By industry
AI agents and automation for human resources and hiring
Screening two hundred applications is one person reading for three days, and the outcome depends heavily on which day. We build systems that do the reading, evidence every judgement against the text of the CV, and hand a recruiter a shortlist they can audit — with rejection left where it belongs, as a decision a person makes.
Two weeks to a written build plan · no obligation to build with us
- Every score points at a line in the CV
- Evidence citedEvery score points at a line in the CV
- People are not rejected by a model
- No auto-rejectPeople are not rejected by a model
- Same criteria applied to every candidate
- StructuredSame criteria applied to every candidate
- To a written build plan, via the audit
- 2 weeksTo a written build plan, via the audit
Where the time goes
If two of these are true, there is something here
Written the way the people doing the work describe it, not the way a process document does.
Screening is three days of reading
Two hundred CVs, one recruiter, a deadline. The first fifty get careful attention and the last fifty get a skim, which is a fairness problem before it is an efficiency one.
Every CV is in a different shape
PDFs, Word files, scans, portfolio links, and a dozen ways of describing the same experience. Comparing them like for like is manual work that never gets done consistently.
Candidates hear nothing for weeks
Acknowledgements, status updates and rejections all need somebody to send them, so they slip — and the silence is what candidates remember about your employer brand.
HR answers the same policy questions all year
Leave balance, policy detail, payslip copies, benefit eligibility. The answers exist in the handbook and the HRIS; the route to them is a person.
What we connect to
The systems you are already running
Hiring stacks are usually an ATS, an HRIS and several sourcing channels that do not talk to either. We connect to what you have rather than replacing it.
- ATS and hiring
- Greenhouse
- Lever
- Workable
- Zoho Recruit
- BambooHR ATS
- Custom pipelines
- HRIS and payroll
- Workday
- BambooHR
- Zoho People
- Keka
- Local payroll systems
- Sourcing and scheduling
- LinkedIn Recruiter
- Job board feeds
- Google Calendar
- Microsoft 365
- Email and WhatsApp
What we build here
Five things worth building for human resources & hiring
Each one names the systems it touches, because an integration that is not named is an integration nobody has thought about yet.
- 01
CV parsing and structured comparison
Every application turned into the same fields — years in a skill, seniority, industry, location, notice period, qualifications — regardless of format, so candidates are compared on the same axes rather than on how well they write a CV.
- Touches
- CV inbox · ATS · job board feeds
- What changes
- Comparison becomes a table instead of a memory of what you read first.
- 02
Evidenced screening against the brief
The role brief becomes explicit criteria with weights agreed before applications are read. Each candidate is scored per criterion with the supporting line quoted from their CV, so a recruiter can check the reasoning rather than trust the number.
- Touches
- Role brief · parsed applications · ATS
- What changes
- A shortlist that can be defended, criterion by criterion.
- 03
Candidate communication that keeps up
Acknowledgements on receipt, status updates at each stage, interview details, and rejection messages drafted for a human to send. Nobody is left in silence because the pipeline moved on.
- Touches
- ATS · email / WhatsApp · calendar
- What changes
- Candidates know where they stand, which is most of employer brand.
- 04
Interview scheduling
Panel availability, candidate preferences and time zones reconciled, slots proposed and confirmed, reschedules handled in the same thread, and everything written back to the ATS.
- Touches
- Calendars · ATS · candidate channel
- What changes
- Scheduling stops consuming a coordinator’s week.
- 05
An HR assistant for the whole company
Leave balances, policy questions, payslip copies and benefit eligibility answered from the HRIS and the handbook after the employee is identified, with the policy clause cited. Anything sensitive routes to an HR person.
- Touches
- HRIS · policy documents · identity provider
- What changes
- HR spends its week on the cases that need judgement.
How it is put together
How an application gets screened defensibly
The design goal is not speed on its own. It is being able to explain, months later, exactly why a particular candidate was or was not shortlisted.
- 01
Agree the criteria first
Criteria and weights are written down and versioned before applications are read. A criterion invented halfway through a batch is how a process becomes indefensible.
- 02
Parse
Documents in any format become the same structured record. Where a field cannot be read confidently it is marked unknown rather than inferred — an assumed graduation year is a real-world harm.
- 03
Score with evidence
Each criterion gets a score and a quoted passage from the application. A score with no supporting evidence is reported as unsupported rather than allowed to stand.
- 04
Suppress what should not count
Name, photograph, age, gender, marital status, nationality and address are excluded from the material the scoring model sees, unless a field is a genuine legal requirement for the role.
- 05
Hand to a person
The recruiter gets a ranked list with the evidence and can re-rank or reject any part of it. No candidate is advanced or rejected by the system itself, and every human override is logged.
Your data
What we will and will not do with it
Stated plainly, and without claiming a certification we do not hold. If a requirement here is harder than what is written, say so on the call and we will tell you whether it is buildable.
No automated rejection
A model never sends a rejection and never removes a candidate from consideration. It ranks with evidence; a person decides. This is a hard rule for us — several jurisdictions now regulate automated employment decisions, and it is the right design regardless.
Protected characteristics are kept out of scoring
Name, photo, age, gender, marital status, nationality and address are stripped before scoring. That does not make a system bias-free — proxies exist in education and employment history — which is why the evidence trail exists and why we recommend you audit outcomes by group.
