Ushiro Labs Get the two ideas

Back to home

From days of chasing feedback to a ranked list in minutes

A Chicago emergency veterinary hospital needed to choose its interns. Here’s what we built and how it worked.

The situation

Every intern class meant the same scramble. More than 40 applicants, and more than 10 physicians who each had a view on every one of them.

The person running the program chased feedback by email, read every comment and wrote summaries by hand, all on top of a full clinical schedule. In the end, the decision leaned on impressions more than on data.

What the hospital asked for

  • Real data in the decision, not just impressions
  • A setup that works for every future class
  • As much of the work done for them as possible
  • No more hand-written summaries
  • Something simple for busy, non-technical physicians to use

What we built

A shared spreadsheet, because everyone already knows how to use one. The clever parts run quietly behind it.

  1. One sheet per physician

    Each physician got their own sheet, with instructions on the first page. For every applicant they pick a score from a dropdown and add comments. They can edit only their own sheet.

  2. Everything in one place

    A background sheet gathers every score and comment, so nothing gets copied by hand.

  3. Comments summarized

    A script reads all the comments for each applicant and writes a short summary.

    The summarizing ran on an AI model hosted on our own machine, so applicant comments were never sent to an outside AI company.

  4. A ranked list

    The scores are averaged, then combined with the summaries into one ranking. The final sheet lists applicants from strongest to weakest, color coded and ready to discuss.

Example

  1. 1
    Applicant A, veterinary technicianStrong comments from every interviewer. Wants more surgical exposure.
  2. 2
    Applicant B, veterinary technicianGreat with clients. Needs a refresher on dental work.
  3. 3
    Applicant C, veterinary technicianSolid skills. Interviewers were split on schedule fit.

The result

A multi-day manual process now takes minutes.

The program lead reviews a ranked list with a short summary of what the physicians said. The same setup works for every future class.

By hand: more than 10 days

With the automation: 7 minutes

Days are counted from the last physician feedback to a finished ranking. Automation time is run time.

Not to scale.

  1. 13 physicians scored the applicants

  2. More than 350 written comments summarized

  3. 41 applicants ranked

What it could look like for you

The same problem exists anywhere several people judge the same decision.

In a general practice the pile is smaller, but the pattern is the same. Say you’re hiring a veterinary technician. The owner, the practice manager and a couple of senior technicians each meet the finalists, and each has an opinion.

How it often goes

  • Opinions arrive by text, email and hallway chat
  • Someone sums them up from memory
  • The decision can lean on whoever spoke last

With a ranked list

  • Everyone scores and comments in one simple sheet
  • The comments are summarized for you
  • You see a shortlist, with the reasons beside each name

Want the two things we would tackle first?

Tell us about your hospital. We’ll look at your practice and reply with two concrete ideas, no obligation.

Prefer email? Write to hello@ushirolabs.com.

We’ll use your details only to reply to you.