KPMG International: The Firm That Certified Accountable AI Could Not Satisfy a Single Accountability Condition on Its Own Deployment
A public-record application of the Vordan Agentic Accountability Baseline
PublishedJune 22, 2026
Assessed EntityKPMG International
Deployment in ScopeAI Research & Citation Workflow | Total Experience Report
InstrumentAAB v0.2 (Public-Record Application)
Assessment IDAAB-PR-001
Evidence BasisPublic Record Only
AssessorVordan
Context
In October 2025, KPMG International published "Total Experience: Redefining Excellence in the Age of Agentic AI," a flagship global report on customer experience and AI transformation. The report was produced using an AI-assisted research and citation workflow. On June 12, 2026, AI detection firm GPTZero published a forensic review finding that only five of the report's 45 citations accurately pointed to real, uncorrupted sources. Forty of 45 citation titles were fabricated, paraphrased beyond recognition, or unverifiable. GPTZero described the pattern as consistent with an LLM research tool prompted to find case studies of agentic AI in the wild, producing what it named "vibe citing", the citation equivalent of vibe coding, where a model stitches fragments of real sources together and invents the rest.
On June 13, 2026, KPMG removed the report from its websites. Named organizations including UBS, the UK's National Health Service, Swiss Federal Railways, and Transport for London disputed the claims made about their AI deployments. KPMG stated it was "reviewing the circumstances surrounding its publication" and that employees "may have breached internal AI-use policies." GPTZero's conclusion was direct: "We suspect no human at KPMG double-checked the citations, the claims, or the sources before Total Experience was published."
This document applies the Vordan Agentic Accountability Baseline (AAB v0.2) to the specific agentic AI deployment that produced this outcome: KPMG's AI-assisted research and citation workflow. The deployment in scope is not KPMG's broader technology estate. It is the autonomous workflow that searched, synthesized, and produced 45 citations for a global publication, without, by all available evidence, any functioning human authorization, oversight, or verification gate.
The Accountability Context That Makes This Finding Significant
KPMG International achieved ISO 42001:2023 certification in December 2025, the world's first international standard for AI Management Systems, certifying a robust system of controls to ensure accountability and trust while mitigating risk in AI deployment. KPMG US achieved the same certification in November 2025. Both certifications were obtained after the October 2025 report was deployed and before the June 2026 retraction.
On April 22, 2026, seven weeks before the retraction, KPMG announced a suite of new AI agents powered by Google Cloud's Gemini Enterprise, describing its Trusted AI framework as providing guardrails ensuring systems "act within clear ethical and operational boundaries." KPMG had by that date deployed Gemini Enterprise to more than 55,000 professionals globally and built more than 150 agents for clients, positioning itself as a firm that had successfully governed AI at enterprise scale.
The firm that certifies AI management systems, sells AI governance advisory to the market, and publicly promotes its accountability architecture scored 0 out of 100 on the AAB when that architecture was applied to its own internal agentic deployment. This is the finding.
Central Finding
Accountability Architecture Insufficient: 0/100
All eight AAB v0.2 conditions are Not Satisfied. No condition is Partially Satisfied. Four conditions carry Critical gap severity. Four carry High. Seven of eight gaps are Structural, meaning they cannot be closed by policy updates alone. The architecture required to satisfy each condition was not present in the deployment. A firm cannot retract its way into accountability. The governance infrastructure that would have prevented this outcome was absent at the point of deployment, not at the point of discovery.
Primary Gap: Authorization Chain Absent
No human authorization record for the research workflow has been identified in the public record. The accountability chain begins with the agent, not with a human decision. KPMG's own guidelines required human validation of AI-generated content. The guidelines were the policy. They were not the authorization record. The distinction is the gap.
Secondary Gap: Forensic Reconstruction Is Not Possible
KPMG stated it is "reviewing the circumstances surrounding its publication." A firm with a complete forensic audit trail reviews records. It does not review circumstances. The language of investigation confirms the absence of an organizational audit trail sufficient to reconstruct what the workflow did, what it accessed, and what decisions it made, without vendor cooperation.
Tertiary Gap: Governance Architecture Existed and Was Not Applied
Condition 2.6 (Decision Auditability) is the one Procedural gap in this assessment. A sensitivity classification framework existed: ISO 42001 certification, the Trusted AI framework, internal AI-use guidelines. The procedural failure was the non-application of that framework to this specific deployment before publication. Every other gap is structural. This one was not. The framework was there. It was not activated.
