Home · Solutions · HR & people
Solution · HR & peopleThe recruiter reads the CVs; the robot does the register, the reply and the calendar
Every application answered, every interview booked
Applications from every channel are registered, checked against the requirements the advert states, answered the same day and booked into panel slots; who is shortlisted stays a human decision.
Executive summary
Applications wait days for an answer while recruiters match a panel's calendars by email.
Nothing new needs buying.
Every applicant hears from you on the day they apply, which is the one thing candidates in this market notice and compare.
the applicant tracking system; Microsoft Bookings and the panel's calendars; the SharePoint candidate file
Business problem
Talent acquisition
Frontline hiring is a volume business. A group replacing a fifth of its store and warehouse staff every year runs dozens of requisitions at once, and every advert says plainly what the job needs: a category C+E licence valid through probation, a forklift certificate, a two-shift pattern, the right to work, a start date within six weeks. Candidates answer those questions when they apply. Then somebody reads the answers again, one CV at a time.
The reading is the smaller half. Once a recruiter wants to meet someone, the interview has to fit a store manager, a shift lead and sometimes a regional manager, none of whom sits at a desk. The slot is negotiated by email and phone, confirmed, then moved when a delivery runs late. Every candidate who accepts a first date and then a second one costs a recruiter a morning meant for the next requisition.
At scale the two halves collide. In a seasonal peak the intake doubles while the panel's availability does not, applications sit unregistered, and nobody can say on Friday how many people applied that week or how many are still waiting. Candidates draw their own conclusion from the silence.
How it works today
- PersonA coordinator copies each application from the recruitment mailbox into the applicant tracking system and files the CV under the requisition
- WaitingJob-board applications wait there until somebody has time, typically two to four working days
- PersonThe recruiter opens each CV to re-check what the advert already asked: licence, certificates, shift availability, distance, start date
- Risk of errorCandidates who left a question blank are guessed about rather than asked, and most are never told anything
- PersonFor anyone worth meeting, the recruiter proposes three times to the store manager and the shift lead, then relays what survives
- WaitingA slot is agreed after four to six messages, and some move again the day before
- Risk of errorNothing shows how long applications have waited, so an ageing requisition surfaces when the hiring manager escalates
Why the current process costs more than it appears
Behind every exception is an hour nobody logged.
- Chasing is the job, not the exception. Four to six messages per interview, multiplied by three panel members and by everyone who reschedules, is where a recruiter's week goes.
- Silence has a price in this labour market. An applicant who hears nothing for a week has usually accepted another offer, so the cost is the vacancy that stays open, not the unanswered email.
- Requirement checks vary with whoever performs them, so whether a licence expiry date is verified depends on the recruiter, the day and the length of the queue.
Cost of inaction
Applications are not the scarce thing here; answered applications are. A group hiring 640 people a year competes for the same drivers and warehouse staff as every other employer in the region, and the company that replies first is often the one that hires. The arithmetic above prices the administration only, and says nothing about shifts covered by overtime while a requisition ages.
A quieter exposure grows beside it: every month adds a few thousand CVs to a mailbox and a shared drive with no decision recorded and no retention date, and the first serious test of that is a candidate asking what you hold about them and why.
A plausible organisation with realistic proportions. The figures are there to be recalculated on your data; they are not a client result.
A retail and logistics group: 210 stores, three distribution centres and its own driver fleet in two countries, about 4,800 employees, six recruiters and two coordinators, an applicant tracking system with a partial API, Microsoft 365 E3.
2,900 applications a month against roughly 90 open requisitions, from the careers page, three job boards, the recruitment mailbox and forms scanned in stores; 640 hires a year, mostly store, warehouse and driver roles.
Applications are copied into the system by hand, read against the advert and answered when there is time; interviews are arranged by email between the recruiter, the store manager and the shift lead.
About six minutes of administration per application before anyone has judged anything, plus the calendar traffic around every interview and a first response measured in days.
