 1.1 million people annually, a toll expected to exceed eight million by 2050. MIT researchers used AI to screen 36 million compounds against drug-resistant gonorrhoea and MRSA, identifying two candidates that target bacteria in ways existing antibiotics cannot. At Cambridge, AI screened billions of molecules in days to surface five promising Parkinson’s compounds, work that would have taken six months and millions of dollars using conventional methods. At Harvard, an AI model found nearly 8,000 approved drugs with the potential to treat 17,000 diseases, offering a path to find treatments for rare conditions that pharmaceutical companies have little financial incentive to pursue. Startups are moving to commercialize these approaches, with Radical portfolio company Nabla Bio, for instance, developing an AI system that designs therapeutic antibodies from scratch against targets once considered intractable.
-
What Entertainment Might Look Like in 20 Years (WSJ)
AI is rewriting how movies and games get made. Generative AI will allow filmmakers to produce realistic crowds and battle scenes without location shoots or hundreds of extras, work that today requires large studio budgets. On the production side, AI test audiences trained on historical viewer data will let directors test plot directions and endings before a film is finished, replacing traditional focus groups. In gaming, AI companions will persist across different games and decades, accumulating shared history with players over time. Democratized access to tools once reserved for studios and publishers will allow creators to bring more ideas to life and richer content for audiences.
-
Don't Get Used to Cheap AI (Axios)
AI labs have been pricing aggressively to build user bases. Margins remain negative across much of the industry (a significant exception being Radical portfolio company, Cohere). With OpenAI projected to burn $14 billion in 2026 and both OpenAI and Anthropic widely expected to go public, investor pressure will demand a path to profitability. “These LLM companies are going to go public, and they’re going to raise prices because they have to,” said May Habib, CEO of Radical Ventures portfolio company Writer. The dynamic mirrors how Amazon and Uber built their user bases on subsidized pricing before eventually charging enough to cover costs and generate profits. Aggregate token pricing has fallen sharply as inference efficiency improves, but total AI spending continues to climb as usage surges.
-
AI is Helping Expand the Frontier of Theoretical Physics (The Economist)
AI is proving its worth in frontier academic research. A group of theoretical physicists used generative AI to crack a problem in subatomic particle interactions that had stalled their work for months. The AI spotted simplifications the researchers had missed, generated original conjectures, and produced a full proof after 12 hours of reasoning. The team then extended the findings to a related problem, with the AI largely working independently. “The physics problem now is not the hard part,” said one researcher. “The hard part is verifying the results and writing it up.”
-
Research: Proactive Real-Time Agentic System Capable of Modeling Human State of Mind (MIT)
Researchers have built an agentic system that reads brain signals from consumer-grade wearables to model a user’s mental and emotional state in real time, then acts on that model without waiting for explicit commands. NeuroLoop runs fully on-device, continuously querying a vector space built from biophysical and neural data to decide when to intervene, simplify a response, or suggest a break. The system runs offline with no cloud dependency, and has been tested in education, gaming, and assistive communication contexts for people with ALS and minimally verbal autism.
Radical Reads is edited by Ebin Tomy (Analyst, Radical Ventures)
