AI screening for small teams: what to buy, and when to skip it

"AI screening" is sold as one thing and is really five, and a small team usually needs one of them, not all five. Each is described below from a vendor's own page, as an example of the kind rather than a recommendation.

Swipe the table sideways to see every column.

Skill tests and psychometric assessments sit alongside these, and some vendors sell them on the same bill: Goodfit counts an AI voice interview, a skill test, a psychometric test and a coding test as one credit each.

The question for a small team is not which of these is best in general. It is which one removes the hours you are actually losing, and whether you are losing enough of them to pay anyone at all.

Here is the arithmetic for one role that draws 200 applicants, screened by somebody whose real job is something else. The minutes per step are assumptions, not measurements; put your own in and the total changes, but the shape does not.

Swipe the table sideways to see every column.

Three working days, for one role, taken out of a week that already had a job in it. That is why a small team rarely does it: the first forty resumes get read properly, the rest get a glance, and 170 people are never spoken to.

Now change the tooling and redo the sum.

• Resume scoring only. Reading drops to reviewing a ranked list, perhaps an hour. The arranging, the calls and the write-ups remain, so you still spend about 18 hours, and the 30 you call are still chosen from documents.
• An AI interview for everyone. Reading, arranging and calling are replaced by reading the reports at the top of the ranking. At 5 minutes each for the best 20, that is 100 minutes, and every applicant was actually questioned.

The full process, step by step, is in how to screen 200+ applicants without a recruiter, and pricing your own hours is covered in what screening applicants really costs.

AI screening software for small business mostly means resume scoring, so it is worth being exact about what the document can carry.

What a resume does tell you

• Eligibility facts, when stated: location, right to work, a licence, a degree the role genuinely requires.
• The claims: titles, employers, dates, the skills the applicant says they have.
• How well the applicant writes about their own work, which matters for some roles and is noise for others.

What it does not

• Whether a claimed skill is real. "Led the migration" reads the same whether they led it or sat near it.
• Scope. "Built the platform" could be two years alone or one quarter as one of thirty.
• How they think. Nothing on the page shows how somebody handles a problem they have not seen before.
• Anything neutral about the person. In Bertrand and Mullainathan's field experiment, identical resumes sent to real job adverts drew 50 percent more callbacks when the name at the top sounded white. A document carries signals that have nothing to do with the job, and a model trained to imitate past choices can learn them too.

That is the limit of resume scoring, however good the model: it ranks the writing. It is fast and cheap, and for a pile that is mostly ineligible it is the right tool. For a pile of plausible resumes it cannot tell you which ones are true.

Swipe the table sideways to see every column.

A simple rule for a small team:

• Most of your pile is plainly unqualified? Put the must-haves on the application form as yes or no questions, or use resume scoring. You do not need an interview to discover that somebody lacks a required licence.
• Most of your pile looks plausible? That is the case resume scoring cannot settle and phone calls cannot scale to. Interview everybody, then read the top of the ranking.
• The job is making something? Ask for the work. A short, paid work sample beats both tools for a designer, a writer or an engineer, and no interview replaces it.

On Career1 there is no resume-only mode: every applicant is invited to the interview while your monthly allowance lasts, and the resume screening happens inside the interview, which asks about the claims on the document and scores the answers. Screening questions on the application form are asked, but nobody is filtered out on them before the interview.

Two costs, and the second is usually larger: the subscription, and the hours it does or does not give back.

Career1's pricing is a free plan of 2 interviews a month with no card, then Starter at $49 a month for 25 interviews, Pro at $99 for 60 and Max at $199 for 150, billed monthly and cancellable from the workspace. Every plan includes the scored report, the transcript and the recording. For one role with 60 applicants, one month of Pro covers everyone for $99.