Robots register every application whatever channel it arrived on, check only the requirements the requisition states in writing, acknowledge every candidate the same day, and open a Microsoft Bookings slot across the whole panel once a recruiter has decided to invite someone.
First response falls to the same working day, the calendar traffic disappears, and 290 hours a month return to the recruiters for the part only they can do; every figure there is modelled rather than measured.
Proposed solution
Nothing new needs buying. Applications from the careers form, the job boards, the shared mailbox and the scanner folder are queued in UiPath Orchestrator, matched against existing records so that one person applying to three stores stays one candidate, and written into the applicant tracking system with the CV attached. The check that follows is deliberately narrow: a robot compares the candidate's own declarations with the requirements the requisition states in writing, records what matched and what did not, and asks once, through a Microsoft Forms link in the acknowledgement, for anything left blank.
The robot does not read the CV for meaning, score anybody, sort the list or produce a shortlist. Ranking people is a human judgement with legal consequences, and we do not automate it. The calendar is where the hours come back. Microsoft Bookings holds one service per interview stage with the panel as its staff, so an invited candidate sees only the times free for everyone who must attend, inside the lead-time and buffer rules the team sets, and confirmations, reminders and reschedules are the booking page's own work. The panel gets the CV in their Teams channel the day before, each interviewer completes one Microsoft Form afterwards, and outcome letters leave only when a named recruiter releases the batch in Teams Approvals.
Microsoft Bookings services, staff and scheduling policies through the Microsoft Graph Bookings API; Microsoft Forms; Microsoft Teams Approvals app; UiPath Orchestrator queues, triggers and audit; UiPath Integration Service connectors for Microsoft Outlook 365, Microsoft Teams and Microsoft OneDrive & SharePoint; UiPath Action Center notifications in Teams
The multi-channel intake and deduplication, the requirement rules per requisition family, the acknowledgement and missing-information loop, the booking and reschedule flow, the outcome-release batch and the Power BI report
The applicant tracking system, through its API where one exists and its interface where it does not; the job-board mailboxes and the careers-page form
How the automated process works
- AutomationApplications from the careers form, the job boards, the mailbox and the scanner folder are queued, matched against existing records and registered with the CV attached
- AutomationEach one is compared with the requirements the requisition states in writing: licence and expiry, certificates, declared language level, right-to-work document, site, shift pattern, start date
- AutomationEvery candidate is acknowledged the same day, and anything missing is requested once through a Microsoft Forms link that writes back to the record
- PersonThe recruiter works the review list in Teams, in arrival order, and decides who is invited, held or declined; the automation offers no ranking, no score and no shortlist
- AutomationFor an invitation, Microsoft Bookings offers only slots free across the whole panel; the confirmed slot creates the meeting and puts the CV in the panel's Teams channel the day before
- AutomationReschedules and cancellations made through the link update the panel, the record and the interviewers' calendars
- PersonAfter each stage every interviewer completes one Microsoft Form, filed against the candidate; the panel's conclusion is reached in conversation
- AutomationOutcome letters wait until a named recruiter releases the batch in Teams; Power BI shows response times, ageing requisitions and no-shows by site
Human-in-the-loop model
Automation handles
- Registering every application from every channel, filing the CV and merging duplicates into one candidate record
- Comparing declarations with the requirements the requisition states in writing, and asking once for whatever is missing
- The same-day acknowledgement, the booking link, the reminders, the reschedules and the panel briefing
- The evidence: which requirement was checked against which declaration, and when each message left
People decide
- Who is invited, who is held and who is declined; no ranking, no score and no shortlist comes from the automation
- Whether a stated requirement is met when the answer is unclear or explained in a covering note
- What the interview establishes, and what the panel concludes from it
- Which requirements belong in a requisition at all, and whether each one is lawful and necessary
Before and after
Systems and integrations
Every entry can be checked in vendor documentation. The evidence class is stated next to each one.
