At a 140-person company with a two-person People team, hiring a mid-level engineer takes 34 days from first interview to signed offer, and eleven of those days disappear inside handoffs nobody owns: the recruiter screens the candidate, then waits for the hiring manager's interview slot; the hiring manager submits feedback, then waits for someone to schedule the panel; the panel wraps, then waits for someone to draft the offer. Multiply that by five open reqs running in parallel and the pipeline runs on whoever remembers to nudge Slack that day. Onboarding is worse: ramp time for a new hire ranges from two weeks to six depending entirely on which manager they land under, because there is no standard 30-60-90 plan, just a Google Doc from 2023 that half the managers have never opened. New hires who land under the wrong manager quietly disengage in week three and nobody notices until the exit interview. Performance review season is the two worst weeks of the year: a shared spreadsheet, calibration meetings that eat four hours per department across six departments, and a completion rate that limped to 62% last cycle even after the deadline got pushed twice. None of this is a headcount problem. It is a process design problem, and it is exactly what real ai agents for hr teams should fix, not by automating the empathetic parts of the job away, but by giving a stretched People team the structured process design it never has bandwidth to build from scratch.
Why ChatGPT and Cursor don't solve this
Ask ChatGPT or Claude.ai for a new-hire onboarding plan and you get a generic 30-60-90 template inside thirty seconds, the same template it would give a hospital, a law firm, or a five-person startup. It has no idea that Loopwell's engineering onboarding needs a staging environment walkthrough by day three, that sales onboarding needs shadow calls by week two, or that the current Google Doc template has a broken link to the benefits portal that three new hires have already hit. It cannot see the applicant tracking system, the calendar of open reqs, or last cycle's review completion data. It produces a document. It does not produce a process, and it has no way to check whether the process it suggested actually got followed the next time a manager onboards someone.
Cursor and GitHub Copilot are not even in this conversation, and that mismatch is worth naming directly. They are autocomplete layers built for people writing code, tuned to finish a function or suggest the next line inside an editor. An HR team's hardest problems, an inconsistent onboarding ramp, a hiring pipeline with unowned handoffs, a review cycle nobody trusts, have nothing to do with code completion. If your People team's actual bottleneck is that no one owns the moment a candidate moves from panel to offer, an editor autocomplete tool cannot help you, no matter how good its suggestions are. The tools solving code problems and the tools solving process design problems are not the same category, and pretending they overlap is how a rollout stalls before it starts.
Meet Folk, and where Keep and Brace fit in
Folk is the People engineer on the Tonone team: it designs org structure, builds hiring pipelines, drafts comp frameworks, redesigns onboarding, and plans performance cycles, treating each of those as a process design problem with real artifacts, not a one-shot document generator. Where a generalist chatbot answers the question you typed, Folk reads the actual state of your People operations, current headcount, open req list, review completion history, and produces a plan that accounts for what is actually broken, not a template for a company that doesn't exist.
Tonone's Folk redesigns hiring pipelines and onboarding playbooks around the actual handoffs a People team is running today, not a generic HR template.
folk-onboard: a ramp plan with owners, not just a checklist
The folk-onboard skill drafts role-based 30-60-90 tracks, one for engineering, one for sales, one for support, each with day-by-day tasks, an assigned owner for every task (not just 'HR'), a buddy-program structure, and a manager readiness checklist that has to be signed off before a req is allowed to make an offer. The output is designed to be tracked, not just read once and filed away.
folk-hire: closing the gaps between pipeline stages
The folk-hire skill maps every stage of the current hiring pipeline, recruiter screen, hiring manager interview, panel, reference check, offer draft, offer approval, and assigns an explicit owner and a target turnaround to each handoff, so a candidate never sits for three days because nobody realized the ball was in their court. It flags the stages where history shows the biggest drop-offs and redesigns those specifically, rather than rewriting the whole pipeline from nothing.
folk-perf: a review cycle people actually finish
The folk-perf skill redesigns the review cadence itself, calibration structure, rating scale, manager prep steps, and a completion-tracking mechanism that surfaces who is behind before the deadline, not after. For a People team running review season out of a shared spreadsheet, this is the difference between chasing completions in week two and knowing exactly which four managers are at risk by day five.
