Skip to content
GLOBALANCE INTELLIGENCEThe economy ahead.
This Week Index 28/100 Sign in
Money & Investments › GI 100 Index
The Great Value ShiftInvestment ThemesGI 100 IndexGI PortfolioStart Future Check
AI FOOTPRINTsolidASSESSMENT 2026GLOBALANCEINTELLIGENCE
Globalance AI Footprint

Assessment as of November 2026 · next assessment 2027.

Result of the assessment by the Globalance research team under the published methodology. Not a certification or quality label.

How we assess ›

How this rating is builtVideo · 0:52
GI 100 Index · Profile

Automatic Data Processing

USAAdoptersAssignment FullData quality Low–medium

Key messagesolid AI Footprint

ADP has anchored oversight of artificial intelligence with its board, describes formal reviews of new AI applications before launch in its corporate responsibility report, and explains unusually openly how its tool for assessing job applications works and that an independent audit found no disadvantage to any group. For the many new AI assistants in payroll and HR processes, however, the company has so far provided hardly any verifiable information on testing, limitations and the data used.

Opportunities

  • Responsibilitysolid

    The 2026 proxy statement documents AI-specific board oversight of the AI governance programme, the charter of the Corporate Development and Technology Committee amended in April 2026, and a responsible AI programme led by the Chief Data Office with an AI and Data Ethics Council since 2019 (S03, S01).

    The 2026 proxy statement documents AI-specific board oversight of the AI governance programme, the charter of the Corporate Development and Technology Committee amended in April 2026, and a responsible AI programme led by the Chief Data Office with an AI and Data Ethics Council since 2019 (S03, S01). The 2025 corporate responsibility report, a formal company report, describes a concrete review procedure with formal AI reviews at proof-of-concept, pilot and release stage, implementation of the NIST AI RMF and monitoring after launch (S12); in addition, there is the periodic independent bias audit of candidate scoring with an announced follow-up audit (S15, S14). Both minimum conditions are therefore met; still missing are a published AI policy with risk classes, a described approval procedure for purchased AI and partner agents (S23, S24), AI training figures and an external audit of AI management, and the report itself is not externally assured (S12). Hence “solid”, not “strong”.

  • Safetysolid

    Data and system protection is documented concretely: Item 1C of the 10-K describes board oversight, quarterly reports and an annual third-party review against the NIST CSF (S01), ADP names ISO 27001/27701 certificate numbers from a certification body (S29) and runs a vulnerability disclosure …

    Data and system protection is documented concretely: Item 1C of the 10-K describes board oversight, quarterly reports and an annual third-party review against the NIST CSF (S01), ADP names ISO 27001/27701 certificate numbers from a certification body (S29) and runs a vulnerability disclosure programme with a safe harbour commitment (S17). For candidate scoring, tests and bias measurements are documented (S14, S15). AI-specific checks such as red-teaming, error rates of the payroll agents or misuse monitoring for ADP Assist are not documented (S19, S31, S26). The only controversy is a data leak at a business partner without an AI link (minor, S11).

  • Transparencysolid

    For candidate scoring, ADP discloses an unusual amount: how it works, the data basis, retention, the base models used (JobBERT-v3, OpenAI text-embedding-3-large), sub-scores, limitations, bias figures and an opt-out (S14, S15).

    For candidate scoring, ADP discloses an unusual amount: how it works, the data basis, retention, the base models used (JobBERT-v3, OpenAI text-embedding-3-large), sub-scores, limitations, bias figures and an opt-out (S14, S15). For ADP Assist and the other agents, by contrast, there is only a general, undated transparency notice without model names, information on training with client data or a labelling rule (S27, S21). The rating “solid” therefore rests essentially on one product with high harm potential and is a borderline judgement.

Risks

  • Risk area

    AI-based candidate scoring and AI interview agent affecting access to work

  • Risk area

    AI agents in payroll and tax processes with a direct effect on wages

  • Risk area

    Protection of very large personal data sets that are also used for AI training

  • Risk area

    Purchased models and third-party agents in the ADP Marketplace

  • No serious controversy recorded.

What lies behind the assessment

Research of 3 November 2026
40searches
32pages and documents read
10primary documents
1controversies checked
5open points, named openly
The full research: pillars, evidence, controversies and open points
AI Footprint · Adopter · As of 3 November 2026
Automatic Data ProcessingOverall rating: solid

ADP has placed oversight of AI with its board, describes formal reviews of new AI applications and discloses its candidate scoring including a bias audit. Information on testing, limitations and data of the new AI assistants is missing.