We do not build behavioural or emotion scoring
No video-interview scoring of tone, expression or personality, and no inferred traits. The evidence base for these is poor, the harm is direct, and they are already restricted in several markets. We will decline that scope.
Proof
We have built recruitment matching and shipped it
Oggrow is a talent recruitment company. We built their matching platform in Flutter for iOS and Android — recruiters and candidates swipe to match, and a profile only opens once the match is mutual, which is a deliberate privacy boundary rather than a game mechanic — plus their brand website and the admin analytics dashboard their team uses to report to corporate clients. Matching, profile disclosure rules and hiring analytics are the same three problems this page is about.
- One Flutter codebase
- iOS + AndroidOne Flutter codebase
- Profiles open only on a mutual match
- Match firstProfiles open only on a mutual match
- Dashboard for corporate clients
- AnalyticsDashboard for corporate clients
Technology
What this gets built with
We pick the boring option unless there is a reason not to — the framework is the part most likely to be abandoned before your system is.
- Parsing
- Document AI
- OCR
- Layout parsing
- Structured output
- Field confidence
- Assessment
- LangGraph
- Criterion scoring
- Evidence citation
- Versioned rubrics
- Integration
- Greenhouse / Lever APIs
- Workday
- BambooHR
- Calendar APIs
- FastAPI
- Controls
- Field suppression
- Audit log
- Override tracking
- Outcome reporting by group
How the engagement runs
Five phases, each with something you can hold
Every phase ends in a named deliverable. You can stop after any of them and keep what has been built.
- 01Week 1–2
Sit with the work
We watch the process as it is actually done, not as the process document says. The output is a map of every step, which system it touches and how long it takes.
DeliverableProcess map and a shortlist of what is worth automating - 02Week 2–3
Design the access layer
Which systems get read, which get written to, what scope each credential carries, and which decisions stop for a person. Agreed in writing before anything is built.
DeliverableArchitecture and permission model - 03Week 3–8
Build against your real data
Built in your sandbox against real records, not a demo dataset. The write path goes through a review queue from the first day, so nothing lands in a live system unseen.
DeliverableRunning system in staging, with the review queue - 04Week 8–10
Run it beside the humans
It runs in parallel with the people doing the job. We compare the two outputs, fix what differs, and only then start letting the routine path through unattended.
DeliverableAgreement rate report and a go-live threshold - 05Ongoing
Hand over and watch
Source code in your repository, deployment in your accounts, cost and failure dashboards you can read without us. Support if you want it, not because you are locked in.
DeliverableRepository, runbook and monitoring
Why Devs Core
Three reasons that are checkable
We run two of our own products
Cadenza and Budge are ours. What an agent costs per run, where it fails and which decisions need a person came from operating them, not from a vendor deck.
The write path is reviewed by default
Anything that changes a record in a system of record goes through a queue a person can see. You choose what graduates to unattended, and when.
You own the code
Source in your repository, infrastructure in your accounts, model keys in your name. There is no runtime of ours you have to keep paying for.
Questions
Human Resources & Hiring — the questions we get asked
Is it legal to screen candidates with AI?
It depends on where you hire, and the rules are tightening — the EU AI Act treats employment screening as high risk, and New York City requires bias audits and candidate notice for automated employment decision tools. The design we build for stays on the safe side of most of it by never making the decision: it ranks with evidence and a person decides. Take your own legal advice for your jurisdictions.
Will it be biased?
Any system trained on historical hiring can be. Stripping names and photographs helps and does not solve it, because proxies survive in education and employment history. What we can give you is the evidence trail and outcome reporting by group, so bias is measurable instead of assumed absent.
Can it reject candidates automatically to save time?
We will not build that. Beyond the legal exposure, rejection is the irreversible action in this process and the one where a wrong call has a real cost to a person. Ranking with evidence captures most of the time saving.
Can it score video interviews?
It can transcribe them and check whether specific, pre-agreed points were covered. It will not score tone, expression, confidence or personality. That category of product has a weak evidence base and is restricted in a growing number of markets.
How does an HR assistant avoid disclosing someone else’s data?
The same way our member platforms do: the employee is authenticated through your identity provider first, and the credential the assistant uses can only read that employee’s record. It is not a chatbot with a database behind it.
We hire fewer than fifty people a year. Is this worth it?
For screening, probably not on its own. For the rest — acknowledgements, scheduling, status updates and the internal HR questions that run all year — quite possibly. The audit is the cheap way to find out which half applies to you.
Related
What human resources & hiring buys most often
- Read more
AI Agent Development
Agents that take real actions in your systems — reading live data, drafting work, and writing back only what a person has approved.
From $6,500 - Read more
Process Automation
Replace the manual steps between two systems — the re-keying, the copy-paste, the spreadsheet that reconciles what neither system knows.
From $4,500 - Read more
Custom Software Development
The operational system underneath — the one the packaged software did not cover. React and Django, built to be handed over.
From $14,000 - Read more
AI Readiness Audit
Two weeks that end in a written plan: which processes are worth automating, what each would cost, and what has to be fixed first.
From $2,400
Tell us which process is costing you.
Thirty minutes, no preparation, no deck. You describe what keeps going wrong and we tell you whether AI is the answer — including when it is not.
Prefer email? contact@devs-core.com