Inputs
- careers-page form
- job-board applications in the recruitment shared mailbox
- agency and referral submissions
- paper forms scanned at store level
Automation layer
- UiPath Orchestrator
- UiPath Robots
- UiPath Integration Service
- UiPath Action Center
Target systems
- the applicant tracking system
- Microsoft Bookings and the panel's calendars
- the SharePoint candidate file
- Power BI
Human touchpoints: the recruiter's review list in Teams; Teams Approvals for outcome letters; the interviewer's Microsoft Form after each stage
Technologies used
queue every application, run the requirement rules, write to the tracking system, retry and log
Aread the mailbox, file CVs, post the panel briefing
Athe recruiter's queue: unclear declarations and exceptions
Aa service per interview stage, panel availability, self-booking and reschedules
Aa named recruiter releases each batch of outcome letters before it is sent
Amissing-information requests and per-stage interviewer feedback
Aresponse times, slot fill, no-shows and channel mix per requisition and site
Athe record of every application, check, interview and decision
CIllustrative economic model
The arithmetic is open, so it can be argued with.
Two figures drive everything below: six minutes of administration per application, and €27 an hour fully loaded for a recruitment coordinator in Central Europe. The six minutes covers registering the application, filing the CV, comparing it with the advert, sending an answer and the interview calendar traffic that belongs to it. Reading a CV and judging a person is outside the model, because it is not automated. What comes out is capacity released, not posts removed.
Run the numbers on your data
An illustrative estimate from your own inputs. It models released capacity; it is not a promise of savings.
Business benefits
- Every applicant hears from you on the day they apply, which is the one thing candidates in this market notice and compare
- Interview slots are agreed in a single exchange instead of six, because the candidate picks from times the whole panel is free, and a reschedule is handled by the link rather than by the recruiter
- Requirement checks are applied the same way to every applicant and recorded against the wording of the advert
- A seasonal peak is absorbed by the intake, not by overtime: twice the applications is twice the robot runs, not twice the coordinators
The management view
- Hiring becomes a service with a measured lead time instead of an assumption that recruitment is busy
- The funnel is visible per requisition and site: what arrived, what was answered, what is booked, what waits on a hiring manager
- Checks and outcome letters carry a record, so an audit question or a candidate complaint is answered from the system, not from memory
Board-level KPIs
Security and governance
An auditor should be able to reconstruct every decision.
- The recruitment mailbox is read by an application identity scoped to that one mailbox under Microsoft Graph RBAC for Applications, never by a recruiter's account
- A CV, the declarations attached to it and the hiring decision live in the applicant tracking system rather than inside a robot; what the robot does touch stays in your Microsoft 365 tenant, and UiPath Automation Cloud runs it in its EU region
- No negative outcome is sent by a machine: declines wait until a named recruiter releases the batch in Teams, and the release is recorded with that person and the time
- Candidate files carry Microsoft Purview retention labels set to the period your privacy notice promises, rather than depending on who empties a folder
- Requirement rules are versioned per requisition, so the advert wording, the rule derived from it and the date it changed can be produced on request
Why now
Regulation (EU) 2024/1689 classifies AI systems used for the recruitment or selection of people, in particular to analyse and filter job applications and to evaluate candidates, as high-risk in Annex III, point 4(a). A rule comparing a stated requirement with a declared answer is not that: Recital 12 puts systems based on rules defined solely by natural persons outside the definition of an AI system. Which side of that line your screening sits on is now a board question.
The frontline labour market is decided on speed. This group hires 640 people a year against the same drivers and warehouse staff every other regional employer wants, and the modelled 290 hours a month of administration is time not spent on candidates still available.
The scheduling half needs no development: Microsoft Bookings exposes services, staff and scheduling policies through Microsoft Graph, so a booking page per interview stage is configuration, and only the intake and the write-back are built.
