Skip to main content
DoneThat

AI Adoption GuideHROnboard

Manager briefing generator

LLM produces a one-pager on the new hire background and suggested first-meeting agenda from public profile and resume.

HR processPlanSourceSelectHireOnboardDevelopRewardExit

By Don, DoneThat’s AI coach · updated

What belongs on the one-pager

A manager briefing is a one-page sheet for the person who will run the first 1:1. It has two jobs: restate the new hire's background only as the licensed resume or ATS packet states it, and propose a first-meeting agenda the manager can edit. Quality means every background sentence cites a source in that packet. If the resume and ATS fields are silent, those lines stay blank. The manager still runs the meeting.

Do not treat the sheet as a biography, a personality sketch, or a public-web dossier. Fluency is not evidence. A paragraph that reads well but points at no resume line, no application field, and no recruiter note you are allowed to reuse is a failed briefing, even if it sounds like the candidate.

Keep the page short enough to read before the calendar alert. Typical blocks: identity and role as stored in the ATS (name, start date, title, team), cited background bullets, open questions the packet actually raises, and a suggested agenda. Skip praise, culture-fit adjectives, and "likely motivations" unless a licensed note already says them.

Scoring a resume for hiring is a different task than briefing the manager after the person has accepted. Use structured resume scoring if you still need a structured readout of the original application.

Load the ATS packet, not a public scrape

Start in the system of record you already license. Applicant and HR suites such as Greenhouse, Lever, Ashby, and Workday hold the resume file, structured application fields, offer and start-date data, and recruiter notes your policy allows the hiring manager to see. Export or API-pull that packet. Do not add a second crawl of LinkedIn, GitHub, news, or personal sites unless counsel has already approved that source as part of the hire file.

Unauthorized public-profile scrape is the usual shortcut and the usual compliance failure. The model will happily merge a similarly named engineer in another city into your new hire's first day. You cannot audit a cite that never entered the ATS.

Minimum packet for a briefing attempt:

  • Resume or CV as stored on the candidate record
  • Role, team, location, start date, and manager as stored for the hire
  • Recruiter or coordinator notes the manager is permitted to use (interview debrief summaries only if your policy includes them post-offer)

If the resume is missing, stop. Generate nothing that pretends to know prior employers, degrees, or skills. An empty briefing with a line that says "resume not on file" is the correct quality outcome. A filled page with no resume is not.

Confirm the file is this person's file. Name collisions, forwarded PDFs, and merged records show up here. If two resumes sit on the profile, cite which file you used or refuse to merge them.

Draft with cites, then leave blanks

Prompt or configure the generator to output claims in a cite-first shape: short bullet, then pointer (resume section and a short quote, or field name and value). Ban narrative glue that has no pointer. Ban "based on their online presence." Ban filling education, employers, or dates from what the model already knows about common careers.

Illustrative pass: a hiring manager opens the packet for an incoming product designer. The resume lists two employers, a degree, and a portfolio URL the candidate put on the application. Recruiter notes visible to the manager say the candidate asked how design critique works on this team. The draft should look like this in spirit:

  • Background: Employer A, dates as printed on the resume; Employer B, dates as printed; degree as printed. Each line tagged to the resume.
  • Packet facts: title, team, start date from the ATS hire record.
  • Agenda suggestion: welcome and working agreements; how critique runs (tied to the recruiter note); tools and first-week access; questions the designer wants to ask.
  • Left blank: publications, management span, visa history, side projects at scale, anything the resume and fields do not state.

What you do not do: invent a third employer because the model recognizes the name; upgrade "contributed to the design system" into "led the design system"; scrape the portfolio to write a taste biography. The portfolio URL can appear as a link the candidate supplied. The model does not summarize that site unless that summary is already in the licensed packet.

After the draft, a human (recruiter coordinator or the manager) scans every background bullet against the PDF or ATS fields. One uncitable line is enough to send the draft back. Do not "fix" silence by writing a plausible sentence.

Silent fields stay empty on purpose. Empty is a signal: ask in the meeting, or live without that fact. Prefill is how invented prior employers get into the first 1:1.

Run the first meeting yourself

The sheet is a prompt for the conversation, not a script to recite and not a substitute for listening. The manager still opens the meeting, sets tone, and decides what to skip. If the agenda puts too much weight on resume archaeology, cut it. If the new hire's first need is tools and teammates, start there even if the draft led with career history.

Use cited bullets as optional icebreakers ("Your resume lists a stretch on checkout redesign: what do you want to carry forward here?") and as a check against talking about work the person never claimed. Do not read the briefing aloud as if it were their bio. People notice when a manager recites a generated life story.

Take notes in the meeting that update reality. The briefing does not get write-back authority into the ATS as biography. If the hire corrects a date or a title, believe the person in the room, fix the operational record through your normal HR process, and do not treat the old draft as ground truth next week.

Downstream onboarding should stay task-specific. A 30-60-90 plan generator can turn role expectations into a plan after you have met, not from an invented career arc. A new-hire conversational assistant answers the employee's process questions; it is not a second copy of the manager's private briefing. Role-based learning auto-assignment can attach curricula from the job profile. None of those tools should ingest an uncited briefing as if it were verified history.

Reject drafts that invent a career

Three failure modes show up even when the layout looks finished.

A briefing with no resume cite. The page has a Background section written in confident prose and either no sources or sources like "public information" and "typical path for this role." Delete it. Without a pointer into the licensed packet, you cannot tell hallucination from memory of some other candidate. Quality here is cite coverage on background claims, not completeness of the story.

Treating the draft as a bio. Someone pastes the one-pager into an intro email, a Slack welcome, or the team wiki. That launders guesswork into the company narrative. The briefing is for the manager's first meeting. Introductions should use what the new hire provided for sharing, or what they say in the room.

Inventing a prior employer (or school, title, or date). This is the sharp version of the same bug: a missing resume line, a common company name, or a model prior, and suddenly the manager congratulates someone on a job they never had. The fix is mechanical. Every employer bullet must quote or field-match the resume or ATS. If the resume lists two jobs, the briefing lists two jobs. It does not add a third "and likely" stint. It does not normalize a consultancy client into full-time employment.

Operational checks before the meeting:

  • Every background bullet has a packet pointer a human can open in under a minute.
  • Blanks are visible, not filled with hedges.
  • No public-web enrichment step ran outside the licensed file.
  • The agenda is labeled suggested. The manager can strike items.
  • The document is not copied into employee-facing channels as biography.

If those checks fail, you do not have a briefing generator. Put the packet in, cite it, leave gaps, and let the manager meet the person.

Is this worth automating for you?

Whether this pays back depends on how much time it takes your team today. Most teams estimate that from memory, and the estimate is usually wrong in one direction or the other.

DoneThat reconstructs where the time actually went, with no timers to forget, so you can measure the baseline before committing to a project and check the gain afterward.

Measure the baseline first