Questions to ask when evaluating Decision Agent

What does the platform assess?

Decision Agent helps hiring teams collect evidence through questionnaires, text scenarios and voice conversations. Recruiters define the role requirements, behaviours and objectives, with assistance from the platform. A bank might explore judgement under competing priorities; a manufacturer might explore how someone handles a safety concern. These are possible assessment designs, not claims of validation for those sectors.

The sample reports are fictional, read-only demonstrations. They show the form of the evidence, not real candidate records or customer endorsements.

Can a recruiter understand and challenge a finding?

Yes. Scenario and Conversation reports connect findings to responses. Conversation evidence includes transcript passages and playable clips. Recruiters can inspect the exchange and correct findings in the UI, affecting that report.

A displayed score combines findings using fixed rules and adjustable weights. The AI interpretation behind a finding can still be wrong, and a transcript can contain errors. Read how to review a report before treating a score as a decision.

What can be automated, and who decides?

Configured questionnaire criteria can automate screening steps, and the Pinpoint integration can carry configured outcomes into the ATS. Question-specific automation is a managed process using recruiter-labelled answers and a validation target agreed in advance. It is not automatic validation of an entire hiring process.

The recruiter is responsible for the requirements, review arrangements and exceptions. Significant hiring decisions need the human oversight required by the applicable agreement and law. Read the human-review guide and agree which steps will be automated before launch.

Is it validated or independently tested for bias?

We do not currently claim that the platform makes validated predictions of suitability or job performance. Question-level validation and inspecting behaviour findings are narrower tasks. They should not be presented as proof that a hiring process is fair for every group or effective in every role.

We do not publish an independent product-wide bias-audit report or a sector-specific predictive-validation study on this site. If you need evidence for a particular population or intended use, ask what is available and agree a suitable pilot. Our validation article explains why the claim, population and intended use must be specified.

What governance and security information can procurement request?

Ask for the proposed DPA, relevant security measures, supplier and transfer information, and the governance documents needed for your intended use. Our enterprise contracting can address confidentiality, rights assistance, breach notification, audit information and customer-data separation. The exact commitments belong in the agreement.

We use internal privacy and security policies, access controls and monitoring. A supplier’s certification is not a certification of Decision Agent. This site does not establish a Decision Agent SOC 2 or ISO 27001 certification, a completed independent employment-bias audit, or approval for a particular regulated use.

For candidate access, ask about supported languages and devices, accessibility testing, cross-accent assessment evidence and suitable alternative formats. We do not publish a WCAG conformance report or a complete language list here.

Articles discussing the EU AI Act or other regulation explain our thinking; they are not a certificate of compliance for your deployment. A bank, healthcare employer or international business should specify its required assurance, documentation and review process. See legal resources and data handling.

What supports the figures on the solution page?

The figures on our solution page—81% lower screening costs, a threefold time-to-hire improvement and 96% candidate satisfaction—were reported from one launch-client trial involving around 3,000 applicants, compared with its previous screening services. They describe that deployment.

They are not independent proof of assessment validity or a forecast for another customer. We do not publish the full study method and underlying dataset here. Ask about the comparison, measurement period and definitions before using the figures in a business case. The brochure’s workload calculator is an illustration, not another measured customer result.

What should an enterprise proposal cover?

The pricing page gives the current starting offer and module credit costs. Larger volumes, team arrangements, integration work and bespoke requirements need an enterprise proposal. Agree credit terms, rollout, training, support coverage, response targets and any uptime or recovery commitments in writing.

We do not publish a universal enterprise uptime SLA, throughput guarantee or implementation timetable. For 1,000 or 10,000 applicants, send the expected volumes and assessment mix rather than extrapolating the starting subscription into a volume price.

Pinpoint is available today. Other connectors, a general API, SSO, regional processing options and particular audit/export requirements should be confirmed in scope rather than assumed. Send your requirements to team@decision-agent.ai.