Eightfold AI has launched Candidate Agent as the candidate-facing component of its Talent Agents 2.0 release. The agent is designed to keep job seekers in one conversation from role discovery through application, scheduling, follow-up and hand-off to Eightfold’s AI Interviewer.

The company says Candidate Agent is now generally available to organisations with Eightfold Career Site enabled. It supports more than 24 languages and can interact through web, SMS, WhatsApp and voice. For existing eligible customers, Eightfold says the agent can be enabled without a separate integration.

A conversational front door for hiring

Candidate Agent replaces parts of the conventional form-based application journey with an adaptive conversation. A candidate can ask about a job, provide experience and availability, search for other roles, complete application steps and move into interview scheduling without repeatedly entering the same information across disconnected systems.

Eightfold says the questions can change with the role. A warehouse candidate might be asked about certifications, location and shift availability, while a software candidate might be asked about languages and frameworks. The agent can also parse resumes, provide reminders and interview-preparation information, and schedule or reschedule appointments.

Matching draws on Eightfold’s Talent Intelligence Platform, which the company says incorporates more than 1.6 billion career trajectories. Eightfold positions this as a way to recommend jobs based on skills and potential rather than relying only on title and keyword matches.

How it connects to Eightfold’s other agents

Candidate Agent is intended to pass a candidate into AI Interviewer while preserving the conversational context. Eightfold describes Candidate Agent as handling engagement and coordination, AI Interviewer as handling structured evaluation, and human recruiters as retaining hiring decisions.

The company is also introducing Avatar, a digital-human interface for AI interviews. Avatar and a planned 360 Interview capability are associated with Talent Agents 2.5 and are expected to become generally available at the end of July. They are therefore separate from the Candidate Agent availability confirmed in the 2.0 announcement.

Eightfold reports that early users of AI Interviewer reduced hiring cycles from 42 days to less than a week and cut time to interview by as much as 90 per cent. These are company-reported outcomes associated with early adopters, not a guarantee that Candidate Agent will produce the same results for every organisation.

Governance questions remain important

Candidate-facing automation touches personal information, employment decisions and potentially sensitive inferences. Buyers should establish which actions the agent may take without human approval, how explanations and audit logs are retained, and how candidates can correct information or request a human response.

Eightfold says agent actions are logged, explainable and auditable, and that its matching model has undergone an independent bias audit under New York City’s Local Law 144. Australian employers will still need to assess the product against their own privacy, discrimination, workplace and record-keeping obligations, as well as rules in every jurisdiction where it is used.

Practical evaluation should include accessibility, language quality, escalation behaviour, data retention, integration with applicant-tracking systems and the risk that conversational convenience obscures an automated assessment. Organisations should also test whether the agent maintains context accurately and avoids making commitments about roles or process that recruiters have not authorised.

The candidate experience needs its own success measures. Faster applications and fewer recruiter tasks are useful, but employers should also track completion rates, candidate complaints, requests for human assistance, accessibility failures and differences in outcomes across groups. A quicker process is not automatically a fairer or more informative one.

Channel support creates additional operational choices. SMS and WhatsApp can make the process more accessible, yet they also create consent, identity, security and message-retention questions. Employers should make it clear when a candidate is interacting with AI, explain how information will be used and provide a practical route to continue outside the automated channel.

Implementation teams should define the source of truth for role requirements, status updates and scheduling. If Candidate Agent receives stale vacancy data or conflicting instructions from connected systems, a fluent conversation could still produce a poor outcome. Permission boundaries, integration monitoring and a tested manual fallback are therefore central to the deployment.

The launch reflects a broader move from AI that recommends actions to AI that executes workflow steps. Eightfold’s release may reduce routine coordination work, but its value will depend on controlled configuration, measurable candidate outcomes and clear human accountability.