Puerto Rico's AI-powered hiring platform: Matching candidates to government jobs by skills

The problem

Recruiting staff into government in Puerto Rico was a process that worked against both the agencies trying to hire and the candidates trying to apply.

Puerto Rico has 56 government agencies, and each ran its own recruitment, much of it on paper and shaped by different rules and required documents from one agency to the next. There was no single place for candidates to see what jobs were open across government, and no standard way for agencies to assess who was qualified. Filling a single vacancy could take months. Faced with a long, opaque process, many candidates gave up or accepted another job before they heard back about their applications. For agencies competing with the private sector, which is strong in Puerto Rico, those delays made it harder to attract the people they needed.

There was also a trust problem. Hiring at the Puerto Rico Department of Education, a large agency, had faced repeated criticism that bias, including the influence of personal connections, could affect hiring decisions. Without a transparent, standard process, agencies found it hard to demonstrate that a hiring decision rested on a candidate's skills and qualifications.

The solution

As part of a broader Civil Service Reform led jointly by the Government of Puerto Rico and the Financial Oversight and Management Board (FOMB), the United States federal board that oversees the territory's public finances, a new digital hiring platform was introduced in 2024. It launched at six agencies at the start of March 2024, and the Puerto Rico Department of Education joined at the end of April, forming a seven-month pilot that grew to nine agencies.

The platform replaced the patchwork of paper-based, agency-specific processes with one standard system. Candidates search, apply and track their progress in a single place, and agencies post vacancies, review applications and manage hiring through the same system. For the first time, recruitment across the participating central government agencies was conducted through a single, consistent process.

The platform is built on a talent intelligence platform supplied by Eightfold AI. It uses machine learning, a type of AI that learns patterns from data, together with natural-language processing, which lets software read and interpret written text, to analyse the experience and qualifications in each application. The system reads a candidate's resume and application, identifies the skills described in them, and matches those against what each vacancy requires. It then shows the applicant other open positions across government that fit their skills, so they can apply for those as well. Someone who comes to the platform for one role may find they qualify for several others they had not considered.

To address the concern about fairness, the platform masks candidates' names during the early stages of evaluation, showing hiring staff only their initials. The aim is for the first assessment of a candidate to rest on skills and qualifications. For agencies where the fairness of hiring has been publicly questioned, this directly addresses that concern.

The AI handles the initial screening to shortlist candidates by skills. Human resources staff then carry out phone and in-person interviews and make the final hiring decisions.

The reform pairs this recruitment platform with talent management measures that support employees once they are in post, including a skills-based evaluation system for current staff, an online learning platform for building new skills, and a professional development model with pathways to promotion and additional pay.

The impact

1. Application volumes surged

In the first two months, the platform received over 12,000 applications for 256 job postings across the participating agencies. To put that in context, that figure equalled half the total number of applications the government's previous hiring portal had received in the entire two years since it launched. The new platform did not just move the old process online — it brought significantly more people into contact with government job opportunities.

2. Hiring times dropped from months to days

At the Department of Economic Development and Commerce, 41 of 57 vacancies were filled within two months of the platform going live, with an average hiring time of 13 business days per role. Under the previous system, filling a single vacancy could take months. The AI carried out the initial skills screening, while HR staff conducted the interviews and made the final decisions.

3. Candidates discovered roles they had not considered

Of the people who used the platform, 3,448 applied for more than one job. The AI's ability to analyse a candidate's skills and suggest additional matching vacancies meant that people who came looking for one role found others they were qualified for. This broadened the applicant pool for each vacancy without requiring agencies to do additional outreach or advertising.

4. The Department of Education saw immediate take-up

The PRDE, which joined the platform in late April 2024, received 8,544 applications for 177 vacancies from 2,276 candidates within its first two weeks. For an agency where the fairness of hiring had been a longstanding public concern, the visible change to a standardised, skills-based, name-masked process matters as much as the speed improvement.

5. The pilot met its stated goals

Over a seven-month pilot across nine agencies in 2024, the participating agencies published 311 job announcements and hired 339 people, more than expected, with posts filled in about two months on average and some in under 13 days. The Oversight Board reported that the pilot's goals, faster hiring, less bias in evaluation and a better overall process, had been met, and it commissioned a study and a public webinar to review the results. Following a change in the territory's government administration, the platform is not currently active, and the Board is awaiting confirmation from OATRH on the project's status. A wider rollout to all government agencies would depend on the government setting clear guidelines for large-scale implementation.

Lessons learned

• Standardising and digitising the process delivered gains alongside the AI. Much of the early improvement came from replacing fragmented, paper-based, agency-specific recruitment with a single standard digital platform used by both candidates and agencies. For governments where each agency still recruits on its own system, creating a single, consistent process is a foundational step that the AI-specific features then build on.

• Showing candidates roles they had not searched for widened the talent pool. Beyond the role a candidate applied for, the platform assessed their skills and surfaced other government vacancies that matched, and 3,448 applicants went on to apply for more than one job. The pattern suggests governments may be missing qualified applicants who simply do not know which other roles are open.

• Name masking responded to a concrete trust problem at specific agencies. Showing only initials during the early evaluation was a direct response to public criticism of hiring bias. The available sources describe the design and its intent and do not report measured effects on hiring decisions, so it is best read as a change to the visible process, in a context where public trust in government hiring had been damaged.

This case study was written with assistance from artificial intelligence.