Matching and scoring engine
Paste in a job description and it returns one ranked shortlist, drawn from all three candidate sources in a single pass.
AI recruitment platform case study · HR tech
From job description to interview shortlist in under a week.
An HR-technology company wanted to give fast-growing startups an alternative to agency fees and weekends lost to CV screening. We designed and built the platform that does it.
- One ranked shortlist from three candidate sources
- Works inside 50+ applicant tracking systems
- Designed for ISO 27001 and GDPR from day one
- Next.js
- AI matching
- ATS integrations
- ISO 27001
- GDPR
In brief
- The client
- An HR-technology company in Asia-Pacific, with customers worldwide. We did the product design, the AI engineering and the platform build.
- The problem
- Newly funded startups had to hire fast. CV screening cost 15+ hours per role; an agency often cost $25,000 to $40,000 per hire.
- The result
- A shortlist ready for interviews in under a week, from a platform that works inside 50+ applicant tracking systems (ATS), at a 95% lower average cost per hire than agency recruiting.
The challenge
Two bad options, and tools that did not join up.
The client's target customers had just raised funding and needed to double headcount. A founder or small HR team could lose 15 hours or more per open role to CV screening, or pay agency fees out of the money they had just raised.
Agencies typically charge around 20% of the new hire's first-year salary, often $25,000 to $40,000 per hire, and send back a shortlist of three to five people.
Our client set out to build an AI hiring assistant that does what a strong in-house recruiter does. It pulls from three places at once: the company's own applicant pool, a network of pre-qualified candidates who have said they're open to a move, and profiles on the open market. Then it hands back a ranked shortlist a hiring manager can start contacting straight away.
What the platform had to get right
- One workflow, not three tools. Existing recruitment software treated sourcing, screening and outreach as separate products, so hiring managers stitched the steps together by hand.
- Work inside the tools teams already use. The market runs on more than 50 ATS, the systems HR teams use to track applicants. A busy team was never going to replace theirs to try something new.
- Protect candidate data from day one. Candidate data is sensitive by nature. Enterprise buyers ask about security and privacy before a pilot, not after early traction.
- Earn the hiring manager's trust. A non-technical hiring manager had to act on a ranking without re-checking every result. If they didn't trust it, they wouldn't use it.
- Serve two kinds of buyer. Operators who want to run the search themselves, and founders who want to hand the whole process off.
What we built
One workflow for sourcing, screening and outreach.
An end-to-end recruitment platform that runs sourcing, screening and candidate engagement as one workflow each customer can configure. We took it from the first architecture decision to production launch, and scaled it after that.
Native integrations with 50+ ATS
Results appear inside the system a team already uses. There is no data migration and no change to how they work.
Chat and voice assistant
A web application built with the Next.js framework, with a conversational assistant for recruiters and for candidates, including voice.
Autonomy controls
The customer decides how far the AI goes: screen applicants and stop, or screen, score and start outreach, or a point in between. It is the same design question behind any AI agent that acts for a user.
Fast onboarding, measured results
Self-serve onboarding gets a new customer to a first shortlist in under five minutes, with a sales-assisted route for larger accounts. In-product analytics track hours saved and cost per hire, so the client's sales and marketing teams have measured numbers to show prospects.
Security and privacy by design
An architecture designed for ISO 27001 and GDPR, the EU's data protection law, from the ground up. It has audited data handling, encryption and access controls, and runs on enterprise cloud infrastructure.
The results
Hiring in the same week, without an agency.
Weeks of manual work and agency dependence gave way to a self-serve process that customers finish inside a week. Figures are the client's own, as reported to us.
- 95%
- lower average cost per hire than agency recruiting
- 15+ hrs
- saved per hiring manager, per open role
- 23 hrs
- saved on CV vetting per hiring round
- <7 days
- until a shortlist is ready for interviews
- 50+
- ATS integrations live at launch
What made it work
Letting each customer decide how far the AI goes.
A ranked shortlist demos well. Whether a hiring manager acts on it is a different matter, and that comes down to trust and control.
Some teams want the AI to screen applicants and nothing more. Others are happy for it to score candidates and send the first message before anyone has looked. We don't think one setting suits every customer, so autonomy became something the customer sets. That was as much a product call as an engineering one.
Because the integrations were live at launch, customers adopted the platform without disrupting the way they already hired.
Why the client chose us
The client spoke to several engineering firms before choosing us. The deciding factor was a combination it couldn't find elsewhere.
- Engineers who build AI. Hands-on AI and machine learning experience.
- Production, not prototypes. A team that had put integrations into production systems instead of stopping at a prototype.
- Security and compliance know-how. People who understood the questions HR-tech buyers ask before a vendor gets anywhere near candidate data.
- A product partner. We sat in on architecture and prioritization calls with the founders. Nobody handed us a fixed spec and waited for it to come back built.
Where the platform is now
The client achieved ISO 27001 certification and GDPR compliance, which opened sales conversations with security-conscious enterprise buyers. High-growth companies in retail, media and B2B SaaS (software sold to other businesses as an online service) now use it as a core part of the hiring stack they put in place after a funding round.
Related services
Services for a project like this.
AI engineering, integrations and security built in from the start. Pick the question closest to yours.
- Our people lose hours to work an AI system could take on, like screening CVs.AI development
- We want AI that acts for our users, within limits each customer sets.AI agent development
- Our product has to work inside the tools our customers already use.AI automation and integration
- Our buyers' security teams will question the AI before they sign.AI security
- Enterprise buyers ask whether we hold ISO 27001.ISO 27001
- We hold candidate or customer data covered by GDPR.GDPR compliance
More case studies: the AI teaching platform for schools, a bank security assessment and an ISO 27001 certification program. Or see every case study.
Let's talk
Building an AI product that has to pass a security review?
We build the AI, and we know what your buyer's security team will ask about it. Tell us what you're building and who has to approve it. We reply within one working day.