Responsibility

40 %
limitedsolidstrong

The board oversees AI, and new applications pass through formal review stages.

  • +Board oversees the AI governance programme Source ↗
  • +Formal AI reviews at proof-of-concept, pilot and release stage Source ↗
  • +Independent bias audit of candidate scoring by BLDS Source ↗
  • –No published AI policy with risk classes found
  • –No AI training figures and no certification of AI management
All sources for this pillar (11)

Safety

40 %
limitedsolidstrong

Data and system protection are well documented, but AI-specific testing of the assistants is not.

  • +Annual third-party review against NIST CSF and quarterly board reports Source ↗
  • +ISO/IEC 27001 and 27701 with certificate numbers Source ↗
  • +Vulnerability reporting channel with safe harbour commitment Source ↗
  • –No red-teaming and no error rates documented for the AI agents
  • –Data leak at business partner Business Systems House in 2024 Source ↗
All sources for this pillar (15)

Transparency

20 %
limitedsolidstrong

Candidate scoring is openly documented, but model and data information is missing for ADP Assist.

  • +Base models, data and retention of candidate scoring disclosed Source ↗
  • +Bias figures from the independent audit published Source ↗
  • –Only a general, undated transparency notice for ADP Assist Source ↗
  • –No rule on labelling AI-generated content found
All sources for this pillar (7)
What is missing for “strong”

These steps are in the company’s own hands.

1Responsibility
Publish an AI policy with risk classes

Publish an AI policy with risk classes and an approval procedure that also covers purchased AI, report training figures and have AI management externally audited.

2Safety
Disclose testing of the AI agents

Document red-teaming, error rates and misuse monitoring for ADP Assist and the payroll agents.

3Transparency
Publish system cards for ADP Assist

Disclose models, training data and labelling for ADP Assist and the AI interview agent.

Context: 101 companies assessed

Distribution of overall ratings, no ranking
strong
none yet
solid
← Automatic Data Processing is one of 35
limited

No company reaches “strong” today. Disclosure on how companies handle AI is still young across the market.

Controversies
1 1 minor

No deduction for controversies.

Largest risk area
Discrimination by AI in recruiting

The new “Recruiting AI Interviewer Agent” is not yet described in the transparency documents read, and critics point to gaps in the bias audit of candidate scoring.

How thoroughly assessed
33 Evidence

10 primary documents read, 40 searches.

Read the full reasoning and all sources
Automatic Data Processing · Full reasoning

Board oversight of AI and open candidate scoring, but little on the AI assistants

Overall rating “solid”. No deduction for controversies. The superscript numbers link directly to the source.

1 · Responsibility

solidWeight 40%

ADP documents board oversight of AI and describes formal reviews of new AI applications, but publishes no AI policy with risk classes.

What is documented
  • The board oversees the AI governance programme; the charter of the Corporate Development and Technology Committee was amended in April 20261
  • A responsible AI programme under the Chief Data Office and an AI and Data Ethics Council have existed since 20191 2
  • The 2025 corporate responsibility report describes formal AI reviews at proof-of-concept, pilot and release stage, implementation of the NIST AI RMF and monitoring after launch3
  • A cross-functional working group reviews AI use cases; the Audit Committee oversees data protection2 1
  • In April 2024, BLDS independently audited candidate scoring by gender, ethnicity and intersections and published figures11
  • Internal tests with more than 48,000 CVs are described, and a follow-up audit has been announced10
The case against
Gaps in policy and purchased AI
  • No published AI policy with risk classes was found
  • Marketplace partners commit to ADP principles, but no review procedure for this is described8 9
  • The corporate responsibility report is not externally assured3
Missing figures and certificates
  • AI training figures and a certification to ISO/IEC 42001 or similar are missing
  • No general AI complaints channel with documented remedy is described
Sources for Responsibility (11)
  1. Automatic Data Processing, Inc. / U.S. SEC: Definitive Proxy Statement 2026 (annual general meeting 2026-11-11) · 24 September 2026 · primary source
  2. Automatic Data Processing, Inc. / U.S. SEC: Form 10-K for the fiscal year ended 2026-06-30 · 5 August 2026 · primary source
  3. ADP, Inc.: Our Corporate Responsibility: Leadership by Design 2025 · November 2025 · primary source
  4. AI Business (Informa TechTarget), Esther Shittu: How payroll giant ADP practices AI governance and human oversight · 21 October 2026 · secondary source
  5. InformationWeek, Joao-Pierre S. Ruth: How ADP's chief AI officer keeps a tight rein on AI without holding it back · 28 July 2026 · secondary source
  6. Automatic Data Processing, Inc. / U.S. SEC: Annual Report to Stockholders fiscal year 2026 (Form ARS) · September 2026 · secondary source
  7. ADP, Inc.: ADP AI Ethics Statement · 7 March 2022 · secondary source
  8. CPA Practice Advisor: ADP Adds AI Partner Solutions to its ADP Marketplace · 22 February 2024 · secondary source
  9. Rallies.ai: ADP Introduces Curated AI Agents in Marketplace with 10 Partners for HR Automation · 2 March 2026 · secondary source
  10. ADP, Inc.: Profile Relevance FAQ and AI Transparency Notice (Last Updated July 9, 2026) · 9 July 2026 · primary source
  11. ADP, Inc.: Candidate Relevancy FAQ (Last Updated December 11, 2024) · 11 December 2024 · primary source

