Adam Hocek

The through-line

Adam Hocek's consistent argument is that a minute or two of speech carries enough signal to tell you who someone is, and that this shortcut around the long written questionnaire is useful precisely where human judgement is slowest and most consequential: hiring, employee wellbeing, student retention, early mental health flags. He describes Echo as analysing "psychometrics and emotions using the voice using tones of the voice," taking "a minute to two minutes of of the speech" and producing "about 70 plus traits from it anything from Big Five conscientiousness openness to uh you know you know emotional uh coping stress" [1]. The positioning is explicitly comparative: "we really position ourselves with the traditional tests of like dis Meyer Briggs those type of tests but we do it quicker and as accurate or even more accurate than some of them" [1].

The second half of the through-line is a refusal to let that stay a novelty. He is deliberate that the technology should "help the community not just be something cool or take away uh uh you know work," and wants it understood as "a helpful tool" [1]. Later, under the Aecho AI name, the same instinct shows up in his framing of recruitment as a place where the human and soft-skill dimension matters more as AI reshapes the process [2][3].

On how a voice model gets built and how accurate it honestly is

Hocek is unusually plain about method and about limits. Echo labelled its training data using the established instruments themselves: "we use some of the common tests dis and hexico and Hogan and we use those those written tests to label the data and then we use that for the voice training in the machine learning model" [1]. He volunteers that the ceiling is set by the labels, not the model: "each test varies on its predictability but Approximately 80% is about what you get uh from any one of those tests" [1]. Voice samples were drawn across "eight different languages very diverse languages" and "a range of uh ages from teenagers to uh elderly people male female across the board," giving what he calls diversity of "culture and uh and age and ethnicity," after which the model was improved by further testing and tweaking [1].

He is equally clear that a result is not a verdict. What he values in the output is the moment of friction: "when you look at your results it brings your attention to something like oh I'm an extrovert I thought I was not," so the reading sits against a person's own self-perception and "starts of course questioning in a in a in a positive in a constructive way" [1]. Self-awareness, in his account, is the point rather than classification, and it is inseparable from the mental health case: "self-awareness with mental health is absolutely important" [1].

On hiring, and what happens after the hire

HR was the first target and he treats that as the obvious move, but the more interesting claim is that assessment should not stop at the offer letter. Echo also evaluates a candidate "once a candidate gets hired as an employee their Journey Through the company how they're performing how they're coping," including the stress points that organisations rarely measure: "when they get uh a promotion are they how are they doing with their new job with the new team maybe the company's acquired which we've seen and how is that affecting the employee" [1]. Under the Aecho AI banner he continues to argue that soft skills are becoming more important in recruitment as AI reshapes how hiring is done [2][3].

On mental health as the largest opportunity

He calls mental health "a huge area" and describes work at several depths of the problem [1]. Commercially, Echo already sits behind other companies as infrastructure: "they do the uh you know the therapist consultants for them they match them up but we help with the analysis," so that a therapist meeting a client "right away when they speak to the candidate they'll know exactly they'll have an idea about them let's say and where they need to explore" [1]. Clinically, the ambition runs further: mild depression is handled today, and the roadmap covers "suicide detecting extreme uh degrees of depression," plus Parkinson's disease, where results from a data repository are "very good" but not yet tested in the field, and Alzheimer's as the natural next case [1]. The justification is access rather than replacement of clinicians: for "people who don't have access are in remote areas to doctors to therapists," a first indication is enough to send them to a professional "based on their concern" [1].

On education and student attrition

Education is the third pillar and the one he argues from a specific statistic: roughly 20 to 25 percent of US college students leave, moving elsewhere or dropping out, and "the biggest reason is emotional reasons whether it's a relationship whether it's you know family" [1]. His proposition is early flagging so that "the team at this College counselors coaches can intervene help the person with before it's too late" [1]. It is a direct application of the same claim that runs through the mental health work: distress is audible before it is declared.

On the consumer side and its limits

Hocek treats B2C as real but deliberately under-invested. The free experience on the site lets anyone record directly or upload a pre-recorded file if "they don't feel comfortable speaking in the mic," and get results [1]. He sees an obvious pull from the Myers-Briggs and 16 personalities audience, "especially you know with younger people," who want to know something about the people they are dating and matching with, but says the company is "just mildly investing in right now" and would need "someone who's really connected in that uh social media and the and the youth" before pursuing that market properly [1]. B2B remains "our main focus" [1].

On team, gaps and knowing what you cannot see

The company he describes is eleven people, six full-time and five part-time, combining a sales and marketing hire, a psychologist and researcher for "the science side of the human mind and the behavior," machine learning data scientists, advisors, and part-timers on user interface and graphic design, with lawyers and accountants used as outside services [1]. He is candid about his own shift: with a background in speech technology, engineering and computer science, "as things evolved I pulled more away from that and onto the business side of things so that's been a a transition for me" [1]. The stated gaps are sales and marketing, since the company only started selling in December and only recently began growing those functions, and someone with "the CTO kind of overarching experience to guide a team and move it in the direction of our goals" [1]. For the mental health expansion specifically, the need is data scientists and researchers to "collect data lab get the data labeled expand on that use the machine learning to uh explore that data finding the best models" [1].

He also asks outsiders for use cases, on the grounds that a founder inside the business cannot see the whole map: "we have our ideas it could be used here and there but people out there have other ideas and can see things that we don't see" [1].

Takeaways

• Voice-based psychometrics can extract 70-plus traits from one to two minutes of speech, positioned against DISC and Myers-Briggs as faster and at least as accurate [1].
• The model was trained by labelling voice data with established written instruments, DISC, HEXACO and Hogan, which caps accuracy at roughly the 80 percent those tests themselves achieve [1].
• Training diversity was treated as a design requirement: eight languages, teenagers through elderly, male and female, across cultures and ethnicities [1].
• Assessment should extend past hiring into the employee journey, including how people cope with promotion, a new team, or an acquisition [1].
• Roughly a quarter of US college students leave, mostly for emotional reasons, which makes early flagging for counsellors a clear intervention point [1].
• The mental health roadmap runs from mild depression to severe depression and suicide risk, with promising but not yet field-tested results on Parkinson's disease and Alzheimer's as a follow-on [1].
• The value of a result is that it challenges self-perception and prompts constructive questioning rather than delivering a definitive label [1].
• The named hiring gaps are sales and marketing, a CTO-level technical lead, and data scientists able to source and label data for the mental health models [1].