Gap Score
AAB Gap Score
0
out of 100
Accountability Architecture Insufficient
8 conditions assessed. 8 Not Satisfied. 4 Critical findings. 4 High findings. 7 Structural gaps. 1 Procedural gap. No conditions satisfied or partially satisfied. Score: 0/100, the lowest classification on the AAB scale.
This assessment is a public-record application of the Vordan Agentic Accountability Baseline (AAB v0.2). All findings are derived exclusively from the public record. The deployment in scope is the AI-assisted research and citation workflow that produced KPMG's October 2025 Total Experience report. This assessment does not evaluate KPMG's broader technology estate, its client engagements, or its internal audit function. KPMG is invited to submit evidence it believes would affect any condition finding. Submissions will be reviewed and the assessment updated if warranted. This document does not constitute legal advice.
Instrument Definition
This assessment applies the Vordan Agentic Accountability Baseline (AAB), Version 0.2, to a named agentic AI deployment using the public record as its sole evidentiary basis. It is designated AAB-PR-001 to distinguish it from direct-access AAB assessments, which require organizational cooperation, practitioner access, and formal workbook completion with signature gates.
Vordan Instrument Registry
VAF Assessment
Requires direct organizational access. Uses the formal VAF workbook (v0.1). Produces an Accountability Gap Score out of 100. Requires signature gates and practitioner interviews.
VEPA
Public record only. No organizational access required. Evaluates observable accountability posture across six fixed components. Produces a Posture Score on a 1–5 scale.
AAB Assessment (Direct)
Requires direct organizational access. Uses the formal AAB Gap Score Workbook (v0.2). Evaluates all eight accountability conditions against organizational evidence. Produces a Gap Score out of 100.
AAB-PR (This Document)
Public-record application of AAB v0.2. No organizational access. Each condition evaluated against the public record. Absence of public evidence of compliance is a gap finding. Produces a Gap Score out of 100. This document is an AAB-PR assessment.
Assessment Basis
Principle I: Public Record Only
All findings are derived exclusively from the public record: published reporting, organizational statements, third-party forensic investigations, press releases, and documented product announcements. Independence from the assessed entity is the source of this assessment's credibility.
Principle II: Absence as Evidence
The AAB's evidence standards require organizations to demonstrate compliance: each condition specifies what must exist, what must be demonstrable on demand, and what must be independently verifiable. Where no public evidence of compliance exists, that absence is treated as a gap finding. This is not an inference that the organization failed; it is a finding that the organization has not demonstrated compliance on the public record. The remedy is the production of countervailing evidence, not a contest of the methodology.
Principle III: Deployment Scope
The deployment in scope is the agentic AI research and citation workflow that produced KPMG International's October 2025 report "Total Experience: Redefining Excellence in the Age of Agentic AI." The specific LLM tool or tools used in this workflow are not identified in KPMG's public disclosures. The assessment evaluates the governance architecture applied to this workflow, not KPMG's broader technology estate or any other agentic deployment.
Principle IV: Right of Response
KPMG is invited to submit evidence it believes would affect any condition finding to hello@vordan.co. Submissions are reviewed against the AAB evidence standards and the assessment is updated in a published revision if the submitted evidence satisfies the relevant tier. The right of response does not delay publication. An organization that contests a finding without producing countervailing evidence has not closed the gap.
Principle V: Evidentiary Scope
Assessment period: October 2025 (report publication) through June 22, 2026 (assessment date). Primary sources include: GPTZero's forensic investigation published June 12, 2026; KPMG spokesperson statements via TechCrunch, City A.M., Financial Times, and BusinessToday (June 12–16, 2026); KPMG's ISO 42001 certification press releases (November and December 2025); KPMG's Google Cloud partnership announcement (April 22, 2026); and KPMG's published Trusted AI framework documentation.
Gap Score Methodology
The AAB Gap Score is calculated as follows: each condition evaluated as Satisfied contributes 2 points; Partially Satisfied contributes 1 point; Not Satisfied contributes 0 points. Maximum score: 16 points, normalized to 100. A score of 0 means no condition was satisfied or partially satisfied, the accountability architecture is absent or structurally broken across all eight conditions. A score of 100 means full Active compliance across all eight conditions.