2 · Safety

solidWeight 40%

ADP documents board oversight, an annual third-party review, ISO certificates and a vulnerability reporting channel, but red-teaming and error rates of the AI agents are missing.

What is documented
  • Item 1C of the 10-K describes a CISO, a CSO, quarterly reports to the board, an annual third-party review against the NIST CSF and emergency exercises7
  • ADP names certificate numbers for ISO/IEC 27001 and 27701 and carries out privacy impact assessments8
  • A vulnerability reporting channel commits to taking no legal action against researchers who follow the rules12
  • Internal bias tests and an independent audit are documented for candidate scoring1 2
  • Incident registers list no incidents for 2023 to 2026, and there is no filing under Item 1.0513 14
The case against
Missing evidence of AI testing
  • Test methods, error rates and red-teaming for the generative ADP Assist agents and the payroll agents are not documented
  • The Chief AI Officer describes guardrails and a restriction to reliable tasks without verifiable details4
  • SOC 2 reports are available only under a non-disclosure agreement9
Misuse and supply chain
  • Usage restrictions and misuse monitoring for the AI agents are not described
  • After a ransomware attack on partner Business Systems House, personal data of employees was published in 202415
Sources for Safety (15)
  1. ADP, Inc.: Profile Relevance FAQ and AI Transparency Notice (Last Updated July 9, 2026) · 9 July 2026 · primary source
  2. ADP, Inc.: Candidate Relevancy FAQ (Last Updated December 11, 2024) · 11 December 2024 · primary source
  3. ADP, Inc.: Our Corporate Responsibility: Leadership by Design 2025 · November 2025 · primary source
  4. InformationWeek, Joao-Pierre S. Ruth: How ADP's chief AI officer keeps a tight rein on AI without holding it back · 28 July 2026 · secondary source
  5. ADP (Hong Kong, press release): ADP integrates new AI agent to elevate Global Payroll accuracy and efficiency · 13 April 2026 · secondary source
  6. ADP, Inc. (press release): ADP Accelerates AI Leadership with Launch of New AI Agents Designed to Solve Workforce Challenges · 28 January 2026 · secondary source
  7. Automatic Data Processing, Inc. / U.S. SEC: Form 10-K for the fiscal year ended 2026-06-30 · 5 August 2026 · primary source
  8. ADP, Inc.: Privacy at ADP · 3 November 2026 · primary source
  9. ADP, Inc.: A partner you can trust (data security) · 8 June 2026 · secondary source
  10. ADP, Inc. (press release): ADP Ranks Among Elite Handful of Companies Worldwide with Approved Binding Corporate Rules for Global Data Protection · 28 February 2018 · secondary source
  11. ADP, Inc. (press release): ADP Assist with Generative AI Features Makes HCM Decisions Easy, Smart and Human · 31 January 2024 · secondary source
  12. ADP, Inc.: Vulnerability Disclosure · 3 November 2026 · primary source
  13. Board Cybersecurity: AUTOMATIC DATA PROCESSING INC Cybersecurity Tracker · 3 November 2026 · secondary source
  14. Rankiteo: ADP – Cyber Incident History · 3 November 2026 · secondary source
  15. The Register: Broadcom employee data stolen by ransomware crooks following hit on payroll provider · 16 May 2025 · secondary source

3 · Transparency

solidWeight 20%

ADP discloses models, data, limitations and bias figures for candidate scoring, but only a general transparency notice for its other AI agents.

