Section 6: Future of Assessment

Section 6: Future of Assessment

Sections 1–5 documented the current state: personality testing has become hiring infrastructure, regulatory pressure is mounting, and adoption remains gated by cost and candidate friction. But the landscape is shifting. Between May 2026 and the end of 2027, three forces will reshape assessment: (1) regulatory divergence between the EU, US states, and global laissez-faire jurisdictions; (2) AI's integration into scoring and item generation, with growing skepticism of LLM-based grading; (3) a parallel momentum toward skills-based filtering over trait-based prediction.

This section bridges compliance, technology, and market prediction. We map the regulatory landscape through 2027, analyze where AI actually helps and where it creates liability, and offer five specific, falsifiable predictions grounded in precedent and trajectory.

Regulatory Landscape 2026–2030: From Compliance to Enforcement

EU AI Act: Full Enforcement Timeline

The EU AI Act (Regulation 2024/1689) classifies hiring assessments as “high-risk AI.” The timeline is unambiguous:

• February 2, 2025: Act entered into force for all provisions except Chapter 2 (high-risk system requirements).
• August 2, 2026: Chapter 2 enforcement begins. Vendors and employers deploying assessment systems must have completed risk assessments, training data audits, and bias monitoring protocols. This applies to all EU-facing assessment platforms, whether headquartered in EU or not.
• Fines: €30 million or 6% of global annual turnover — whichever is greater. For Workday, HRIS, or mid-tier vendors, this is an existential penalty.

By August 2026, expect the first wave of compliance certifications. Vendors with published bias audits and documented governance will gain competitive advantage through risk reduction. Those without will face either rapid remediation (costly) or market exit from EU.

US State Law Proliferation

The US federal government has issued guidance (EEOC 2022–2023, DOJ 2024, OFCCP 2024) but no federal mandate yet. Instead, state legislatures are acting:

• NYC Local Law 144 (effective Jan 1, 2024): Enforcement is now live. We predict NYC DOC will issue guidance clarifying the 80% disparate-impact threshold and audit standards by Q3 2026.
• Illinois HB 5053 (effective Jan 1, 2024): Requires informed consent for automated screening. Likely expansion to other Midwest states by 2027 (Wisconsin, Minnesota legislative interest noted).
• California SB 1162 (effective Jan 1, 2025): Mandates disclosure of automated systems in job postings and candidate communications. Will trigger similar transparency laws in Colorado, Washington, Massachusetts.
• Federal floor: If EEOC enforcement accelerates (complaint volume +45% YoY in 2024), expect Senate commerce committee hearings on algorithmic hiring by mid-2027. A federal mandate remains unlikely before 2028, but guidance expansion is certain.

International Divergence

EU will enforce strictly. UK ICO guidance (June 2024) is lighter-touch but signals convergence toward documentation and bias testing. Asia (Singapore, Tokyo) and Latin America (Brazil, Mexico) remain laissez-faire for now, creating arbitrage opportunities for vendors targeting non-compliant markets.

Emerging Assessment Frameworks: Skills vs. Traits

A fundamental shift is underway. Rather than trait-based prediction (MBTI, Big Five as job-fit proxies), employers are moving toward skills-first hiring. This is not new, but the velocity is accelerating.

Skills-Based Hiring: The Shift

LinkedIn Talent Solutions data (2026) shows skills-based job filters growing from 12% of posts (2024) to 23% (2026). This trend reflects employer skepticism of personality-to-job-performance pipelines and compliance anxiety around bias in trait assessments. The implication: employers want direct evidence of capability, not inferred trait profiles.

This favors:

• Work sample tests (coding problems, design briefs, case studies)
• Structured interviews (behavioral and situational)
• Skill-based assessments (Excel, Python, technical writing)
• Credential verification and portfolio review

Personality tests, by contrast, face headwinds. Employers deploying Big Five or DISC without explicit validation data risk regulatory challenge. MBTI, already weak on job-prediction, is increasingly repositioned as “self-discovery” (JobCannon's framing) rather than hiring signal.

Adaptive Testing (CAT): Mainstream by 2027

Computerized Adaptive Testing (CAT) is not new—it powers the GRE and military screening—but adoption in hiring assessments has been slow. This is changing. By 2027, expect 35–40% of enterprise vendors to offer CAT or CAT-adjacent technology. Advantages:

• Reduced test duration (candidate experience improves)
• Increased precision (fewer easy/hard items wasted)
• Lower dropout rates (shorter tests = more completions)

Pymetrics and Plum, already deploying adaptive logic, will see accelerated adoption. Traditional linear tests (fixed-length, all-candidates-same-items) will be repositioned as “entry-level” or “research” offerings.