Gap findings are classified by type (Structural / Procedural / Technical) and severity (Critical / High / Medium / Low) per AAB Section 4. Critical severity applies to structural gaps in Authorization Provenance, Handoff Traceability, Forensic Reconstructibility, or Model Substrate Integrity.
Eight conditions assessed. Eight Not Satisfied. Findings are presented in AAB sequence. Evidence notes reference primary sources; full citations appear in the Sources tab.
Not Satisfied. No human authorization record for the AI research workflow exists in the public record.
Harness question: Is there a record of when and how this agent was deployed?
Must Exist
A human authorization record for each agent deployment containing the authorizing role, timestamp, scope, permitted data sources, and expiration condition. No such record exists in the public record for this deployment. KPMG's own AI-use guidelines required "human oversight to validate content and verify independent sources", but a policy is not an authorization record. A policy describes what is required. An authorization record proves it was done. No authorization record has been identified.
Must Be Demonstrable on Demand
KPMG stated it is "reviewing the circumstances surrounding its publication." An organization with a producible authorization record does not review circumstances; it reviews the record. GPTZero's conclusion, "We suspect no human at KPMG double-checked the citations, the claims, or the sources before Total Experience was published", is consistent with an authorization chain that begins with the agent, not a human decision. No authorization event has been identified in the public record.
Must Be Independently Verifiable
No authorization record stored in organizational infrastructure has been referenced in any public disclosure. Independent verifiability of an authorization record that does not appear to exist cannot be assessed. The absence is the finding.
Gap Description
The accountability chain for this deployment begins with the agent, not with a human decision. No named owner, role, timestamp, scope, or expiration condition for this workflow is identifiable in the public record. The guidelines that required human oversight existed. The authorization record demonstrating that oversight was applied did not. This is a Structural gap: the authorization architecture was not built into the deployment.
2.2 · Scope IntegrityHigh · Structural
Not Satisfied. No permission manifest or task-bounded scope control exists in the public record for this deployment.
Harness question: What permissions does this agent have?
Must Exist
A permission manifest documenting specific systems, data sources, and tools the agent may access, with expiration conditions tied to task completion. No such manifest is referenced in any public disclosure. The hallucination pattern is directly consistent with a workflow operating without temporal or topical scope boundaries: GPTZero found the tool cited a 2019 East Japan Railway press release as evidence of agentic AI, despite the term not entering wide use until 2024. A task-bounded workflow with a defined date range would not retrieve a 2019 source as evidence of a 2024-era technology.
Must Be Demonstrable on Demand
No evidence that KPMG can demonstrate permission states active at any point during the workflow's operation. The post-incident review references no permission manifest or scope boundary review.
Must Be Independently Verifiable
No permission state audit log is referenced in any public disclosure. Permission events (grant, expiry, revocation) are not evidenced.
Gap Description
The workflow operated with permissions broad enough to retrieve temporally and factually mismatched sources without constraint. The accountability condition, narrow, temporary, revocable permissions bounded to a specific task, was not satisfied by design. Structural gap.
2.3 · Memory GovernanceHigh · Structural
Not Satisfied. No memory access log, purpose boundary, or owner assignment exists in the public record for this deployment.
Harness question: What can this agent read from and write to memory?
Must Exist
A memory access log recording every read and write operation, the task context, the authorization under which it occurred, and the data owner. No such log is referenced in any public disclosure. The cross-source conflation pattern in the output, authors of one source paired with titles of another, fabricated components grafted onto real sources, is consistent with ungoverned memory accumulation across source retrieval operations, though the architecture is not confirmed in the public record.
Must Be Demonstrable on Demand
KPMG cannot demonstrate from the public record the purpose or authorization under which any memory access occurred. No write operations are traceable to human-authorized decisions.
Must Be Independently Verifiable
No persistent memory governance documentation: owner, purpose, expiration condition, is referenced in any public disclosure. Ungoverned memory is a confirmed finding where no governance record exists.
Gap Description
No memory governance architecture is evidenced for this deployment. The absence is the finding. Structural gap: procedural remediation alone cannot close a gap that requires architectural intervention.
2.4 · Handoff TraceabilityCritical · Structural
Not Satisfied. The required human review handoff was mandated by KPMG policy but did not function. No handoff record exists in the public record.