Two of Folk's related agents carry over disciplines from other parts of the business that map onto People ops more directly than it first looks. Keep spends its time scoring customer account health so a Customer Success team knows which accounts are at risk before they churn. That same weighted-signal approach, missed check-in, incomplete task, no manager 1:1 logged, applies just as well to flagging a new hire's disengagement risk in the first 90 days, before it turns into a resignation nobody saw coming. Brace spends its time designing SLAs and escalation paths for support ticket queues. Most People teams are quietly running a ticket queue of their own, benefits questions, leave-of-absence requests, equipment requests, offboarding checklist items, without ever calling it one or giving it a response-time target. Brace's SLA and escalation design work applies directly.
A worked example: Loopwell's onboarding redesign
Loopwell is a 140-person B2B SaaS company hiring roughly 15 people a quarter, run by a People team of two: Priya, Head of People, and Marcus, People Ops Coordinator. Before bringing in Folk, their onboarding ramp time averaged 21 days to first productive contribution and swung as high as six weeks depending on the manager. Only 40% of new hires had a completed 30-60-90 plan on file by day 30. The hiring pipeline averaged 34 days from first interview to signed offer, with three specific handoff gaps averaging 2.5 days each. Priya handed Folk the actual state: current onboarding doc, the open req list, and last quarter's ramp-time data by department.
Folk, Onboarding + Pipeline Redesign, Loopwell
Input: current onboarding doc (2023), 5 open reqs, Q1 ramp-time data by dept.
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
HIRING PIPELINE, before -> after
Recruiter screen -> HM interview gap: 2.5d -> 0.5d (auto-notify on feedback submit)
HM interview -> Panel scheduled gap: 3d -> 1d (owner: recruiter, 24h SLA)
Panel -> Offer draft gap: 2.5d -> 0.5d (owner: HM, template pre-filled)
Offer-to-signature total: 34d -> 19d
ONBOARDING PLAYBOOK, role-based 30-60-90
Day 1: Manager readiness checklist signed off (blocks offer if incomplete)
Day 1-3: Role-specific environment/tooling setup, owner: eng lead or sales ops
Day 5: Buddy assigned, first 1:1 logged
Day 30: 30-day check-in, task completion vs plan reviewed
Day 60: Peer feedback collected, risk flag if any milestone missed
Day 90: Ramp complete, manager sign-off
METRICS
Ramp time to first productive contribution: 21d avg -> 6d avg (engineering)
30-60-90 plan completion by day 30: 40% -> 94%
Review-cycle completion rate (next cycle target): 62% -> 90%+
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Next: pilot on the 2 open engineering reqs, expand to sales/support next quarter.None of these numbers came from a template. The 2.5-day gap between panel and offer draft showed up because Folk read the actual req history Priya provided and found the pattern: offer drafts consistently stalled until the hiring manager was reminded by hand. The fix was not a new tool, it was assigning an explicit owner and a 24-hour SLA to a handoff that had never had either. Marcus rolled the new onboarding playbook out to the two open engineering reqs first, tracked the day-30 completion rate directly, and expanded it to sales and support the following quarter once the pattern held. By the second quarter, Priya was pulling the same 30-day completion number into her board update as a People metric, something she had never had a reliable number for before, because the plan itself now generated the data instead of relying on managers to self-report. The performance review redesign followed the same logic: instead of a spreadsheet everyone filled out under deadline pressure, folk-perf built a calibration schedule with per-department checkpoints two weeks before the deadline, so Priya could see which four managers were behind while there was still time to intervene, rather than finding out on the due date.
Tonone's Folk turns a two-person People team's process gaps into a tracked playbook with named owners, not another document nobody opens after week one.