What is documented
  • The transparency notice on candidate scoring names its function, data basis, three-year retention and the base models JobBERT-v3 and OpenAI text-embedding-3-large1
  • Sub-scores for experience, role and skills, the absence of exclusion thresholds, limitations and user obligations are described1 5
  • Candidate Relevancy was trained on earlier selection decisions; the bias figures are published5
  • The 10-K names risks arising from the use of AI3
The case against
Gaps for ADP Assist
  • The general transparency notice names no models and does not clearly govern the use of client data for training2
  • ADP trains its own AI on its HCM data set; there are no model or system cards for this6
  • Whether applicants are informed about the new AI interview agent is not documented7
Criticism of candidate scoring
  • A critic points to few technical details, no derivation of the weights and the exclusion of small groups4
  • An opt-out cannot be distinguished from a technical failure by employers1 4
Sources for Transparency (7)
  1. ADP, Inc.: Profile Relevance FAQ and AI Transparency Notice (Last Updated July 9, 2026) · 9 July 2026 · primary source
  2. ADP, Inc.: Artificial Intelligence Transparency Notice · 3 November 2026 · secondary source
  3. Automatic Data Processing, Inc. / U.S. SEC: Form 10-K for the fiscal year ended 2026-06-30 · 5 August 2026 · primary source
  4. Dr. Chao Wu (blog, Maryland Delegate): AI hiring disclaimer from ADP, Part 1/2 · 14 August 2025 · secondary source
  5. ADP, Inc.: Candidate Relevancy FAQ (Last Updated December 11, 2024) · 11 December 2024 · primary source
  6. ADP, Inc. (press release): ADP Assist with Generative AI Features Makes HCM Decisions Easy, Smart and Human · 31 January 2024 · secondary source
  7. Automatic Data Processing, Inc. / U.S. SEC: Form 10-Q for the quarter ended 2026-09-30 (Q1 fiscal year 2027) · 30 September 2026 · primary source

4 · Why the harm rule does not apply

The harm rule does not apply: in the research period no AI-related judgment, no AI fine and no pattern of serious incidents was found.

5 · Controversies (1)

6 · What remains open

  • Testing, validation and red-teaming procedures for the generative ADP Assist agents
  • Labelling of AI-generated content and information of those affected by the AI interview agent
  • Rules on the use of client data for model training
  • AI figures on training, incidents and complaints and an external audit of the responsible AI programme
  • Full report of the BLDS bias audit and results of the announced follow-up audit

7 · How it was assessed

Cut-off date: 3 November 2026 · 40 searches, 32 fetches.

Primary documents read (10)
  • Form 10-K fiscal year 2026 (Items 1, 1A, 1C, 3; two retrievals)
  • Definitive Proxy Statement 2026 (three retrievals)
  • Definitive Proxy Statement 2025 (extract)
  • Form 8-K Item 5.07, November 2024
  • Form 10-Q as at 2026-09-30 (through Note 13 and MD&A)
  • Annual Report to Stockholders 2026 (letters)
  • Corporate Responsibility Report 2025 (two retrievals)
  • ADP AI Ethics Statement 2022
  • Profile Relevance FAQ (2026-07-09) and Candidate Relevancy FAQ (2024-12-11)
  • AI Transparency Notice, privacy, data security and vulnerability disclosure pages

Research of 3 November 2026 under Globalance AI Footprint rulebook version 0

We assess documented practice in dealing with AI; this is not an investment recommendation. Three pillars, weights 40/40/20, harm rule.

Research of 3 November 2026 under Globalance AI Footprint rulebook version 0. Documented observations from public sources, no investment recommendation.

Important Notice. This material is provided for information purposes only and does not constitute an offer or solicitation to buy or sell any financial instrument, nor does it constitute investment advice or a recommendation. The GI 100 Index is a rules-based strategy index of Globalance Bank. Investments involve risk, including the loss of capital invested. The final product documents (including FinSA) shall prevail in all cases.

Your company? Information on inclusion in the GI 100 Index →

What comes next