Capital Sciences recently concluded its summer internship.  Our program selects 1 to 2 individuals per summer and focuses heavily on helping them both take ownership of work and develop the skills necessary to be a good engineer whether it’s software or hardware.  This year, our intern was a rising junior – software focused – who’s experience was primarily limited to python coding.   He produced two applications, one written as a script that creates a report and supporting files in the requested formats and the second app was a data pipeline monitoring tool written as a full stack application, both projects helped expand his knowledge drastically.
                These two tools were chosen because they were necessary applications for current customer projects that were easily compartmentalized and scoped to an intern’s ability with support from senior personnel.   The question we faced was whether to provide our interns with access to the AI toolset that our company recently started using.  The use of AI was experimental for the company at the start of the internship, so only key personnel had access to it early on.  The project he started with was smaller and primarily leveraged python.  It was more compartmentalized and the team of mentors were intimately familiar with the scope and framework.  The second project was more complex and utilized a framework which, while it was the best for the job, wasn’t one that the mentors had a extensive experience with.  Along with the remainder of the CapSci team, access was provided to the AI tools in mid-summer, which coincided with the intern’s transition to the second project.

What seemed to work well:

• Continue to utilize task management software and break down our work into manageable tasks that were under 16 hours of work ideally.  (most schools don’t seem to cover using collaborative tasking tools)
• Utilize the same tech we use, task-based pull request for any update to the code (most interns we’ve worked with have limited to no experience with using git at the complexity needed for bigger projects)
• Git branching per task and demonstrating how trackability is important, even on small projects like this (even more important to teach how to resolve merge conflicts, using stash/pop, branch naming conventions, etc.)
• Have senior personnel conduct subject-101 intros to any subject that was new to the intern to help frame it out for him so that he could learn the structure before asking anything of AI
• Focus on good engineering skillset development such as good UX, understanding requirements, big picture thinking, thinking through edge cases, and testing your own changes before sending for review.
• Encouraging them to use AI as a secondary tool. e.g. Asking AI for suggestions on how to approach something, asking it to help scan the code for bugs, asking for suggestions on very complex solutions but in a planning mode, not implementation.
• Senior personnel were able to utilize it to find bugs as well as suggestions for improvements which helped them create new tasks if the backlog got low (development was conducted in a very agile format).
• Senior personnel could utilize it to help gain knowledge quickly and narrow the gap between what they knew, and the nuances of the framework selected.
• The same thing we told the whole team.  “Anything that leaves your computer is still your responsibility.”

What could go wrong with having an intern using AI and how to mitigate?

• Any of the same data security concerns which exist for our team in general
  • Giving the intern mock databases instead of real data access alleviated this.
• The intern could utilize AI too much and it be a crutch that hinders their learning
  • Having frank conversations about why they should try to solve problems without so they may understand the task better.
• Why a window of time with no AI might be necessary:
  • If we had let the intern utilize AI from the start, we would not have had the same insight into his existing abilities or shortcomings in his current skills.  This window helps shape how we focus an intern’s support through their internship.

We feel sharing this experience can help others think about how to adapt internship programs to this evolving technology.  While no one wants to see jobs disappear, the reality is that some of the ‘grunt’ work is going to be offloaded to AI.  What’s irreplaceable in this industry long term are the skilled engineers who can utilize AI as a tool to complete the same soundly engineered software we all want to make.