Folk vs the alternatives
The comparison below is specific to what a People team actually needs when hiring, onboarding, and reviews are the bottleneck, not a general AI capability comparison.
| Capability | Tonone | Generalist chatbot | Cursor / Copilot |
|---|---|---|---|
| Redesigns onboarding as a role-based 30/60/90 playbook with owners | Yes, via folk-onboard, grounded in your actual roles and current docs | No, gives a generic template with no org-specific detail | No, not a process design problem code autocomplete addresses |
| Maps hiring pipeline stages and assigns owners/SLAs to handoffs | Yes, via folk-hire, using your actual req history to find gaps | No, describes pipeline stages in the abstract | No |
| Redesigns performance review cadence and calibration process | Yes, via folk-perf, with a completion-tracking mechanism | No, suggests generic rating scales | No |
| Applies SLA/escalation discipline to HR request queues | Yes, Brace's brace-sla design work applies directly to People ops tickets | No, doesn't treat HR requests as a queue with SLAs | No |
| Flags new-hire disengagement risk before it becomes attrition | Yes, adapting Keep's keep-health scoring approach to onboarding signals | No, has no ongoing visibility into your onboarding data | No |
| Plans comp frameworks and org design changes | Yes, via folk-comp and folk-org | Generic advice, not grounded in your actual bands or structure | No |
Tonone's Folk designs the org, builds the hiring pipeline, and plans the onboarding ramp, treating People ops as a process design discipline instead of a document-generation task.
If your onboarding ramp time depends on which manager a new hire lands under, that is the signal to start with folk-onboard. Hand Folk your current onboarding doc and last quarter's ramp-time data, and it will draft the role-based playbook with owners attached, not another checklist that sits unread after week one.
Install and try
Tonone is free and MIT-licensed. Install it once and all 100 agents, including Folk, Keep, and Brace, are available in your Claude Code session. You pay only for the Claude Code token usage during the work itself.
1. Add to marketplace
2. Install Folk
Frequently asked questions
What does Tonone's Folk agent do for HR teams?+
Folk is the People engineer of the Tonone team. It redesigns hiring pipelines by assigning owners and SLAs to handoffs, drafts role-based 30-60-90 onboarding playbooks, designs performance review cadence and calibration processes, and builds comp frameworks and org structure, all grounded in your actual People ops data rather than generic templates.
How is Folk different from asking ChatGPT for an onboarding plan?+
ChatGPT produces a generic 30-60-90 template with no visibility into your org's actual roles, current docs, or ramp-time history. Folk reads your actual onboarding materials and req data, then designs a playbook with named owners per task and a way to track whether it is actually being followed.
Can an AI agent fix a hiring pipeline with slow handoffs?+
Yes. Tonone's Folk uses the folk-hire skill to map every stage of your hiring pipeline, from recruiter screen to offer approval, and assigns an explicit owner and target turnaround to each handoff, closing the gaps where candidates currently sit waiting for someone to notice.
What is the folk-onboard skill?+
folk-onboard drafts role-based 30-60-90 onboarding tracks with day-by-day tasks, a named owner for every task, a buddy-program structure, and a manager readiness checklist, built from your actual onboarding materials rather than a generic HR template.
How does Brace apply to HR teams if it's a support agent?+
Most People teams run a request queue, benefits questions, leave requests, offboarding checklist items, without treating it as one. Brace's SLA and escalation-path design work, normally applied to customer support tickets, maps directly onto that HR request queue, giving it response-time targets and clear escalation rules.
How does Keep apply to flagging new-hire risk?+
Keep scores customer account health using weighted signals to flag churn risk before it happens. That same approach, applied to onboarding milestones like missed check-ins or incomplete tasks, flags new-hire disengagement risk in the first 90 days before it becomes a resignation.
Is Tonone free to use for HR and People ops work?+
Yes. Tonone is MIT-licensed and free to install. Folk, Keep, Brace, and the rest of the 100-agent roster are all available in your Claude Code session. You only pay for the Claude Code token usage during the actual work.
What results can a small People team expect from redesigning onboarding this way?+
In a worked example at a 140-person company, redesigning onboarding with Folk cut ramp time to first productive contribution from a 21-day average to 6 days and raised 30-day plan completion from 40% to 94%, while closing 8 days out of a 34-day hiring pipeline through explicit handoff ownership.