This post was adapted from a summary of a meeting transcript generated by NotebookLM. Carleton’s enterprise agreement with Google ensures that information given to NotebookLM by Carleton accounts will not be used to train Google’s AI models.
In March 2026, The AI Coordinating Team had the opportunity to sit down with Rebecca Ditsch ‘93, a former religion major and attorney who now serves as a Senior Director of Product Development at Thomson Reuters. Thomson Reuters is a technology company that offers tools and resources to professionals in the legal, tax, and supply chain sectors. It has also made significant investments in AI and offers its own suite of generative AI tools. Ditsch is also the President of Carleton’s Alumni Council.
Committee members had the chance to ask Ditsch questions about AI trends and developments in the corporate world, and how they inform the postgraduate landscape. Questions ranged in topic from the specific AI skills employers look for to the practices used by Thomson Reuters employees to verify AI output. Below are some highlights from the conversation, which have been edited and condensed.
Question: As students prepare for the workforce, what specific “AI skills” do employers actually expect them to have?
Rebecca Ditsch: It’s less about being an ‘AI expert’ and more about effective prompting and open-mindedness. You need to know how to describe a persona, provide clear examples of desired output, and explicitly tell the tool what to avoid or when to indicate it doesn’t have an answer.
Question: Should we look at AI fluency as requisite for all or most jobs, or should we see it as specialized knowledge?
Rebecca: If your job requires you to work with a computer, you will likely need some level of facility with AI. While specialized roles in marketing, data science, or analytics might require “super users,” most office roles will require at least a basic understanding of how these tools can assist with data and daily tasks. In the near future, AI fluency will be as assumed as proficiency in Word or Excel, which people used to note on their resumes!
Question: How do you perform a quality check when using AI to analyze massive, years-long datasets?
Rebecca: When dealing with that much data, I am usually looking for trends and patterns rather than 100% precision. I use visualizations to spot anomalies and frequently upload the same data into multiple different LLMs to see if they produce consistent observations. In general, though, the degree of precision required depends on the task at hand. If I were using AI to make medical diagnoses, for instance, the story would be different. It’s also worth mentioning that humans can also make mistakes and struggle with precision. You always have to be careful.
Question: In a corporate environment, what does the verification process look like for AI output? Is “the AI made a mistake” a valid excuse for an error?
Rebecca: At Thomson Reuters, we emphasize a “human in the loop” model. Whether it’s an AI or a junior associate drafting a brief, a subject matter expert must review the work. We use embedded citations so users can immediately verify source documents. Accuracy is paramount; you would be summarily marched out if you submitted work without verifying it first. So “the AI made a mistake” is absolutely not a valid excuse.
Question: There is a concern that an AI-fueled transition from authoring to editing might produce adverse cognitive effects. If more jobs consist in the boring work of checking and deciding “that looks fine,” employees may not have as many opportunities to use and sharpen their expertise. How do you view this shift?
Rebecca: While the work is changing from authoring to editing, it doesn’t have to be boring. For one, there is the fact that some people like to be copy-editors. For another, automating the “grunt work”—like dry, informational updates—actually frees up employees to do more interesting, analytical work. This allows our teams to focus on deep-dive interviews and high-level strategy rather than just routine data gathering.
Question: If AI automates the lower-level tasks usually assigned to new hires, what happens to entry-level positions?
Rebecca: My specific department doesn’t hire straight out of college, but I know of departments that do hire recent college graduates. From what I understand, this is a challenge that organizations, especially law firms, are currently grappling with. While some may hire fewer people, another path is using these efficiency gains to allow the same number of staff to produce a higher volume of work more quickly. You still need to understand the “expected output” to guide the AI, so foundational knowledge remains essential.
Question: If the number of entry-level positions decreases in favor of efficiency gains closer to the top, where will those in the top come from in ten years?
Rebecca: I don’t know. It’s something that I worry about. It is concerning how quickly things are changing. For a while there were lots of job postings for “prompt engineers,” but now prompt engineering skills are expected of everybody. Some companies have even developed “prompt coaching” tools or “prompt libraries” to fill this role. This trend toward automation isn’t unique to AI either, and we all need to be mindful of its consequences. For example, I personally try to do most of my shopping in person. I want people to have more job opportunities than just working in a fulfillment center or driving a delivery truck.
Question: Do you personally feel like you are losing any cognitive skills by delegating tasks to AI?
Rebecca: Not yet. I view it as a partner for discrete tasks. For example, if I am frustrated by a situation, I might use AI to help “soften the tone” of an email to ensure I remain professional. It helps prevent me from sending a crummy communication when I’m under pressure. This can also save time in the long run, especially if the recipient might have otherwise escalated and complained to my supervisor.
Question: How are risk-averse clients, like lawyers, reacting to this transition?
Rebecca: Lawyers are naturally risk-averse, and many initially banned AI entirely. However, with more familiarity and the use of RAG (Retrieval-Augmented Generation)—the practice of directly connecting AI tools to trusted source documents—clients are becoming much more comfortable.
Question: Has the massive investment in AI actually proven to be profitable?
Rebecca: I don’t know an absolute answer in terms of expenditures and revenue, but we are seeing very high adoption rates for our AI-enabled research platforms. Customers value these capabilities enough that they are willing to convert to higher-tier, AI-driven services.
Question: If you were calling the shots in law school pedagogy, to what extent would you have law students learn how to integrate AI into their work?
Rebecca: I would make it mandatory. It is the only way to ensure students learn to use these tools ethically and responsibly. You have to use these tools to understand what they are good at and, more importantly, what they are not good at.
Question: How is your company hedging against the possibility that the current AI “hype” might eventually collapse?
Rebecca: We are building our own legal-specific LLM. By focusing on internal, trusted sources rather than the entire internet, we are trying to improve accuracy and reduce reliance on external providers. Even if the broader “hype” dies down, the practical, high-value applications that work will remain.
Ditsch concluded by sharing her appreciation for the chance to hear how Carleton community members were thinking about AI. She conveyed she found it interesting to hear what people are grappling with, where they’re having successes, and how students are talking about these tools and capabilities.