From admin-heavy to people-centric recruitment
Hiring at Hyde & Associates ran on admin: CVs read by hand, interviews arranged by email, and the owner meeting every candidate. We helped the firm hand the admin to AI tools under written rules. The recruitment lead now spends her time with candidates, and the owner meets only the shortlist.
515applications for one vacancy, in about 10 days
~80%less screening effort for the recruitment lead
10 → 3.5owner interviews per hire: 30 for three hires in 2024, seven for two in 2026
0 → 17hrecruitment lead time with candidates per vacancy
Hyde filled two associate roles from a single vacancy. The recruitment lead now spends about half of her recruiting time talking to candidates rather than handling paperwork. Because the people who reach the owner are better screened, he meets about a third as many per hire.
What was going wrong
| Area | What we found |
|---|
| Intake | Each vacancy posted by hand on five to eight job sites, each with its own login. No presence on LinkedIn. |
| Screening | Every CV opened by hand, 2 to 10 minutes each: 30 to 50 hours per vacancy. Candidates judged mostly on feel. |
| Selection | The owner interviewed every candidate, usually for 90 minutes or more. In 2024 he held 30 interviews to hire three lawyers. |
| Coordination | CVs zipped and emailed. Interviews arranged by hand across European and Asian time zones. |
| Definition of good | The owner valued independent judgement and legal reasoning, but nobody tested either before probation. |
| Memory | No talent pool. Good runners-up were lost after every round. |
The cost: the firm judged three of its five hires since 2024 to have been avoidable, at around €25,000 each in salary, onboarding and recruitment time.
What we did
- Put numbers on the old process. Two interviews and a simple business case for what hiring cost the firm, in money and owner time.
- Kept the scope tight. Mapped the whole employee journey, then focused on the steps before the hiring decision, where the time and the risk sat.
- Agreed what a good hire looks like. Turned the owner's view of a strong hire into knockout questions, a scoring rubric and signs to look for.
- Built case-study interviews for both roles, with a structured scorecard and planned curveballs that test what past hires were missing.
- Chose the tool to fit the process. Twenty weighted requirements and scripted vendor demos, such as screening 500 CVs. Hyde chose Workable.
- Moved people's time to the end of the funnel. Machines clear the obvious mismatches. People only meet the shortlist. A first interview with the recruitment lead now comes before the owner's final interview.
- Added an AI scoring layer. Claude, connected to the platform, scored 160 candidates against the firm's own criteria and quoted evidence for every score.
- Set rules for the AI and wrote down the process. A ten-section operating manual, owned by the recruitment lead.
- Checked the results against the evidence. Platform records compared with the March business case, including what did not work.
The AI agents and their rules
| Agent | What it does | Rule |
|---|
| Sourcing | Found more than 100 relevant candidates beyond those who applied | Outreach stays off until a person reviews it |
| Screening | Checks every application against five yes/no questions and ranks the rest | Only the yes/no questions can reject; a person checks borderline cases |
| Rubric scoring | Scored 160 candidates against Hyde's own profile, quoting evidence | It recommends; a person decides |
| Communication | Acknowledges every application in seconds and sends self-booking links | Templates approved by the firm; personal notes after interviews |
| Assessment | Sends and marks aptitude tests | A pass mark, not a ranking |
Results against the business case
| Measure | March target | September result |
|---|
| Screening effort | 75% lower | About 80% lower. Around 12 hours for 515 applications. |
| Owner interviews per hire | 50% lower | About 65% lower. Seven interviews for two hires, against 30 for three in 2024. |
| Recruitment lead's time | Not targeted | From about 50 hours, mostly admin, to 32 to 35 hours. About half now goes on candidates, up from none. |
| Finding the eventual hire | Not measured | Flagged two days after applying. Final interview booked 15 days later. |
| Hires | One per vacancy | Two from one vacancy. |
| Candidate experience | Not targeted | A candidate who did not get an offer described the process on Glassdoor as quick and professional. |
| Quality of hire | Fewer avoidable hires | Too early to call. Ninety-day reviews are due in Q4 2026. |
“I saved admin time to give more added value to the recruitment process.” “Definitely, I wouldn't go back [to the old ways of working].”
Ana Domingos, Recruitment Lead, Hyde & Associates
What we would tell another owner
- Agree what good looks like before you buy anything. A tool set up around a vague profile finds the wrong people faster.
- Move people's time from the start of the funnel to the end. Your busiest person should only meet candidates who passed every check a machine can do.
- Measure where the hours went, as well as how many were saved.
Basis of figures: application volumes, pipeline stages and interview counts from Workable system data, June to September 2026. Hours and screening effort: the recruitment lead's own estimates, September 2026. Baseline and targets: the business case agreed with the client, March 2026.
