Prism: Listing Each Candidate's Weaknesses Too

Key points

• Prism finds the hiring bar in past hires and candidate feedback, not just in the job posting.
• Each person on a shortlist comes with where they clear the bar and where they fall short.
• Prism published its search benchmark score together with the system’s weak spots.

A company that sells the work recruiting software leaves undone

According to Prism’s launch post, a founder usually spends about a third of the working week on hiring. Even after a job is posted, most applicants who come in are closer to noise that does not fit the company’s bar. Reaching out to talent directly means competing with the flood of messages other companies send. Handing the job to a recruiting agency is expensive, and the process is slow and opaque.

Recruiting software that arrived to fix this brings another worry. It puts thousands of profiles on the screen that are only loosely related to the role. The work of reading each profile, picking the right ones, and bringing them to the interview stage is still left to the customer. Prism’s founders point to this gap: “The hardest part of hiring isn’t picking the tool. It’s doing the work.”

Prism was founded in 2025 and took part in the Y Combinator Spring 2026 batch. It calls itself an “AI-native recruiting agency.” CEO Theo Kitsberg ran recruiting and training for a mental health helpline while at Cambridge. CTO Axel La Pira worked on in-house recruiting infrastructure at Alan, a French healthtech company. Prism does not sell software subscriptions. Like a traditional recruiting agency, it sells the finished work to its customers: finding candidates, reaching out, screening them, and handing over a shortlist ready for interviews.

Seven steps from a request to a finished shortlist

The process in which a customer hands over a hire starts with a messenger conversation. In a Slack channel shared by Prism and the customer, the customer describes the role to hire, as the instructions put it: “You brief us on a Slack channel.” It also supports entering the job requirements in a short web form, and Prism starts its candidate search the moment the form is done.

Once it receives the request, the system analyzes the job description together with the customer’s past hiring records. Prism says that at this step it finds the criteria hidden behind the job description and the past hires. The job description says what the customer asks for, but the past hiring records show whom the customer actually chose.

Once the criteria are set, Prism searches three places at the same time: its own candidate pool, the customer’s company network or applicant tracking system (ATS), and the open web. In its launch post, Prism says it now runs dozens of searches in parallel and instruments every step. According to the company’s technical report, the system finds up to 50 ranked candidates per role and puts together each candidate’s work history and the reasons for the match.

Early in the search, it takes quick feedback from the customer to correct course. When the system shows the customer a few early candidates it picked out and gets a thumbs up or thumbs down, the search system fine-tunes its criteria based on those choices.

After candidates are found, outreach begins. Prism sends each candidate a message tailored to the role through LinkedIn, email, and WhatsApp. Most of the people Prism finds never applied to the company in the first place.

Candidates who respond and pass screening make up the final shortlist. According to a post the founder wrote on Product Hunt, this ranked list sets out for each candidate where they clear the bar, where they fall short, and what to look into further in the interview. The customer then runs the interviews and extends the offers, as the instructions put it: “You step in to interview and make the offer.”

There is a job seeker flow too. A job seeker who is thinking about a move has a “10-minute conversation with Prism, your AI agent,” and talks about their experience and the conditions under which they would change companies. The system matches this to the customer’s open roles, and a job seeker’s profile is not shared with an employer without the person’s approval. The founder explains that a candidate who does not fit one company may fit another company’s search, so the overall candidate pool keeps growing as searches go on.

A small number of people stand behind this work. The YC page lists the team as three people, and the launch post says four. Besides the two founders, there is a founding engineer and an in-house recruiter, and the last two were hired through Prism’s own system. According to the website, Axel La Pira builds the systems and Theo Kitsberg runs recruiting operations. On Product Hunt, Theo Kitsberg wrote, “we built our own AI tooling and we run the searches ourselves.” The large-scale searching, screening, matching, and outreach are left to agents and software, and people’s time goes, in the launch post’s words, “only where it actually matters.” What Prism’s own staff check before a shortlist goes to the customer, and how often they step in, has not been made public.

The pricing differs from page to page. The launch post and the YC profile say Prism charges a $500 retainer when a search starts, plus 15% of first-year salary when a candidate Prism found signs the contract. The current website, by contrast, describes a fee of 15% only when a hire is made, with no retainer or monthly fee, and carries the line “No hire, no bill.”

Where human labor was replaced by agent search

A pricing structure built on a success fee has long been used in the recruiting agency business. What Prism changed is not the condition under which it collects the fee, but the way work actually runs behind the fee.

The first thing that changed is who does the searching. At a traditional agency, a recruiter looks for résumés one by one, reads the profiles, and writes the messages. At Prism, the system runs dozens of searches in parallel. People set the search criteria, manage the relationship with the customer, and step in only where a human touch is needed. The founders explain that AI should not try to replace the human parts of hiring, such as judgment, relationships, and context. It should instead clear the way so people can focus on that essential work.

