Yaniv Shimoni

The through-line

Yaniv Shimoni's fixed idea is that for the vast majority of sales jobs, the résumé is the wrong instrument. His co-founder Kaushik spent two years researching skill-based hiring with academics and organisations including Burning Glass and Erudera, and the finding that seeded the company was that "a lot of these high volume roles that skills data is not really useful to evaluate fit for," which pushed the question to "how can we leverage demonstration data to basically help companies find the best fit job applicants" [1]. Everything else follows from that: if you cannot infer capability from a document, you have to watch someone do the work, and until recently watching someone do the work at scale was impossible.

The second half of the through-line is that generative AI has just now made the impossible version practical. Shimoni frames the ideal as unattainable and then claims the near-miss: "in an Ideal World right companies could evaluate their job applicants on the actual job before decid[ing] who to hire... but that's obviously not practical until recently but we're building the closest thing possible" [1]. Take2 AI, founded in 2021, generates tailor-made simulations of real-world sales scenarios that vet applicants for skills, behavioural attributes and culture fit [5][6].

On why high-volume sales hiring is broken

Shimoni is deliberate about scoping the market away from where most vendors point. "A lot of companies are targeting Tech B2B sales we're actually focused on the 80% the high volume repeatable entry-level sales roles" [1], meaning call centre sales, inside sales and in-store sales, against a backdrop of "over 1.4 million open sales [roles] in the United States each and every month" [1]. He gives three reasons these roles are unusually hard to fill. Skills and fit are very hard to infer from résumés, so employers fall back on "interviews and gut feeling" [1]. Degree requirements have been stripped out, so "sometimes you have north of a thousand job applicants for every open role which means it's an endless amount of time reviewing and interviewing talent" [1]. And expectations are misaligned in a way that only surfaces after the start date: someone takes a call centre role because "on paper it sounds cool to do cold calls but some people just join the job and after two days get fed up and tired of having people hang up on their faces so they just quit" [1].

He sizes the consequence bluntly. "One in three salespeople churn each year that's three times more than any other occupation," with "an average cost of 100 k for every sales hire that quits and four month to backfill," leaving companies "constantly backfilling roles... because half of their work force quits every year essentially" [1]. The cost across employers, in his telling, runs to billions of dollars [1].

On simulation as the closest thing to doing the job

The product logic is to test the actual behaviours the role requires: "whether you can call leads whether you can manage customer objections whether you can craft emails with the right etiquette whether you can pitch a product" [1]. In practice the flagship modality is a ten-minute phone call with a bot playing the customer, scored on "their tone their customer centricity how much empathy did they employ did they deliver the right content did they talk about the budget and Authority" [1]. Shimoni argues the value is two-sided: the simulations "don't only provide High signal performance data for the employers but they also immerse the job applicants in the role which leads to fewer Mish hires and reduce[d] short-term attrition" [1][5]. The candidate who discovers in the simulation that they hate being hung up on is a candidate who never becomes a ninety-day statistic.

He credits a specific technical threshold for making this viable: "we got to an inflection point where llms can actually help us generate hyper realistic... Real World Sales scenarios in the context of a specific company that actually allows you to measure things such as customer centricity empathy tonality delivery" [1]. Company and role specificity is not decoration in this argument, it is what makes the signal predictive of "who would be [the] quickest ramp up highest fit job applicant" [1].

On the results he claims, and on widening the funnel

Across customers Shimoni cites "three times reduction in time to hire 50% boost in Quota attainment and 2x increase in employee retention" [1]. His worked example is an unnamed energy company with more than a hundred open sales roles a month and 90-day attrition above 52 percent [1]. The mechanism of improvement there was as much about who got considered as about who got rejected: the company had been hiring only people with prior solar experience, and the simulation let them "attract and evaluate people who otherwise they wouldn't evaluate... solely based on their resumés" [1]. Within four months attrition fell from 52 percent to 20 percent and sales quota rose 15 percent [1]. He quotes the customer's CEO saying Take2 "has brought a lot more discipline and Clarity in a recruiting process," that "the correlation between the scores we see for a candidate and the quality of hire is extremely high," and that they now use it "as the sole screening criteria at the top of our funnel" [1]. That last point matters to him as a pattern: adoption starts at the top of the funnel and then rolls out across the organisation, into training as well as hiring [1][5].

