← Back to home Case study

One AI-native system for a hiring firm's HR and ops.

How we turned work that depended on people remembering into a system that runs on its own, without asking anyone to learn a new tool.

Hiring firm3 business units~120 people2 Google Workspaces~35 hrs/week saved (est.)
01What they came with

Everything worked, as long as someone remembered.

New hires learned the job by asking around.Onboarding lived in HR's head and a few shared sheets.
HR answered the same leave and policy questions every day.Across three business units and two separate Google Workspaces.
Leadership had no regular way to hear from 120 people.Problems surfaced at exit interviews.
Recruiters downloaded résumés that never reached the database.Most sat in personal folders.
Clients paid late, and nobody owned the follow-up.Promised dates got lost in email threads.
Founders chased managers for updates.There was no single view of what needed attention.
02How we worked it out

We mapped how the work gets done, then taught agents to do it.

Map

We sat with the founders, managers, HR and staff and wrote down how each task really gets done, step by step.

Train

Agents learned those workflows, plus the rules, exceptions and past decisions the team had built up over time.

Act

We built systems that use that knowledge to do the work on people's behalf, one workflow at a time.

Decide

People talk to their own agent in Slack, WhatsApp or email, and only handle the calls that need them.

03The system

One knowledge core. Systems that act on it. People who just decide.

1 · Map how work gets done2 · Train agents on it3 · Systems act on their behalfFounderreports, approvals, updatesManagersstatus chasing, leave, reviewsHRquestions, onboarding, check-insEmployeesleave, payslips, daily tasksThe knowledge coreEvery workflow, step by stephow each task is really done todayWhat the team has learnedrules, exceptions, past decisionsWho can see and do whataccess by role and teamAnswers HR and policy QsRuns new-joiner onboardingReads quarterly check-insChases status updatesSends payment remindersPrepares hiring approvals4 · People handle the higher-level calls, by chatFounderdecides, approves, asks anythingManagersunblock the team, review workHRonly the cases that need a personEmployeesask, request, get it doneIn Slack, WhatsApp or email, or from Claude and ChatGPT through MCPUnderneath · tools are just data and connectorsData and actionsGmailWhatsAppCalendarSheets and DriveBuilt withn8nApps ScriptSupabaseVercelClaude
04How it's built

Designed so the work runs itself, and nobody's day gets harder.

AI-native

AI agents answer, chase, sort and summarise. The system acts first instead of waiting for someone to click.

People only for exceptions

Disputes, borderline test scores and flagged answers go to a person. Everything routine doesn't.

Nothing new to learn

Staff use WhatsApp, Slack and Gmail as before. New hires were onboarded without a single training session on the tool.

Access follows the org chart

Founders see everything, managers see their team, staff see their own. Answers that should stay anonymous stay anonymous.

Available anywhere, through MCP

The same system can be used from Claude, ChatGPT or any AI tool that supports MCP, so founders and HR can ask questions or take actions from wherever they already work.

05The result

What each piece gives back.

  • HR assistantStaff ask in chat, HR only sees what it can't answer.
    ~6 hrsa week of HR time
  • Payment remindersChases until paid, pauses when a date is promised.
    ~5 hrsa week of accounts time
  • Founder assistantAnswers about any part of the business, from every system.
    ~3 hrsa week of chasing
  • Needs-your-attention digestOne daily list for each manager and HR.
    ~2 hrsa week per manager
  • New joiner onboardingA three-month plan, reminders and a chatbot for every hire.
    ~8 hrsper new hire
  • Training and testOne recorded session became a course, a test and a helper.
    ~10 hrsof senior time per batch
  • Quarterly check-insThemes by team, compared with last quarter.
    2 days → 1 hrper quarter
  • Résumé captureEvery download saved to one searchable place.
    Allrésumés kept, up from a few
~35 hrs a weekEstimated across the team, based on its size and how the work was done before.

Tell us one thing that's eating your team's time.

A free 30-minute call. You leave with a plan to fix it.

Book a call
nishiljain7024@gmail.com
Or chat with Nishil on WhatsApp