Harness question: Which agents did this agent transfer work to?
Must Exist
A handoff record for every transfer of work containing the originating agent, receiving agent, task transferred, authorization context carried forward, and timestamp. KPMG's AI-use guidelines explicitly required a human review handoff, "human oversight to validate content and verify independent sources." GPTZero concluded no human checked the citations, claims, or sources. The handoff either did not occur, or occurred without carrying the authorization context forward. Either finding satisfies this condition as Not Satisfied. No handoff record is referenced in any public disclosure.
Must Be Demonstrable on Demand
No downstream action, the publication of a 45-citation global report: it cannot be traced to the original human authorization event through an unbroken handoff chain. The chain broke at or before the human review gate. KPMG's post-incident posture ("reviewing circumstances") confirms the absence of a traceable handoff record.
Must Be Independently Verifiable
No handoff record is referenced in any public disclosure. Tamper-evidence of a record that does not appear to exist cannot be assessed.
Gap Description
A global research report was published to a professional audience with no functioning human verification gate between the agentic workflow's output and distribution. The policy required the handoff. The handoff did not function. No record of it exists. This is a Critical Structural gap: the accountability architecture required a human checkpoint, and that checkpoint was absent or non-functional at the point the output was released to the world.
2.5 · Prompt IntegrityHigh · Structural
Not Satisfied. No context log or prompt integrity control is evidenced for this deployment.
Harness question: What data entered this agent's context?
Must Exist
A context log for each agent session recording what entered the agent's context, the source of each input, and whether any input was flagged or filtered. No such log is referenced in any public disclosure. The hallucination pattern, source fragments conflated, authors confused with subjects, fabricated titles stitched from real components, is consistent with an agent whose context included unverified external data treated as authoritative input with no control layer evaluating source integrity.
Must Be Demonstrable on Demand
KPMG cannot demonstrate from the public record what entered the workflow's context, what source was used for any specific citation, or whether any input was flagged or filtered. No context audit is possible from public evidence, and KPMG's own investigation posture suggests the same limitation applies internally.
Must Be Independently Verifiable
No contemporaneous context log is referenced. Retroactive reconstruction of what entered the workflow's context is not possible from the public record. The AAB evidence standard explicitly disqualifies retroactive reconstruction as satisfying this tier.
Gap Description
The workflow ingested external sources without a control layer validating inputs against authorized instruction channels. The output, 40 fabricated or unverifiable citations, is the product of an agent whose instruction surface was ungoverned. Structural gap: no prompt integrity architecture was present in the deployment.
2.6 · Decision AuditabilityHigh · Procedural
Not Satisfied. A sensitivity classification framework existed. It was not applied to this deployment. No prospective approval record for the workflow's outputs exists in the public record.
Harness question: Is there a log of what actions the agent took?
Must Exist
A sensitivity classification governing document predating the deployment, and a prospective approval record for each sensitive action. KPMG holds ISO 42001 certification (KPMG US: November 2025; KPMG International: December 2025) and operates a published Trusted AI framework with an AI Control Framework and Responsible AI Toolkit. Both were in place when the October 2025 report was deployed. A sensitivity classification framework existed. No sensitivity classification governing document for this specific research workflow, predating its deployment, is identifiable in the public record.
Must Be Demonstrable on Demand
No prospective approval record for the AI research workflow's output exists in the public record. The decision to publish a 45-citation global report is not traceable to an identifiable human approval event for the AI-generated content. KPMG's statement that employees "may have breached internal AI-use policies" confirms the policy framework existed and was not activated for this deployment.
Must Be Independently Verifiable
ISO 42001 certification was obtained in the same period the report was deployed. Whether the certification scope covered this specific research workflow is unknown. Certification is not a governing document for a specific deployment. No deployment-specific sensitivity classification predating October 2025 is identifiable.
Gap Description
This is the one Procedural gap in this assessment. The governance architecture existed: ISO 42001 certification, a Trusted AI framework, internal AI-use guidelines. The activation mechanism failed: the framework was not applied to this deployment before publication. This distinction matters for remediation: unlike the seven Structural gaps in this assessment, Condition 2.6 is closeable without rebuilding the architecture. What is required is the consistent application of an existing framework to every agentic deployment before output is released. That did not happen here.
Not Satisfied. KPMG cannot reconstruct what this workflow did from its own audit trail. The language of its post-incident response confirms this.
