BBVA's Internal Audit teams needed to analyze large volumes of data from multiple sources and heterogeneous formats, design and run increasingly complex analytical tests, verify data quality, document assumptions and decisions, and preserve full traceability without relying on inconsistent spreadsheet work or ad hoc code created by individual auditors.
Reported impact
BBVA expects the assistant to make analytics use in audit engagements more consistent, controlled, and scalable, reduce manual and repetitive programming and documentation work, free auditors for higher-value risk assessment and anomaly analysis, and improve audit quality while maintaining traceability and human accountability.
BBVA created a purpose-specific assistant within ChatGPT Enterprise for its Internal Audit function to standardize and strengthen large-scale data analysis. Used inside BBVA's controlled corporate environment, the assistant guides auditors through a structured workflow: validating the data initially included in an audit, running quality and consistency checks, helping design specific and complex analytical tests, interpreting the results, and producing reproducible Python or SQL code together with structured analytical explanations for documentation. The implementation is intentionally human-supervised rather than autonomous: auditors review each step, apply professional judgment, and remain responsible for validation and final conclusions. BBVA also applied its governance framework for AI assistants, adding controls for suitable data use, traceability, reproducibility, ongoing monitoring, and periodic review.
Internal Audit brings data from multiple internal sources and heterogeneous formats into the assistant-led workflow. Within ChatGPT Enterprise, the assistant helps verify the input data, performs quality and consistency checks, supports the design of analytical tests, interprets outputs, and generates reproducible Python or SQL plus structured explanations. Auditors then review those artifacts, document findings, and make the final conclusions.
"Analyzing large volumes of data from multiple sources and in heterogeneous formats."
"At each stage, the system documents the constraints and record counts used, giving auditors full visibility into the analysis performed by the assistant"
"BBVA has created an assistant with ChatGPT Enterprise to support data analysis at Internal Audit, from validating and designing tests to interpreting results."
"It also generates reproducible code in programming languages commonly used in the audit function, such as Python or SQL, as well as structured analytical explanations ready for auditors to use."
"The assistant guides auditors step by step through verifying the data initially included for each audit, performing quality and consistency checks, designing specific and complex analytical tests, and interpreting the results."
"it is expected to help increase productivity by around 10% in each data-driven audit."
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
OpenAI Sample Oct 16, 2026 Benefit Street Partners Benefit Street Partners needed to answer investment-committee follow-up questions faster, compare lender marks across shared loans, and review 300-page credit agreements for clauses that could shift value away from… 10–12 hours total. BDC mark tracker · Codex · Credit agreement analyzer OpenAI Sample Oct 16, 2026 Itaú Itaú needed to make financial guidance simpler and more personalized for millions of customers while also scaling document-heavy internal workflows, all under the governance, observability, and economic controls… 100% of the customer base is the goal by the end of 2026 Customer DNA · IARA · OpenAI models OpenAI Sample Oct 14, 2026 Block Block needed to shrink the time required to tune detections, investigate alerts, and move from a high-level vulnerability description to a validated patch, while reducing the analyst effort consumed by false positives… 18 minutes from submitting a high-level vulnerability to an automated proposed patch Block security harness · Codex · OpenAI Daybreak 