TNS Agents Livestream: Matan-Paul Shetrit, Writer

Key Concepts

• Enterprise AI Focus: Writer prioritizes enterprise-level AI solutions with built-in security, compliance, and data control, differentiating itself from research-focused competitors.
• Agentic Systems as the Future: The industry is evolving towards agentic systems – autonomous, skill-based AI – representing a significant leap beyond basic LLMs and custom applications.
• Model Control & Customization: Maintaining control over underlying LLMs (like Writer’s Palmyra series) is crucial for data privacy, stability, and tailored enterprise solutions.
• AI Augmentation, Not Replacement: AI will primarily augment human capabilities, requiring a shift in skills towards managing, evaluating, and interacting with AI systems.
• Proactive Workforce Adaptation: Preparing the workforce for AI-driven changes, particularly in sectors like trucking, is a critical social policy challenge requiring investment in retraining and new skill development.
• The Underestimated Safety Benefit of Autonomous Vehicles: The potential of self-driving technology to drastically reduce traffic fatalities is a significantly overlooked aspect of the conversation.

Writer’s Evolution & the Rise of Agentic Systems

Writer began five years ago alongside the early development of LLMs, uniquely focusing on enterprise applications from the start. This meant building “core primitives” for data retention, security, and compliance – features often overlooked by competitors originating as research labs. Initially associated with marketing content, Writer’s scope has expanded to encompass all forms of “the written word” across various departments including engineering, product, sales, and customer support. The company is now heavily invested in building agentic systems – autonomous, skill-based AI capable of invoking tools and making decisions – representing a “phase three” in AI evolution. Despite the increasing capabilities of off-the-shelf models, Writer continues to develop its own LLMs (the Palmyra series) to maintain control over data privacy (avoiding IP contamination), model stability (preventing unexpected changes by model providers), and customization for specific enterprise use cases.

Building a Knowledgeable & Adaptive AI

Writer is developing sophisticated context graphs to capture organizational knowledge, going beyond simple data storage to represent workflows, decision-making processes, and implicit knowledge. This is coupled with a move towards “self-evolving models” that learn and adapt continuously from usage data, reducing reliance on large, expensive models. This architecture involves smaller, specialized models that are more efficient and effective than larger, general-purpose models. The company is also exploring techniques like RAG (Retrieval-Augmented Generation) to improve LLM accuracy.

The Impact of AI on Labor & Society

The discussion highlights the potential of autonomous vehicles to save “hundreds of thousands if not millions” of lives globally, a benefit often overshadowed by other concerns. However, the rollout of this technology, initially envisioned for long-haul trucking, presents significant labor market implications. The Teamsters union, representing “hundreds of thousands if not more” drivers, plays a vital role as “the arteries of our economy.” The initial phase of autonomous trucking is expected to involve drivers acting as “navigators” in more complex urban environments, similar to roles in seaports.

A critical argument is the need for proactive workforce preparation. Current policy discussions are lacking in addressing how to equip individuals with the skills needed to thrive in an AI-driven economy. Skills like “prompt engineering” are predicted to become obsolete as AI improves, while skills in “interacting with [AI], managing agents, [and] building evals for your domain” will be crucial. Both companies and governments (federal, state, and local) have a responsibility to invest in “retraining” and “on-job training” initiatives.

The Evolving Skillset & the Importance of Critical Thinking

The speaker illustrates the changing skill landscape with the example of their wife, a writer. AI doesn’t eliminate the need for writing, but elevates the importance of skills like “editing” and “factchecking,” particularly in an era of “fake news” and misinformation. AI acts as a “forcing function” for critical thinking, requiring individuals to evaluate the accuracy and validity of AI-generated content. The retiring “boomer generation” is being replaced by tech-savvy individuals and “AI native” generations, necessitating a rethinking of educational and professional development approaches.

Political Challenges & the Urgency of Preparation

The speaker acknowledges a political challenge, noting that “politicians are always not always proactive when it comes to these things.” There is a concern that “people are not prepared for what’s coming at them and people don’t know what’s coming at them,” emphasizing the urgency of addressing these issues.

Conclusion

The conversation underscores a pivotal shift in the AI landscape – from basic LLMs to sophisticated agentic systems designed for enterprise needs. Maintaining control over models, prioritizing data security, and focusing on specialized solutions are key differentiators. Crucially, the discussion highlights the societal impact of AI, particularly the need for proactive workforce adaptation and a renewed emphasis on critical thinking skills. The potential for AI to save lives through technologies like autonomous vehicles is significant, but realizing this benefit requires careful planning and investment in preparing individuals for the evolving demands of an AI-driven future.