Multi-Modal Assessment: Text + Voice + Behavior

Some vendors (HireVue, Pymetrics) have experimented with video+audio analysis. The reality: construct validity is poorly documented, and regulatory risk is high. Expect a bifurcation: (1) North American vendors continue experimenting with video analysis despite lack of peer-reviewed support; (2) EU vendors de-emphasize video analysis due to GDPR Article 22 + AI Act liability. By 2027, multi-modal approaches will be niche, not mainstream.

AI's Actual Role: Where It Helps, Where It Harms

Large language models are reshaping assessment development and scoring. But the impact is mixed. Here's where AI clearly helps and where it creates risk:

Where AI Clearly Helps

• Item generation: LLMs can draft assessment items quickly, with human review. A 50-item Big Five assessment that once took 40 hours of psychologist labor can now be drafted in 4 hours + 8 hours review. Cost reduction is real.
• CV parsing: Extracting skills, experience, education from resumes via LLM is faster and more accurate than keyword regex. No liability if used for routing (enrichment), not decision-making (screening).
• Response summarization: Open-ended essay responses can be summarized by LLM before human review. Faster review cycles, same human decision-making.

Where AI Is Unproven or Risky

• LLM-graded open-ended scoring: Some vendors propose using GPT-4 to grade essay responses on a rubric. The problem: LLM grading accuracy is not peer-reviewed; consistency across diverse writing styles is undocumented; and regulatory frameworks (NYC LL 144, EU AI Act) expect explainability. If an LLM-graded score screens out a candidate, can you explain why? Not reliably.
• Scoring “cultural fit" or “team alignment": Some platforms use LLMs to infer soft traits from responses. This is liability-rich. Soft traits are hard to operationalize, and LLMs amplify demographic bias in training data (female candidates penalized for “assertiveness,” racial minorities penalized for non-standard dialect, etc.). Expect enforcement actions against vendors making vague “fit” predictions by 2027.
• Automated item difficulty calibration: Using LLMs to adjust item difficulty on-the-fly is interesting but untested. Unintended consequence: if the algorithm systematically eases items for protected groups, you've created demographic parity but destroyed criterion validity. Not worth the risk until peer-reviewed.

What We Predict for 2027–2028: Five Falsifiable Predictions

Below are five specific, dated, measurable predictions. Each is grounded in regulatory trajectory, vendor behavior, and market data. We assign confidence (high/medium/low) and reasoning.

Prediction 1: EU AI Act Enforcement Pushes Mid-Tier Vendor Compliance Spend to 4–8% of Revenue

By December 31, 2027: Mid-tier assessment vendors (annual revenue €50M–€500M) with EU-facing products will be spending 4–8% of revenue on EU AI Act compliance (audit, documentation, tooling, legal). Demand will flow to audit firms (Deloitte, EY, PwC), assessment platforms (Pymetrics, Plum, Workday), and specialized compliance-tech startups building out the services.

Confidence: High (75%). Rationale: €30M fine (or 6% of global turnover) creates existential incentive to fund compliance at that scale. Enforcement action by 1–2 vendors in late 2026 will accelerate spend in 2027.

Prediction 2: NYC LL 144 Enforcement Expands to 60% of Identified AEDS Vendors

By September 30, 2027: NYC Department of Consumer Affairs will have issued audit or compliance letters to 60% of the 54 identified AEDS vendors (32 vendors). By this date, 45–50% will have completed at least one bias audit.

Confidence: Medium (65%). Rationale: As of June 2024, only 37% (20 vendors) had published bias audits. NYC enforcement has been deliberate but not aggressive; agencies typically ramp enforcement after 18–24 months of the law's effective date. By September 2027 (45 months post-enactment), escalation is likely. Unverified: whether NYC will issue penalties or focus on remediation notices.

Prediction 3: One Major Vendor Faces EEOC or Mobley v. Workday Settlement by Mid-2027

By June 30, 2027: A second major vendor (other than Workday) will announce a settlement or have one imposed in an EEOC action or class action related to hiring assessment bias. Likely candidates: HireVue (video analysis age/race bias), Hogan Assessments (test-taker demographic parity claims), or a mid-tier vendor via NYC LL 144 audit findings.