Harness question: Do we have logs sufficient to investigate an incident?
Must Exist
An organizational audit trail, stored in infrastructure the organization controls, sufficient to reconstruct the complete sequence of agent actions and authorization states. No such trail is referenced in any public disclosure. KPMG stated it is "reviewing the circumstances surrounding its publication." That phrasing is the signal. A firm with a recoverable audit trail reviews the trail. A firm without one reviews circumstances, the external evidence, the accounts of people involved, the artifacts that survived. The language of the post-incident statement is a forensic finding in itself.
Must Be Demonstrable on Demand
KPMG has not demonstrated that it can conduct a full forensic reconstruction of the research workflow without vendor cooperation. The infrastructure vendor for the LLM research tool is not identified in any public disclosure, meaning forensic dependence on vendor-held logs cannot be ruled out. The AAB standard is explicit: if the reconstruction requires vendor cooperation at any point, that dependency is a gap finding.
Must Be Independently Verifiable
No tamper-evident audit trail is referenced. The complete sequence of the workflow's actions is unrecoverable from public evidence. Whether it is recoverable from KPMG's internal records is not established, and the post-incident framing suggests it may not be.
Gap Description
KPMG cannot independently establish what its agentic research workflow did, what sources it accessed, what it wrote to its context, or what decisions it made, from its own audit trail, without vendor cooperation. This is a Critical Structural gap. The organization that deployed the agent bears accountability for what it produced. That accountability requires the independent capacity to reconstruct the agent's actions. That capacity is absent.
2.8 · Model Substrate IntegrityCritical · Structural
Not Satisfied. The model identity of the deployment is unknown. No model identity record, verification method, or documented acknowledgment of the absence of verification exists in the public record.
Harness question: Is there a record of which model was deployed?
Must Exist
A model identity record documenting the authorized model, version or checkpoint identifier, source from which weights were retrieved, date of retrieval, and verification method applied. The specific LLM tool or tools used in the research workflow are not named in any public KPMG disclosure. KPMG deployed Gemini Enterprise to 55,000+ professionals globally by April 2026; whether the research workflow used Gemini Enterprise, a third-party LLM research tool, or an internal tool is unconfirmed. No authorized model, version, retrieval source, retrieval date, or verification method is identifiable from the public record.
Must Be Demonstrable on Demand
KPMG cannot demonstrate from the public record that the model executing the research workflow was the model authorized to do so, because no authorized model is identified and no verification log is referenced. The model identity question cannot be answered in the absence of a model identity record.
Must Be Independently Verifiable
No cryptographic or technically grounded verification mechanism is referenced. The AAB evidence standard also requires that where no verification mechanism exists for the deployed model, the organization must explicitly document that absence and assess the residual risk. No such documentation exists in the public record. Failing to assess and document the absence is itself a condition finding under AAB 3.8.
Gap Description
The model identity of the deployment is unknown. No identity record exists. No verification was applied. The absence was not documented and the residual risk was not assessed. Condition 2.8 is the precondition for every other condition in the Baseline, Authorization Provenance cannot be satisfied if the identity of the model taking action is unverified. This gap therefore compounds every other finding in this assessment. Critical Structural gap.
Accountability Gap Score
AAB Gap Score
0
out of 100
Accountability Architecture Insufficient
The lowest classification on the AAB scale. Fundamental accountability architecture is absent or structurally broken. No conditions satisfied. No conditions partially satisfied.
Condition Breakdown
Condition
Name
State
Type
Severity
2.1
Authorization Provenance
Not Satisfied
Structural
Critical
2.2
Scope Integrity
Not Satisfied
Structural
High
2.3
Memory Governance
Not Satisfied
Structural
High
2.4
Handoff Traceability
Not Satisfied
Structural
Critical
2.5
Prompt Integrity
Not Satisfied
Structural
High
2.6
Decision Auditability
Not Satisfied
Procedural
High
2.7
Forensic Reconstructibility
Not Satisfied
Structural
Critical
2.8
Model Substrate Integrity
Not Satisfied
Structural
Critical
Severity Summary
Severity
Count
Conditions
Critical
4
2.1, 2.4, 2.7, 2.8
High
4
2.2, 2.3, 2.5, 2.6
Medium
0
none
Low
0
none
Gap Type Summary
Type
Count
Conditions
Implication
Structural
7
2.1, 2.2, 2.3, 2.4, 2.5, 2.7, 2.8
Architecture must change. Policy updates alone cannot close these gaps.
