Willo Review: Asynchronous Interviews, AI Assistance, and Buying Tests

Willo is an asynchronous candidate-screening platform that collects video, audio, text, multiple-choice, and file responses for later human review. Its newer AI functions can transcribe, summarize, suggest questions, and surface areas for follow-up. The product is best used to standardize early evidence collection, not to replace live interviews or make unreviewed hiring decisions.

The previous article overstated Willo as an independently proven AI assessment system. Public sources support a narrower conclusion: Willo has a documented screening workflow, published pricing, ATS integrations, and security materials. Its hiring outcomes and AI validity still require buyer-run testing.

Direct answer: what does Willo do?

Willo lets an employer create a set of questions, invite candidates, collect responses on the candidate’s schedule, and share or review those responses. Its current feature page describes multiple response formats, scorecards, transcription and summaries, question generation, identity checks, anti-cheating tools, an API, and third-party integrations.

These are vendor-described capabilities. Buyers should verify which features are included in Lite or Enterprise, available in their country, and appropriate for the selected role. The platform does not conduct a synchronous two-way interview in the ordinary meaning of that phrase.

Asynchronous screening solves a scheduling problem

Recorded responses can remove calendar coordination from an early stage and give every candidate the same core questions. Recruiters and hiring managers can review at different times, which is useful across locations and time zones.

The format also removes live clarification and rapport. A candidate may interpret an ambiguous question differently, struggle with recording conditions, or be unable to ask about the role. Employers should therefore limit asynchronous screens to questions that can be answered fairly without a conversation.

Good starting uses include:

• a short explanation of relevant experience;
• a work sample or demonstration with clear instructions;
• consistent motivation and availability questions;
• simple eligibility checks with an accessible alternative;
• preparation for a later human interview.

It is a poor replacement for negotiation, sensitive discussion, executive assessment, or any stage where mutual exploration is the point.

Question design determines evidence quality

Recording the same weak question does not make a process structured. Begin with a job analysis and a scoring rubric tied to observable behavior.

For each question, document:

Element	Required definition
Competency	Job behavior being assessed
Prompt	Clear question with no irrelevant personal inquiry
Evidence	What a strong response must demonstrate
Anchors	Examples for low, medium, and high scores
Time	Preparation and response limits
Review	Number and role of reviewers
Alternative	Equivalent accessible route

Keep the screen short enough that it does not become unpaid speculative work. Pilot question comprehension with people outside the hiring team before inviting real applicants.

AI summaries are reading aids

Willo markets “Willo Intelligence” features that transcribe and summarize responses, identify skills or gaps, and generate follow-up questions. These outputs can help a reviewer navigate a large set of recordings. They should not replace the underlying response.

Require summaries to point to timestamps or transcript passages. Reviewers should be able to correct transcripts, see uncertainty, and distinguish what the candidate said from what the model inferred. Generated follow-up questions must remain within approved job-related competencies.

Do not permit an AI summary or benchmark to auto-reject a candidate. A concise but inaccurate summary can create stronger automation bias than a raw transcript because it appears authoritative.

Verify human review in the workflow

Willo states in current company materials that a human makes progression decisions and that the AI prepares evidence rather than automatically rejecting applicants. That is a vendor policy claim, so buyers should confirm it in their configuration and contract.

Test whether administrators can create score-based automation, whether an ATS integration can move or reject a candidate automatically, and whether defaults change when AI modules are enabled. Establish a policy that preserves meaningful review even if the software technically permits stronger automation.

Record the reviewer, evidence, rubric score, override, and final decision separately. That makes it possible to assess whether humans are reviewing or merely accepting machine suggestions.

Greenhouse documents the integration workflow

Willo’s feature pages advertise broad integration coverage. Direct connectors and no-code automations do not all provide the same depth.

The Willo Greenhouse guide documents adding a Willo stage to a Greenhouse interview plan, selecting a Willo template, sending invitations, viewing status, and opening a completed response report. The instructions also warn that API keys provide access to sensitive data and should be transferred securely.

For any ATS, test:

1. job and template selection;
2. candidate identity and duplicate applications;
3. invitation delivery and bounce status;
4. completion and withdrawal states;
5. score, report, transcript, and recording writeback;
6. permissions for shared report URLs;
7. retries after an outage;
8. deletion and credential revocation.

An integration that returns only a link may require different reporting and retention controls from one that writes structured scores.

Candidate experience needs direct measurement

Asynchronous interviews shift scheduling work from recruiters but ask candidates to prepare, record, and troubleshoot on their own. The trade can be fair when instructions, duration, support, and alternatives are clear.

Measure:

• invitation delivery and start rate;
• completion and abandonment;
• time to complete;
• mobile and desktop success;
• support contacts and resolution time;
• retakes and technical failures;
• accommodation requests;
• candidate survey response and score;
• withdrawal after invitation.

Vendor NPS, response-rate, and customer outcome figures should be treated as company-reported metrics unless the cohort and method are published. Use the employer’s own applicants as the acceptance population.