How a customer’s request becomes hiring criteria is also different. An ordinary agency leans mainly on the job posting and meetings with the customer. Prism adds the customer’s past hiring records and thumbs-up or thumbs-down reactions to the first candidates. It sets the criteria by taking into account not only the conditions the company officially asks for, but also the traits of the people it actually chose.

The form of the result the customer receives differs too. Recruiting software hands over a list of thousands of profiles, and a traditional agency tends not to show openly the grounds on which it made a recommendation. Prism sends a ranked shortlist with reasons attached for why each candidate fits the bar and where they fall short. A commenter on Product Hunt called this approach a great answer to the black-box nature of traditional agencies, whose process you cannot see.

Its attitude toward verifying its own performance also differs from a traditional agency. Prism published a technical report using a public people-search benchmark. In an evaluation measured on 30 recruiting queries, Prism reported a score of 89.6, ahead of 68.2, the highest score among published products. In the report, the company itself clearly notes the limits, calling it a “vendor self-evaluation on a benchmark authored by a competitor.” But what stands out more is that the agency laid open the weak spots of its own search. The report candidly admits that the system performs worst when a requirement is not on the public web, such as the undisclosed sale price of a previous employer or a graduation year.

What changes when the line between software and service falls

If a search system and a few people can do most of the searching and first screening that an agency team used to do, the fees in the recruiting agency market could come under strong downward pressure. That is because Prism can make a profit even while charging much less per hire. In fact, Prism’s website has dropped the mention of an up-front retainer. But whether this cost structure will hold up when the number and range of roles it handles grow a lot has not yet been tested.

The wall between software and professional services could also fade. Prism sells a finished hiring result, not permission to use software. At the same time, according to its technical report, it also sells its own system to recruiters. If one engine can support both a professional service and a software product, existing agencies may have to compete with work tools, and software companies may have to face off against professional agencies.

The lessons other organizations can take from the way Prism works are clear too.

The first is to use past choices as data. Whom a company actually hired in the past, and what feedback it gave on early proposals, show the company’s bar more clearly than the words in a job posting. Many in-house hiring teams already hold this kind of data and do not use it.

The second is to give the grounds for judgment with every result. When a list names the points where each candidate clears the bar and falls short, instead of handing over only a scored list, the person reviewing can check the result more easily and trust it.

The third is to decide clearly where people stay. Prism automates the large volume of repetitive work of finding and sorting candidates, hands the final interview and the offer to the customer, and keeps its own staff focused on recruiting operations and relationships between people. It is a division of labor that hands the volume to the system and keeps people only where judgment and context matter.

What has not yet been fully proven

How sustainable and effective this approach really is remains to be seen.

The workflow, benchmark scores, customer list, and testimonials made public so far are all material that Prism published itself. How many searches the system has handled so far, how many people were actually hired, how many days it takes to hand over a shortlist, and what the candidates’ response rate is have not been made public.

It is also unclear what quality checks staff go through before the final list goes to the customer. At the service’s launch, a user asked how it handles US rules against hiring discrimination, but the founder gave no substantive answer.

Some points in the published company information also conflict. The team size is given as three or four depending on the page, and the company’s location is shown differently as London and San Francisco. Whether the $500 search retainer exists also differs between the website and the YC page. There are also traces that Prism once operated under the name HireGlide, so whether this new way of working can take root steadily across many hiring settings can be judged only after more real hiring results build up.

Sources

1. Y Combinator — Prism: AI-native recruiting agency (company page) — https://www.ycombinator.com/companies/tryprism — accessed 2026-10-22
2. Y Combinator — Launch YC: Prism: AI-Native Recruiting Agency — https://www.ycombinator.com/launches/QOK-prism-ai-native-recruiting-agency — accessed 2026-10-22
3. Prism — Homepage — https://tryprism.com/ — accessed 2026-10-22
4. Prism — About — https://tryprism.com/about — accessed 2026-10-22
5. Prism — Candidates page — https://tryprism.com/candidates — accessed 2026-10-22
6. Prism — Employer intake page — https://tryprism.com/start — accessed 2026-10-22
7. Axel La Pira (Scrambld, Inc.) — Prism on PeopleSearchBench-Recruiting: A Technical Report — https://tryprism.com/papers/people-search-benchmark.pdf — 2026-05-03
8. Product Hunt — Prism: Hire the best candidates, not just the available (launch and maker comments) — https://www.producthunt.com/posts/prism-00e67431-1004-462c-ba5d-3124fde2f5ce — 2026-05-09
9. EveryDev — Prism (tool listing) — https://www.everydev.ai/tools/tryprism — accessed 2026-10-22