Skip to main content
DoneThat

AI Adoption GuideHROnboard

30-60-90 plan generator

LLM tailors ramp goals to role, level, and manager input from a structured template.

HR processPlanSourceSelectHireOnboardDevelopRewardExit

By Don, DoneThat’s AI coach · updated

What a usable draft contains

A usable 30-60-90 draft is a proposed ramp plan in which every filled goal cites the role template, the manager's input, or both. If those sources are silent on a topic, the field stays empty. The model does not complete the page with a familiar KPI, a line borrowed from another role, or a generic ramp slogan.

Citation is the quality bar, not coverage. Three cited goals and two blanks beat a full 90-day grid with no sources. The manager still owns the plan. The draft is not assigned until they accept, rewrite, or drop each line.

Hiring managers and onboarding leads use this when a role template already exists (level, competencies, windowed outcomes, required activities) and manager notes exist or can be collected. Lattice, Workday, and 15Five can hold the live performance record later. Greenhouse can hold the hire. None of those systems authorizes a guess in an empty cell.

Load the template and manager notes first

Do not generate from a job title. Load two artifacts, then draft.

Pass the template as structured fields, not as a pasted job description. Include role title, level, function, each window's outcomes, and required activities as separate items. Pass manager input as dated notes with named priorities and constraints. If a field is absent, send it empty. Do not substitute a library example for a silent field.

The role template is the structured source: function, level, competencies, day-30 / day-60 / day-90 outcomes the company already uses, and required activities (shadowing, certifications, tool access). If the template says observe a discovery call with a senior AE at day 30, that line may appear in the draft. If it never mentions pipeline coverage, pipeline coverage stays out unless the manager wrote it.

Manager input is the second source: this person, this team, this quarter. Useful notes name the book of work, the buddy, surfaces they should not touch yet, and what good looks like in this seat. Collect those notes in a stable shape with a manager briefing generator (priorities, constraints, people to meet) rather than a loose paragraph.

If the template is missing, stop. If manager input is missing, draft only template-backed lines and leave every personalized field blank. Do not infer level from the title when the template has no level. Do not add a weekly forecast review because you assume the role needs one.

Map fields you can actually read from Greenhouse, Workday, Lattice, or 15Five. Do not copy a number from a vendor screen into the draft unless that number is also in the template or the manager notes.

Cite every line, or leave the field blank

Each drafted goal, activity, or checkpoint carries a cite a reviewer can check: Role template, Day 30 outcomes, complete product certification, or Manager notes, 12 Aug, shadow two enterprise renewals. A goal with no template cite and no manager cite is a failure mode. Delete it, or replace it with a blank and the reason (no template row; manager silent).

Draft in this order. Parse the template into windows (0-30, 31-60, 61-90) and types (learn, do, demonstrate). Overlay manager notes: they may add, narrow, or defer a template item. They may not create a KPI both sources omit. Write a candidate line only where at least one source supports the substance. Put the cite on the same line or in a source column. Where a template heading has no content, output an empty cell.

Empty stays empty. TBD plus a made-up target is still invention. A blank labeled quota / attainment: not specified in template or manager notes is the honest output.

The no-cite failure often looks polished: Achieve 50 percent of ramp quota by day 60. If neither source named a quota or a percentage, the line is invalid. The same rule applies to an NPS target or a time-to-first-ticket figure that never appeared.

A role-based learning auto-assignment path can sit beside the plan. Course enrollment is a 30-60-90 goal only when the template or the manager named that course as a ramp outcome.

Example: AE-2 with incomplete manager notes

The AE-2 template lists, for days 1-30: complete CRM navigation training, attend four live discovery calls, pass the internal product cert. For days 31-60: own two inbound demos with a manager on the call, write a territory map. For days 61-90: run a full sales cycle on a named inbound lead, with manager review at proposal. It does not state a quota, win-rate, or pipeline multiple.

The manager notes say the hire sits on mid-market, must not take enterprise accounts in the first 60 days, and should spend week one with Jordan on renewals. They do not add demo volume, a KPI, or a cert deadline beyond the template.

A valid draft cites: CRM training and product cert (template, days 1-30); four discovery-call observations (template); week-one renewals with Jordan (manager notes); no enterprise accounts before day 60 (manager notes, constraint); two inbound demos with manager present (template, days 31-60); territory map (template); one full cycle on a named inbound lead with proposal review (template, days 61-90).

A valid draft leaves blank: quota, ramp percentage, extra activity volume, and any forecast-accuracy target. No source.

An invalid draft adds hit 30 percent of annual quota by day 90 because that is a familiar AE shape. That invents a KPI. It is also invalid to write those lines into Lattice or 15Five as assigned goals, or to send them to the new hire as the plan, before the manager edits and owns them.

If the new hire asks a new-hire conversational assistant what they must hit in 60 days, the assistant reads the manager-owned plan, not the unreviewed draft. An ai roleplay skill coach for discovery belongs on the plan only if the template or manager named that practice as a ramp activity.

Assignment stays with the manager

Generating a draft is not assigning a goal. Auto-assign is a failure mode. The manager decides what is in, out, or delayed, and whether any line becomes a tracked goal.

Deliver the draft with cites and blanks visible. Do not hide empty cells. Require accept, edit, or reject per line, or one I own this version action that records the manager as author of the live plan. Only then sync anything to Lattice, Workday, 15Five, or an onboarding checklist. Leave Greenhouse as the hiring record; do not write ramp goals onto the offer. Tell the new hire the plan is manager-owned. A generator document is not their scorecard.

Map only accepted lines into vendor goal fields. Do not bulk-create objectives from uncited draft text. If a form requires a metric, leave it empty or mark not specified. Do not type a number to satisfy the form.

The manager may add a KPI. That is authorship. If they add pipeline coverage of 3x on inbound by day 90, the live plan cites them, with a date. The generator stopped at the blank.

Treat the draft as assigned too early and you get calendar holds, learning enrollments, and performance objectives firing on generate. Pause that automation until ownership is recorded. After the checklist (every filled cell has a source; no invented KPI; still draft), the manager edits. That edit is the work. The generator only assembled the template and the notes onto one page.

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