Accessibility requires more than device coverage

Willo’s home and security materials describe multilingual access, WCAG alignment, browser-based use, and candidate support. These are useful commitments but not proof that every question and configuration is accessible.

Test keyboard navigation, screen readers, captions, transcript access, color and focus states, zoom, additional time, low bandwidth, camera and microphone permissions, and switching devices. Offer an equivalent text, phone, or human option where needed.

The US ADA guidance on hiring technology explains that automated hiring tools can unlawfully screen out qualified people with disabilities or create disability-related inquiries. The employer remains responsible for accommodation and equal access even when a vendor hosts the screen.

Anti-cheating and identity tools are signals

Willo markets anti-cheating analysis, employer verification, candidate identity verification, right-to-work checks, and background-check options. Each has a different purpose and legal basis. Do not combine them into a single trust score.

AI-use or behavior detection can produce false positives. A candidate may look away to use assistive technology, read an approved note, experience latency, or speak in a rehearsed style. A flag should trigger review and a chance to explain or retest, not automatic rejection.

Identity and right-to-work checks should be introduced only when necessary, with clear notice, approved providers, retention limits, and access controls. Verify that checks occur at a legally and operationally appropriate stage.

Security materials are relatively visible

Willo’s information security page describes encryption in transit and at rest, AWS hosting, monitoring, vulnerability scanning, and an ISO 27001:2022 claim. The Willo Trust Center lists certification, policy, architecture, encryption, privacy, and accessibility resources, some of which require access.

These are vendor disclosures. Enterprise buyers should obtain the certificate and scope, current audit evidence, penetration-test summary, incident history, recovery test results, and remediation status. Confirm which legal entity and environments are covered.

Video, audio, transcripts, identity results, reviewer comments, and derived AI outputs all need explicit roles, retention, export, and deletion behavior.

Subprocessors depend on enabled features

Willo’s subprocessor documentation describes international-transfer mechanisms and links to a current list. The Trust Center identifies providers used for hosting, transcription, and optional AI functions and says some enterprise AI services have model training disabled.

Map the actual configuration rather than accepting a generic list. Ask:

• which provider receives audio, video, transcript, or prompt data;
• whether a module is optional and disabled by default;
• processing and storage regions;
• retention in primary systems, logs, and backups;
• human access by Willo or a subprocessor;
• training and product-improvement terms;
• change-notification and objection rights;
• deletion verification after termination.

The employer should be able to disable an AI module without losing access to the underlying recorded interview workflow.

Pricing is public and date-bound

Willo’s pricing documentation, reviewed August 5, 2026, describes Lite for occasional hiring and Enterprise for ongoing screening. Current company pricing materials list Lite at $59 per live role and Enterprise from $3,799 per year in US dollars. Verify the live checkout or quote because prices, currencies, and packaging can change.

The fair-use terms, reviewed July 7, 2026, publish response and SMS allowances for named 2026 plans plus overage rates. Those plan names do not perfectly match every current marketing page, so the signed order form should control.

Model total cost using active roles, candidate responses, SMS, storage, identity or background checks, integrations, implementation, support, and overage. Compare cost per completed reviewable screen, not only license price.

Public outcome evidence is mostly vendor-controlled

Willo publishes customer counts, NPS, completion, time-saving, and case-study results. These can identify reference customers and possible use cases, but public pages generally do not provide controlled comparisons or enough methodology to infer a universal effect.

Ask references about the same role family, application volume, region, ATS, assessment length, candidate population, and rollout stage. Determine whether reported time savings were measured or estimated and whether quality, accessibility, and candidate withdrawal changed.

The product can clearly remove some scheduling activity. It does not follow that it improves the validity of the hiring decision.

A four-week acceptance pilot

Use one high-volume role and one lower-volume professional role to expose different failure modes.

1. Approve the job analysis, questions, rubric, notice, and alternative process.
2. Configure retention, roles, ATS fields, and AI modules in a test environment.
3. Run accessibility and device testing before candidate invitations.
4. Have two reviewers independently score a historical or consented sample.
5. Compare human review with AI transcripts and summaries, recording material errors.
6. Launch prospectively and measure completion, support, agreement, overrides, and progression.
7. Review flagged cheating cases for false positives.
8. Test export, deletion, outage recovery, and contract termination.

Expansion should require written thresholds for both operational performance and candidate protection.

Best fit, poor fit, and unknowns

Willo is most plausible for teams that need consistent asynchronous first-stage evidence across substantial applicant volume and can provide human review, accommodations, and ATS governance. The published pay-per-role option may suit occasional use, while enterprise buyers need a full security and integration review.

It is a weaker fit for live relationship-led interviews, very low volume, roles that require immediate clarification, or organizations seeking automatic candidate ranking. It should not be used where reviewers will not inspect source responses.

Public sources do not establish independent AI-summary accuracy, anti-cheating precision, role-specific validity, fairness outcomes, causal hiring ROI, or reliability for every advertised integration. Those remain buyer tests. Willo can organize early screening, but the employer must determine whether that screen is valid, accessible, and worth the candidate burden.