Confidence: Medium-High (68%). Rationale: iTutorGroup settled in March 2023 for $365K. Mobley v. Workday certified in February 2025 affecting 200M+ applicants. The EEOC complaint volume for AI hiring increased 45% YoY in 2024 (88 → 127 complaints). Extrapolating, a second settlement by mid-2027 is statistically probable. The question is magnitude: likely range €500K–€5M depending on class size.

Prediction 4: Skills-Based Hiring Posts Reach 35% of US Tech Job Market by Q4 2027

By December 31, 2027: Skills-first job posts (filtering by skill tags, not degree) will comprise 35% of job postings on LinkedIn, Indeed, and Greenhouse in the US tech sector (up from 23% in Q2 2026).

Confidence: Medium (60%). Rationale: LinkedIn data shows 23% in 2026 (up from 12% in 2024), a near-doubling in 18 months. Extrapolating linearly, 35% by Q4 2027 is reasonable. However, this assumes sustained momentum; if regulatory friction on trait-based tests increases faster, adoption could accelerate. Downside: if recession hits in 2027, employers may revert to credential filtering.

Prediction 5: Consumer-Priced Assessment Platforms Reach 500K+ Annual US Users by December 2027

By December 31, 2027: Consumer-facing assessment platforms (JobCannon, Hirable, Truity, etc.) will cumulatively serve 500K+ annual active users in the US. This represents adoption of self-directed testing at scale, outside employer-mandated channels.

Confidence: Medium (62%). Rationale: JobCannon alone is tracking ~70K–80K annual tests (as of mid-2026). The self-directed market is fragmented across 8–10 platforms. Estimated aggregate 2026 volume: 250K–300K. Growth rate: 40–50% YoY. By end of 2027, 500K+ is attainable if user acquisition holds. Downside: if paid tiers face pricing resistance (candidate friction remains high), growth could plateau at 350K.

What We Still Don't Know: Limitations of This Report

This report is constrained by available data and scope. Three major unknowns linger:

• Long-term job-performance outcomes: We have no multi-year longitudinal study linking personality test scores taken in 2023–2024 to job performance and retention in 2026–2027. Any claim about “true” ROI is speculative.
• LLM-grading validity: No peer-reviewed study compares human-graded vs. LLM-graded open-ended responses on any hiring assessment. Vendor claims are marketing.
• Non-English-speaking markets: This report focuses on US and UK. Assessment adoption in non-English markets (India, Brazil, Southeast Asia) is largely unmeasured.

Future research should prioritize these gaps. Regulatory enforcement will outpace research; vendors will make claims ahead of evidence. Practitioners should demand peer-reviewed data before adopting new scoring methods.

How to Cite This Report

APA Format (7th Edition):

JobCannon Research. (2026). State of Personality Testing 2027: An evidence-based analysis of adoption, regulation, outcomes, and bias. Retrieved from https://jobcannon.io/state-of-personality-testing-2027

BibTeX Format:

@report{jobcannon2026stateofpt,
  title={State of Personality Testing 2027: An evidence-based analysis of adoption, regulation, outcomes, and bias},
  author={JobCannon Research},
  year={2026},
  month={May},
  url={https://jobcannon.io/state-of-personality-testing-2027},
  note={Published May 17, 2026}
}

DOI (Digital Object Identifier): If citing this report in an academic database, use: https://doi.org/10.59921/jobcannon-pt-2027 (placeholder DOI for future indexing)

About This Report

The State of Personality Testing 2027 is a six-section, 18,000-word evidence-based analysis of the personality assessment landscape in 2026. This report synthesizes data from SHRM surveys, Gartner research, regulatory filings, peer-reviewed studies, and JobCannon's internal platform analytics. Sections 1–5 document adoption volume, regulatory frameworks, hiring outcomes, bias, and adoption barriers. Section 6 forecasts the future through 2027 and beyond.

This report is published as a research resource for HR leaders, assessment vendors, policy makers, and researchers. Findings are accurate as of May 17, 2026, and should be read in context of the report's limitations (noted in Section 6).

Take Your Assessment

This report was written to inform and empower professionals making assessment decisions. If you're exploring your own personality, career fit, or working style, JobCannon offers free personality assessments grounded in peer-reviewed psychology.

Explore our full assessment library — Big Five, MBTI-style typology, RIASEC career matching, EQ, DISC, and more. All tests are free to start; premium deep-dives available for subscribers.