Procedural
1
2.6
Architecture exists. Governance activation mechanism failed. Closeable without architectural change.
KPMG International achieved ISO 42001 certification, the international standard for AI Management Systems, in December 2025. Its US member firm did so in November 2025. The firm operates a published Trusted AI framework. It sells AI governance advisory to clients across finance, healthcare, and regulated industries. It has publicly positioned itself as a firm that has successfully governed AI at enterprise scale, having deployed agentic platforms to more than 55,000 professionals and built more than 150 agents for clients around the world. In April 2026, seven weeks before its retraction, KPMG announced new AI agents built on the premise that its Trusted AI framework ensures systems act within clear ethical and operational boundaries.
The AI-assisted research and citation workflow that produced the Total Experience report satisfies none of the eight conditions of the Vordan Agentic Accountability Baseline. Not one. The Gap Score is 0 out of 100. Four of the eight gaps are Critical. Seven of the eight are Structural, meaning they cannot be closed by the policy documents and framework certifications KPMG has already obtained. The architecture was never built into this deployment. The framework that KPMG certified, published, and sells did not govern the internal agentic workflow that cost it a global retraction and the credibility of four named organizations that disputed its claims.
This is not a finding about AI hallucination. Hallucination is a known and documented property of how large language models generate text. Every organization deploying an AI research workflow in 2025 knew this, or should have. What the AAB measures is not whether the model made mistakes. It measures whether the organization built the governance architecture to prevent those mistakes from reaching the world, the human authorization gate, the handoff record, the forensic trail, the scope boundary, the model identity record. KPMG built none of these for this deployment. A firm that cannot satisfy Condition 2.4, Handoff Traceability, allowed an agentic workflow's unverified output to be published to a global professional audience, cited in news articles, and fed into the research pipelines of other institutions, without a documented human review gate. The harm did not stop at the retraction.
The accountability gap between what KPMG claims its governance architecture delivers and what that architecture produced in practice is the subject of this assessment. Vordan publishes this finding not to characterize KPMG's intent, but to name the structural condition that makes this outcome possible: a firm can certify an AI management system, publish a Trusted AI framework, and fail every accountability condition on a production agentic deployment simultaneously. The certification and the framework are necessary. They are not sufficient. The AAB defines what sufficient looks like. This assessment documents the distance between the two.
Disclosure
Vordan has no commercial relationship with KPMG International or any of its member firms. No financial interest, consulting relationship, or prior engagement exists. This assessment is conducted under Vordan's Editorial Independence Declaration (effective May 2026). KPMG is invited to submit countervailing evidence to hello@vordan.co. The assessment will be updated in a published revision if submitted evidence satisfies the relevant AAB evidence standard.
Primary forensic source. 45-citation analysis. GPTZero conclusion on absence of human review. "Vibe citing" taxonomy. Used in findings for conditions 2.1, 2.2, 2.4, 2.5.
Source 02
KPMG Pulls Its Agentic AI Report After Hallucinations Surface
Organization disputes (UBS, NHS, Swiss Federal Railways, Transport for London). "Reviewing circumstances" statement. EY Canada precedent. Used in findings for conditions 2.1, 2.4, 2.7.
Source 03
KPMG report on AI found riddled with AI hallucinations
KPMG spokesperson: "We expect all our people to follow our guidelines on the responsible use of AI, including human oversight to validate content and verify independent sources." Employees may have breached policies. Used in findings for conditions 2.1, 2.4, 2.6.
Source 05
KPMG's AI report becomes an accidental demo of AI hallucinations
ISO 42001 certification obtained December 2025, after October 2025 report deployment, before June 2026 retraction. Used in findings for conditions 2.6, and Vordan Position.
Source 07
KPMG LLP Among First of the Big Four in the U.S. to Receive ISO 42001 AI Certification
KPMG announcement of new AI agents and Trusted AI framework positioning, April 22, 2026, seven weeks before retraction. 55,000+ professionals on Gemini Enterprise. Used in findings for condition 2.8 and Vordan Position.
Source 09
KPMG named 2026 Google Cloud Partner of the Year for Global Industry Solutions: Breakthrough