Emirates NBD has won the Best AI-Powered Customer Experience and Engagement Initiative in the Middle East award for Document Intelligence, its enterprise platform for processing customer documents. The category forms part of the AI Excellence Awards under the TAB Global Business Achievement Awards 2026.

Document Intelligence combines optical character recognition, machine learning and generative AI to convert unstructured customer documents into validated, reusable data. Emirates NBD says the platform supports onboarding, servicing and compliance processes, reducing repeated data collection and manual document handling.

Emirates NBD reports that standard due diligence turnaround time has fallen from approximately 19 days to 4.7 days, while enhanced due diligence has declined from 41.5 days to 11.4 days. The platform pre-populates around 100 of 250 onboarding fields and has achieved approximately 98% precision in document classification and attribute extraction.

Turning documents into reusable data

Business-banking onboarding and compliance reviews depend on information held across identity, legal, financial and business documents. Emirates NBD is using Document Intelligence to convert that information into validated data that can be used by its onboarding and compliance systems.

Customers submit documents through the bank’s digital channels. The platform identifies each document type, extracts relevant information from structured and unstructured content, checks whether the information is complete and passes validated data to other systems.

The documents processed include trade licences, memoranda of association, board resolutions, passports, Emirates IDs, tenancy contracts and other legal, financial, business and identity records required for onboarding and compliance reviews.

Extracted information is enriched with metadata and assigned confidence scores. It is checked against application requirements and internal records before being made available through application programming interfaces to downstream systems.

During onboarding, Document Intelligence pre-populates approximately 100 of 250 application fields. This has reduced form-filling effort by about 40%, according to the bank. Customers can review and amend the extracted information before proceeding.

Emirates NBD describes the change as a move from a “submit, wait and rework” process to an “upload once, review and proceed” experience. Improving the interface will not remove much of the friction if customers must still re-enter information contained in their documents or respond to repeated requests for missing information. The platform addresses the document processing behind the interface.

Combining generative AI with established technologies

Document Intelligence uses generative AI alongside optical character recognition and machine learning. Generative AI is applied to document classification, context-aware attribute extraction, and gap analysis and validation.

Document formats and layouts can vary considerably even when they serve the same purpose. The classification function identifies what has been submitted so that the appropriate information can be extracted and assessed.

Large language models then identify attributes such as document numbers, expiry dates, shareholder information, beneficial-owner details, addresses and other know-your-customer data. This enables the platform to process standard fields as well as information contained in less consistently structured documents.

The gap-analysis and validation function checks extracted information against business, onboarding and compliance requirements. It identifies missing, inconsistent or expired information that could prevent an application or review from progressing.

Generative AI is only one part of the process. OCR captures the content, AI interprets and structures it, and validation services determine whether the information can be reused. The customer benefit depends on these components working together and connecting to the bank’s operational systems.

Measuring speed and processing quality

Emirates NBD measures the platform’s performance through operational dashboards and process tracking across its onboarding and compliance workflows. For digital account opening, it compared processing times before and after implementation.

Standard due diligence turnaround time fell from approximately 19 days to 4.7 days, a reduction of about 75%. Enhanced due diligence declined from around 41.5 days to 11.4 days, a reduction of approximately 72%.

Document Intelligence is also used in high-risk annual reviews. Turnaround time for these reviews decreased from approximately 66 days to 21 days, a reduction of about 68%. Clean-case submission rates improved by 22%, reducing rework and further customer follow-up.

The bank monitors extraction quality through model validation and precision assessments. It reports approximately 98% precision in classifying documents and extracting attributes. Confidence scoring and validation services assess the quality and completeness of the resulting data before it is passed to downstream systems.

Precision measures whether the platform has correctly classified and extracted information. It does not establish that a submission is complete or that the information should be accepted without review. Emirates NBD retains validation controls and human intervention for incomplete, inconsistent and higher-risk cases.

The measures cover both speed and processing quality. Turnaround times show whether applications and reviews progress faster, while clean-case submissions indicate whether fewer cases require rework. Precision and confidence scores assess the reliability of the extracted information.

Moving from pilot to operational deployment

Emirates NBD says Document Intelligence is used across onboarding, servicing and compliance processes, providing a common layer for customer document processing and data reuse. Its reported turnaround-time improvements cover UAE business-banking onboarding and high-risk annual reviews.

The platform connects through API-based services to the bank’s digital sourcing platform, onboarding workflow systems, compliance processes and core banking platforms. Validated information can therefore move between systems without each process having to interpret the original documents separately.

Emirates NBD said 100% of its onboarding and compliance staff use the capability as part of their day-to-day work. Extracted information must reach the systems and employees responsible for progressing an application or review if improvements in model performance are to translate into shorter customer turnaround times.

Designed as an enterprise capability, the platform can be extended to additional business units, product journeys, regulatory processes and servicing workflows without rebuilding its document-processing capabilities for each use case.

Beyond individual document-processing tasks, the platform creates a reusable customer data asset that can support onboarding, servicing and compliance, as well as future customer-lifecycle applications. Reusing validated information across these processes reduces the need to collect the same data repeatedly.

Retaining human accountability

The platform operates through both human-in-the-loop and human-on-the-loop controls. Customers and authorised employees can review, amend and validate extracted information before it is committed to downstream processes or decisions. Employees monitor outputs and intervene in exceptions, corrections and higher-risk cases.

Document Intelligence is governed through Emirates NBD’s enterprise AI governance framework. An AI Guild is responsible for standards, controls and the reuse of AI capabilities. The Model Risk Management function oversees model validation, risk assessment, monitoring and lifecycle controls.

An independent AI Ethics Council reviews high-impact customer-facing applications and considers fairness and societal impact. Cybersecurity and technology-control stakeholders also participate in the governance process.

Before deployment, models undergo validation, risk assessment, bias testing, explainability checks, security reviews and privacy controls. Monitoring after deployment covers model performance, drift detection, data lineage, audit trails, model versioning, automated compliance checks and continuing governance oversight.

Full data lineage links source documents to extracted attributes and subsequent actions. This allows the bank to trace information back to the original document, review how it was processed and correct it where necessary. Validation controls also cross-check extracted information against application requirements, internal records and authoritative sources before it is reused.

These controls become more important when extracted information is distributed across multiple systems. An inaccurate attribute could otherwise be reused at several points in the customer lifecycle. Confidence scores, traceability and human review provide checks before the information moves further into the bank’s processes.

Testing whether the platform can scale

Emirates NBD has built Document Intelligence as a common service that can classify, extract, validate and distribute information across connected systems. A shared platform can reduce the need to build separate document-processing tools for each onboarding or compliance journey.

Its current results show where AI can affect customer experience without appearing as a standalone customer-facing feature. Customers encounter the outcome through shorter forms, fewer requests for information and faster processing rather than through direct interaction with a generative AI application.

The reductions in onboarding and review times are substantial, but wider deployment will bring additional demands. More document types, systems and customer journeys will increase the number of points at which information must be validated, monitored and traced.

Replicating the current results beyond business-banking onboarding and high-risk reviews will depend on Emirates NBD maintaining extraction quality, effective human oversight and reliable system integration as volumes and use cases expand.