On moving from hand-holding to a product-led motion

Shimoni is candid that early enterprise deployments were not clean. "There was a lot of handholding with these employers and Catering our product to every specific employer" [1]. The stated priority is to remove that: build applicant tracking system integrations so "the employers and the companies don't need to change their behavior" [1], and give customers an admin panel where they configure their own scenarios [5][6]. Concretely, that means letting a company tune the AI customer itself, deciding "do we want the customer to be more heated... to be less patient," and supplying their own prompts and industry and product context "so that the simulation is hyper realistic and Hyper focused on that company" [1]. His target horizon was three to six months to reach a fully integrated, self-managed product, on the reasoning that only self-service scales "across tens of thousands of companies," which in turn drove active hiring for backend, full stack and AI engineers [1]. The platform has since been described as an AI agents platform purpose-built for healthcare recruiting, whose first agent, an AI Interviewer, conducts phone interviews around the clock, evaluates candidates and records calls, alongside a $14 million Series A [7].

On who thrives on a small AI team

Two criteria govern hiring on the technical side, and Shimoni states them as a pair rather than a preference: big tech experience, "people who have experience working as a full Stack at Amazon or Google," combined with "that drive or that hunger to work in a smaller team in an early stage startup where you have a lot of impact where you have a lot of ownership" [1]. The other filter is temperament toward change. He wants people "extremely interested in Innovation" who are energised rather than unsettled by the fact that in NLP and LLMs "things are changing on a weekly basis" [1]. Culturally he backs in-person work while staying nominally flexible, on the grounds that when you are "lean and small" and "you want rapid feedback loops," the office wins [1]. Across engineering and go-to-market alike, the ethos he names is "moving fast breaking things learning quickly and adapting," with constant iteration against customer and prospect input [1].

On the pitch to candidates

His recruiting argument has three legs. First, credibility by association and traction: the same advisory board as large HR tech unicorns, backing from future-of-work investors including Tempo, Variance and Reach Capital, and Fortune 500 customers already secured "despite being so early," so "you're joining a winning team... and the people that join us early on will be the leaders of Tomorrow" [1]. Second, a distinction he clearly finds important between real AI work and veneer: "a lot of companies say that they use AI right and eventually it's simple open AI Integrations or chat GPT," whereas at Take2 "we're building and integrating llms as a core of our product and everything that we do revolves around AI" [1]. He believes the gap between public perception of AI as "just easy simple prompting with chat GPT" and actual current capability is wide enough that a few days inside the company closes it [1]. Third, flatness: "we're really hard workers but everyone here is equal... everyone has a say here we value everyone's opinion" [1].

The advisory bench is treated as an operating asset rather than a logo strip. Professors involved in the founding research now back the company, alongside HR leaders from Visa, Disney, HP and Google, and Shimoni describes them presenting on Take2's behalf at large HR and sales conferences, making introductions to colleagues and helping the company reach the early-customer results that secured the seed round [1].

Takeaways

• Target the neglected 80 percent of sales hiring, the high-volume repeatable entry-level roles in call centres, inside sales and stores, rather than tech B2B sales, with 1.4 million US openings a month [1].
• Where résumés carry no usable signal, substitute demonstration data: have candidates perform the job's actual tasks before you hire [1].
• Simulation cuts attrition from both ends, giving employers high-signal performance data and giving candidates a realistic preview that filters out those who will quit in days [1][5].
• Reported outcomes: 3x reduction in time to hire, 50 percent boost in quota attainment, 2x retention, and at one energy customer 90-day attrition falling from 52 percent to 20 percent with quota up 15 percent in four months [1].
• The biggest hiring gain can come from widening the pool, evaluating candidates without the "expected" background who would have been screened out on résumé alone [1].
• Enterprise deals that require hand-holding do not scale; the fix is ATS integrations plus a self-serve admin panel where customers tune scenario context, prompts and even how impatient the AI customer behaves [1][5].
• Hire engineers who combine big tech depth with a hunger for ownership at an early-stage company, and expect weekly-cadence change in LLM work [1].