There is a catch at the size in this article's title. Career1's largest plan is 150 interviews a month. A role that draws 200 applicants in a month outruns it, and when the allowance runs out your live jobs pause until it renews. That keeps the bill fixed, but it also means the order people applied in decides who got interviewed, which is the problem you were trying to fix. If a single role will reliably draw more than 150 applicants a month, choose a tool that sells overage or unlimited interviews.

For comparison, a few small-team prices as their vendors publish them:

• Truffle: Starter $49 a month, monthly only, 200 credits, where scoring one resume uses 1 credit.
• InterviewFlowAI: Pilot $59 a month for 20 interview credits, with resume screening "Included across all plans".
• Goodfit: $5 per completed assessment, 20 free, then a minimum purchase of 200 credits that never expire.
• Hireflix: $75 a month billed yearly for companies under 50 employees, with unlimited one-way interview responses.
• Willo Lite: £49 per live role, no contract.

Every vendor's published price, and the ones that publish none, are compared in AI interview software pricing.

A demo shows the best case. These questions show how a tool behaves in a company of eight with one open role.

1. Does it work without an applicant tracking system? Many tools assume one. If you hire from an inbox, you need a job page and an application form, not an integration.
2. Can you pay by the month, or by the use? An annual contract for a team that hires in bursts pays for the quiet months.
3. Does it reject anybody automatically? You want it to rank and explain, and a person to decide. Truffle says "Nothing is auto-rejected"; Career1 ranks and recommends and rejects nobody.
4. Can you see the evidence behind a score? A number with no quote or transcript behind it cannot be checked, or defended when a candidate asks.
5. What does the applicant go through? How long, on what device, at what hour. Your best applicants have other offers and the least patience.
6. How long is setup? For a small team the right answer is minutes, not a project.
7. What happens at the limit? Overage, a hard cap, or a pause. Know before a busy month tells you.

The fuller checklist, including data retention and fairness, is in how to evaluate an AI interview platform.

You post a role and get a job page with an application form. Applying starts the interview: the AI recruiter speaks with each applicant in the browser, on a laptop or a phone, for about eight minutes, with nothing to install and nothing to book. Its questions come from your role and the applicant's resume, and it follows up when an answer is thin, which is where borrowed experience runs out.

Each applicant arrives with a match score against the role, an overall score, a recommendation from strong hire to no hire, scored sections with written reasoning, a ledger of which claimed skills were demonstrated and the quote that shows it, risks, and questions for your own round. The transcript and recording sit beside it. There is a complete sample interview report to read before you trust one. The scores are decision support: your team reads the shortlist and decides, and Career1 rejects nobody.

• You need an applicant tracking system. Several roles, several interviewers, candidates from agencies and referrals: that is a coordination problem. Career1 has a careers page, application forms and simple stages, and nothing beyond that, and it does not sync with any ATS. The difference is in ATS or AI interviewer.
• You hire one person a year. Five applicants are five phone calls. Save the money.
• Most of your applicants are plainly ineligible. Yes or no questions on the form do that job for free.
• The job is making something. Use a paid work sample.
• One role draws more than 150 applicants a month. Career1's largest plan stops there, with no overage.
• You interview in a language other than English. The interview is English only at the time of writing.
• You need customer references or an independent bias audit before buying. Career1 is new and cannot offer either yet.

Every page below was opened and read on 28 September 2026. Career1 is not affiliated with any other company named here.

• Bertrand and Mullainathan, "Are Emily and Greg More Employable than Lakisha and Jamal?", NBER working paper 9873: nber.org/papers/w9873.
• Truffle homepage: hiretruffle.com, and pricing: hiretruffle.com/pricing.
• Paradox homepage: paradox.ai.
• Sapia.ai candidate guide: sapia.ai/candidate-explainer.
• Hireflix pricing: hireflix.com/pricing.
• Goodfit pricing: goodfit.so/pricing.
• InterviewFlowAI pricing: interviewflowai.com/pricing.
• Willo pricing: willo.video/pricing.
• Career1 plans: career1.ai/pricing.