HR that kept up with the hiring.
The client listed in June 2024 and kept hiring hard, and the policies grew faster than the people answering questions about them. We guided and co-built three automations with their technology team: personalized HR answers in Slack, travel planning without the email trail, and CV screening at hundreds a day.
The mandate. A newly public company, hiring aggressively, watched its HR, travel, and related policies grow more complex just as the number of people who needed answers about them kept climbing. Repeated questions, email-thread travel bookings, and a daily flood of CVs were absorbing time the teams did not have.
The outcome. Three automations in production, guided by Livehopper and co-built with the client's own technology team: a personalized HR assistant in Slack, a travel-planning bot that hands the travel desk a clean request, and a screening agent that reads hundreds of CVs a day, explains every score, and works the hiring software on screen.
The shape of it
- Three automations shipped into the tools people already use, Slack first.
- Hundreds of CVs a day screened and scored, every score with a written reason on the profile.
- Every screening session streamed live and recorded, an audit trail end to end.
- No APIs required. The hiring software offers none, so the agent operates it on screen, the way a person would.
The brief: hiring faster than the policies
The client listed in June 2024 and kept hiring hard. Going public did what going public does: the HR, travel, and related policies grew longer and more consequential by the quarter, while the number of people asking about them kept climbing. HR fielded the same questions on repeat. Travel ran as negotiation over email, with the thread itself serving as the tracking system. And every position posted drew hundreds of CVs a day, into hiring software that offers no APIs.
In 2025 we set out to take that work back - Livehopper guiding the design, co-building with the client's own technology team, so that everything shipped would be theirs to run.
An HR assistant that knows who is asking
The first build was a personalized HR assistant in Slack, where the team already lives. It answers the questions people actually bring to HR - not just what the leave policy says, but how much leave you have and how to plan it, what your last appraisal said, where you stand on performance. Answers are drawn from each person's own record and shown only to that person.
The team took to it for the simplest reason: it met them where they worked, and it answered like it knew them. It did.
Travel planning without the email trail
The second was a travel-planning bot, also in Slack. It knows the travel policy, gathers the trip, and hands the travel desk a complete request in exactly the format they want. The desk still makes the booking, deliberately: that is where the best discounts get negotiated, and there was no reason to automate away an advantage.
What went was everything around the booking - the back-and-forth over details, and the email threads that doubled as a tracking system. Requests now move through proper service tracking instead, with a status rather than a search through the inbox.
Hundreds of CVs a day, on the software's own screen
The third was the heaviest. Hundreds of CVs arrived daily for every open position, and the hiring software offered no way in through an API. So the screening agent works the way a person would: on a computer, inside the software itself.
While it works, its screen streams live, and every session is recorded, so the agent's work is auditable end to end. Each CV it scores carries a written comment on the profile explaining why it earned that score - a recruiter reviewing the pipeline sees the reasoning, not just a number. And when a profile is a complete mismatch, the agent sends the decline itself, in the company's brand language, with messaging developed together with the HR team.
The lines we drew, deliberately
- People kept the judgment. The agent screens, scores, and explains; recruiters decide. Only complete mismatches are declined automatically - and in words the HR team wrote with us, not words a model improvised.
- The travel desk kept the booking. Human negotiation is where the discounts come from. The bot removed the coordination around it, not the advantage.
- Nothing invisible. An agent working software on a live-streamed, recorded screen leaves an audit trail by design, not by afterthought.
- Built to be owned. We guided and co-built; the client's technology team shipped alongside us and runs all three today.
Three processes back in proportion with the company they serve. Quietly, in production, and in the client's own hands.