Relevant executive roles
Hiring becomes a measured service: every applicant answered, every requisition with a lead time, and screening that can be explained to a works council
Stores and distribution centres are staffed sooner, because the constraint moves from the recruiter's calendar back to the interview itself
The administrative share of recruitment becomes a planned number instead of a cost absorbed inside an HR headcount
Every candidate record has a decision, a retention label and a named person behind every outcome sent
Common questions and objections
A ranking built on the fields of an application form ranks form-filling, not people. And a system that evaluates candidates enters the high-risk category the EU AI Act describes, with the obligations that follow. We check what the advert states; the judgement, and the accountability for it, stay with your recruiters.
An intake problem, not an intelligence problem. Each board's format is read by a rule written once, matched against existing candidates and written into your system; a seventh board is a rule, not a project.
Today they already are: a mailbox rule and a folder. What a candidate notices is an answer on the day they applied, a booking link offering real times, and a message when something changes. Silence is what feels mechanical.
When this is not the right solution
- You want candidates scored, ranked or shortlisted automatically. We do not build that, and we would advise against buying it: the automation checks only requirements the requisition states in writing, and every judgement about a person, including every decline, is made and released by a named recruiter.
- Fewer than roughly three hundred applications a month across a handful of requisitions, where a disciplined mailbox routine and good templates beat anything we would build.
- Your adverts do not state their requirements, or the applicant tracking system is not where hiring is recorded. Fix the wording and the record first; otherwise there is nothing stable to check against.
A question for the next management meeting
Do we know how many people applied to us last month and never received an answer, and would we be comfortable if that number were published?
Implementation approach
Delivery runs in stages, so it can be stopped at any point.
We deliver
- A channel audit: where applications actually arrive, in what formats, and how many never reach the system
- The requirement rules per requisition family, written from your published adverts and reviewed with HR and any works council
- The intake robots, the deduplication logic and the write-back into your applicant tracking system
- The acknowledgement, missing-information and outcome-release flows, in wording your employer-brand owner approves
- Microsoft Bookings services per interview stage, the panel staff model, lead-time and buffer rules, the per-stage feedback forms and the Power BI report
- A pilot on one requisition family, then rollout with hypercare and a runbook for the coordinators
We need from you
- Three months of application history by channel and requisition, with the dates you answered
- Your adverts as published, so we can separate a stated requirement from an unwritten preference
- Technical accounts for the mailbox and the applicant tracking system, plus the panel members per site
Stages
Discovery
Channels, volumes, requisition families, panel structures and your real first-response times
Rules
Stated requirements per family, the wording of every candidate message, escalation and retention rules
Build
Intake robots, requirement checks, Bookings services, Teams touchpoints, Power BI
Validation
A parallel run on live applications, and a dry run of the booking flow with real panels
Go-live
One requisition family and one region first, then the rest, with hypercare
Departmental. Effort is driven by the number of intake channels, how clearly your adverts state their requirements, and whether the applicant tracking system offers an API or only a screen.
Your next driver applied on Sunday and heard nothing until Thursday.
Send us one month of application counts by channel and requisition, plus three of your adverts as published. We come back with the requirements that can be checked exactly as stated, the ones that cannot, and where your first-response time is going.
Test the checks on one requisitionThe neighbouring process usually has the same problem
New starters wait days for access; leavers keep theirs for weeks. Both are the same missing handover.
View solution HR & peopleShift cover filled in minutes, with the rules checkedA sick call should not cost a supervisor an hour of phone tag and a payroll correction.
View solution HR & peopleQualifications and training tracked before they expireYour training register is only as current as the last time someone remembered to update it.
View solution Operations & qualityOne front door for every request the building generatesBroken air conditioning, a badge, a desk move: all of it lands on one office manager's memory.
View solution Case studyAutomated new-hire onboardingNew hires ready to work sooner — less paperwork for HR.
View case study Case studyThe HR agent on TeamsAn employee asks on Teams about leave, a certificate or a benefit.
View case studyIndustries we deliver this in most oftenTransport & logisticsRetail & e‑commerceServices & IT