Markel AI Automation Intern
Underwriters, claims handlers and operations teams at Markel spend hours on work a computer can do well: reading broker emails, keying the same data into two systems, checking documents against rules and rebuilding quote letters by hand. We build working tools that take that work off their plate—AI-assisted underwriting workbenches, automated email flows and links between the systems people already use.
We move quickly and we build for real people. A typical project goes from a conversation with a business leader to a working proof of concept in days. Then we test it, harden it and hand it over. You'll join as the second person on this work, so you'll learn fast and see your work in use early. Along the way you'll learn how specialty insurance runs and how to put AI to work safely inside a regulated business.
What you'll do
– Build and improve proof-of-concept tools, including browser-based workbenches that underwriters use to read submissions, price risks and produce quotes.
– Connect systems through APIs: send requests, handle sign-in and tokens, map the data that comes back and deal with errors gracefully.
– Write Python scripts that turn spreadsheets, PDFs and Word documents into clean data—and turn data back into finished documents.
– Build and maintain Power Automate flows, SharePoint lists and Copilot Studio agents that triage email, review documents and send reminders.
– Put large language models to work: write the prompts and extraction schemas, then build test sets that show whether the AI got it right and what it cost.
– Test before anyone relies on your work, using automated browser checks, unit tests and side-by-side checks against the source data.
– Deploy tools on our cloud data platform, Databricks, while keeping access, secrets and cost under control.
– Write the documentation that lets someone else pick up a project: setup notes, runbooks, user guides and change logs.
– Join meetings with underwriters and business leaders, capture what they need and turn it into clear requirements.
What you'll bring
– A bachelor's degree
– Working skill in at least one programming language, ideally Python or JavaScript, shown through coursework, an internship or your own projects.
– A basic understanding of how web APIs work, including HTTP requests, JSON and authentication.
– Strong Excel skills (lookups, pivot tables, formulas) and comfort working with large data sets.
– Clear writing. You can explain what you built and why to someone who doesn't code.
– Curiosity about AI, plus the habit of checking an answer instead of trusting it.
– Good judgment with data. You'll handle customer and broker information, so you ask before you share or send
– Business acumen