High-volume recruitment is decided in the hours after someone applies. An applicant contacted within minutes converts; the same applicant contacted two days later has usually accepted something else, and the advertising spend that produced them is wasted. Recruiters know this and cannot act on it, because reading CVs, chasing applicants, arranging interviews and rekeying the same details into three systems takes most of the week. The brief was therefore not “add AI to recruitment” but “remove the delay between application and first contact, without adding headcount and without letting a machine decide who gets the job”.

That last clause shaped everything. The public does not trust AI to decide about people, and the law restricts decisions made solely by automated processing. So the platform screens, scores and explains; a recruiter decides. The score is advisory and its reasons are shown, so a recruiter can disagree in one click.

Capture first: applications from job boards and inbound email turned into structured records automatically, because nothing downstream works if a person still has to rekey the application. Then screening: every CV read against the role and the client’s must-haves, scored, and the score explained in writing, so the recruiter starts the day with a ranked list rather than an inbox. Then the chase: applicants who have not completed a profile, confirmed availability or answered a question are messaged in the window while they are still engaged, over WhatsApp, SMS and email, with the conversation handled until it needs a person.

Booking followed, with interviews arranged, confirmed and reminded straight into the interviewer’s own Google or Outlook calendar. Voice came last and deliberately: inbound calls answered when the desk is busy, outbound first-pass screening calls, and a fallback designed before the voice, because from running it roughly one outbound call in eight reaches a human on the first attempt. Onboarding, document signing and right-to-work checks were brought into the same place as the hiring, so the record is complete from advert to start date. Native iOS and Android apps came with the platform, because recruiters are not at a desk.

The outcomes on this page are the ones we publish: more than ninety hours of manual CV screening removed per office every month, eight in ten applicants reaching a booked appointment rather than being lost, minutes rather than days from application to first contact, and every candidate action captured for reporting and audit. The change that recruiters describe is simpler: the morning starts with a shortlist and a set of booked interviews, and the day is spent on the conversations that place people rather than on the administration between them.

Reporting changed as a side effect. Because every action is recorded once, in one place, the numbers an owner asks for on a Friday are a query rather than an afternoon of reconciling spreadsheets that disagree, and they agree with the invoice.

Every model call on the platform is attributed to a customer, a feature and a model, and spending ceilings are enforced in real time and have been watched firing. That instrumentation is where most of what we tell clients comes from. The model is rarely the expensive part; the channels are, because a text message or a minute of telephony costs more than the model behind it. The small model does most of the work: across the platform the large frontier model handles roughly one call in eighty, and routing one high-volume feature from the frontier model to the small one made it several times cheaper and faster with no drop in success. A CV is read and scored in under two seconds in the typical case.

Those are the figures we are prepared to publish. They are also the reason the running cost of anything we build for a client is stated before go-live rather than discovered on the invoice.

Candidate data is processed under a data processing agreement with each agency, with the lawful basis, the retention and the automated-processing position written down. Screening is advisory with a human deciding, candidates are told AI is used, and scores carry their reasons so a decision can be reviewed. Augustova is registered with the ICO. The same discipline is what we apply to any client system that touches decisions about individuals.

If you are buying an AI system that must read documents, contact people by message or voice, book into real calendars and keep a record fit for audit, this is the same team, the same stack and the same practices. The timelines on our service pages are real because this is where they were measured.