Twenty-six applicants before coffee
That’s an ordinary morning on a busy req, plus eleven unread candidate emails and a hiring manager wanting “a quick pipeline update” by noon. Most of the AI tools for recruiters we’ve tried want to write your InMails. I keep thinking they picked the wrong ten minutes of the day. The other hours go to the ATS and the inbox, and to a side spreadsheet that exists only because the ATS report never quite shows the hiring manager what they asked for.
dassi is our attempt at those hours. It works inside the browser you already use, signed in to your ATS and Gmail as you, and you teach it a job the way you’d teach a temp: say it in plain English, or do it once while it watches. It keeps that as a workflow and runs it when you ask or on a schedule, with whichever model you pick, including a local one if résumés shouldn’t go to a cloud provider.
One rule I’d set before anything else: nothing goes to a candidate and nobody changes stage until you approve.
The unclear pile
Screening usually goes first. On weekday mornings dassi opens Greenhouse (or whatever ATS you use in a tab), reads each new application against the scorecard you wrote, and leaves a note on the profile quoting the résumé lines behind its score. Strong ones get a phone-screen invitation drafted. The example run on our screening page also has two applicants it declined to score, a scanned PDF it couldn’t read reliably and a “5+ years” that was really four plus an apprenticeship.
I’d almost rather it produced only that pile. You’d have called the obvious yeses and nos correctly yourself, tired or not.
To score a résumé the model has to read it, so the text goes to the AI provider you picked (via dassi’s relay if you’re on included credits), and your candidate privacy notice should mention that. Some places regulate AI in hiring, New York City (Local Law 144) among them. I don’t know which apply to you. Check before a score feeds a decision.
Forty scheduling emails and a Friday spreadsheet
“Thursday works, or Friday after 2” needs a calendar check every time, and dassi can triage the inbox each morning so those replies are drafted before you open it. Agency recruiters will probably want LinkedIn exchanges logged to the candidate record, with notice period and salary expectations pulled out. That’s about as far into LinkedIn as I’d take it. The site caps weekly invitations and its user agreement restricts automated activity, so dassi reads there at a person’s pace and never sends connection requests in bulk.
In-house, the backlog is more likely offer paperwork to re-key into the HRIS, a job that stops on any row failing validation and names the field. The Friday pipeline count for the hiring managers’ sheet can join the rest of your recurring admin on a schedule.
Monday to Thursday on one req
Pick one role with a steady flow of applicants. On Monday, write the must-haves and nice-to-haves as you’d say them to the hiring panel, then go through four applicants while dassi watches and explain each call. Tuesday you run it by hand and read every note, which is often where you find out “senior” meant three different things to you. Rewrite that line of the scorecard and save the workflow.
By Thursday it can run at 8am on weekdays, in your own browser, which therefore has to be open at eight. If the ATS logged you out overnight, dassi signs in with whatever Chrome’s password manager fills in and hands you the two-factor prompt.
Paying for it
Free credits to try dassi, good for 7 days. No card required. After that it’s $10 a month on the Developer plan with your own API key, paying the model provider for usage, or $25 to $300 a month on a plan with credits included. Thirty applicants cost more than one scheduling reply, and a big model more than a small one, but how much more on your ATS is something the trial week will tell you better than I can.