Skip to content
Brocode SolutionsAI Software Development

Brocode Research · Volume 4 · 2026 · 6 whitepapers · 312 pages · 14 named authors · 9 external reviewers

Substantive, GCC-grounded research — for boards that read each other's briefings.

Long-form whitepapers on Arabic LLMs, sovereign cloud, model risk in CBUAE-supervised banks, and the primary-research GCC Enterprise AI Benchmark — co-authored with regional operators and academics. Executive summary, methodology appendix, and one full chapter are open. The rest is one email away.

Download the GCC Enterprise AI BenchmarkBrowse the library

Contact policy: we email you the paper, send at most four research updates a year, and never call you uninvited. One-click unsubscribe.

Bound research documents arranged on a desk beside a tablet displaying analytical charts

Flagship · current

The GCC Enterprise AI Benchmark 2026 — 72 pp

218 respondents · 6 countries · 6 sector breakdowns

  • 312 pp

    Across six current whitepapers

  • 14

    Named Brocode authors

  • 9

    External co-authors and reviewers

  • 4 / yr

    Quarterly research-newsletter cadence

Current flagship · February 2026

The GCC Enterprise AI Benchmark 2026.

A primary-research survey of 218 CIO/CDO/CAIO respondents across the UAE, KSA, Qatar, Kuwait, Bahrain, and Oman — sector breakdowns for banking, energy, telco, government, healthcare, and retail.

Headline finding

Only 14 % of GCC enterprises with an approved AI roadmap have a production model serving customers today.

The gap is not strategy, and it is not budget. The gap is data foundation and governance — both of which are addressable inside a single budget cycle when sequenced honestly. The benchmark unpacks the finding sector by sector, with full methodology, raw anonymised data, and a public reproducer.

Table of contents (open)

  • 1. Methodology and sample (open)
  • 2. AI roadmap maturity (open)
  • 3. Production AI rate by sector (gated)
  • 4. Sovereign deployment posture (gated)
  • 5. Buying patterns and procurement (gated)
  • 6. Sector deep-dives × 6 (gated)
  • Appendix A: raw anonymised data (gated)

Free download

GCC Enterprise AI Benchmark 2026

72 pages, 218 respondents, six countries, six sector breakdowns. Full methodology and raw anonymised data in the appendix.

  • 218-respondent primary survey across six GCC countries
  • Six sector deep-dives — banking, energy, telco, government, healthcare, retail
  • The 14 % production-AI finding with sector-level data
  • Raw anonymised data and public reproducer code
  • Methodology appendix with sampling frame and cut-off date
PDF

GCC Enterprise AI Benchmark 2026

Instant download. No spam. Unsubscribe any time.

The library

Five further titles, all written by named authors.

Each card shows page count, author, external reviewers, and a full abstract. Click through to download — same form, same contact policy.

  • EngineeringAvailable — gated PDFJanuary 2026

    The State of Generative AI in GCC Banking 2026

    64 pages · 18 data exhibits · methodology appendix

    How four GCC tier-1 banks moved their first GenAI use case into production. Architecture choices, regulator-facing evidence, total cost of ownership over 24 months, and what we would not do again. The paper includes the redacted board paper from one of the four banks, published with explicit permission.

    Authors
    Layla Mansoor (Principal ML Engineer), Head of Data, tier-1 GCC bank
    External reviewers
    MBZUAI faculty member and a former CBUAE supervisor
    Download the whitepaper
  • EngineeringAvailable — gated PDFNovember 2025

    Arabic LLMs: an Evaluation Framework and Benchmark Results on 9 Open and Closed Models

    48 pages · 22 data exhibits · methodology appendix

    An open evaluation harness for Arabic LLMs covering MSA and Khaleeji dialect, with benchmark results on Jais, Falcon, Llama-3, GPT-4o, Claude, Gemini, and three open-weight regional variants. The harness is published on GitHub; the anonymised test set is published on Hugging Face under a permissive licence.

    Authors
    Yasmin Al Marzooqi (Head of Arabic NLP)
    External reviewers
    KAUST researcher and a senior Arabic linguist
    Download the whitepaper
  • EngineeringAvailable — gated PDFOctober 2025

    AI Sovereignty in the UAE: Architectures, Trade-offs, and Cost Models

    56 pages · 14 data exhibits · methodology appendix

    Reference architectures for AWS UAE North, Azure UAE North, OCI Abu Dhabi, G42 Cloud, and on-prem appliances — with 36-month TCO models, regulator-mapping for PDPL, and honest discussion of where sovereign approaches still trail hosted alternatives.

    Authors
    Khaled Al Otaibi (Principal Architect), former Federal CIO (co-author)
    External reviewers
    Senior partner at a UAE federal entity and a sovereign-cloud architect
    Download the whitepaper
  • GovernanceAvailable — gated PDFSeptember 2025

    Model Risk Management for AI in SAMA- and CBUAE-Regulated Banks

    52 pages · 16 data exhibits · methodology appendix

    Three-lines-of-defence operating model, AML/fraud/credit model evidence packs, bias and fairness test packs, model cards, periodic validation playbook, and the specimen notification triggers for CBUAE and SAMA. A specimen regulator-grade evidence pack is included as an appendix.

    Authors
    Aisha Al Hosani (Head of AI Risk), ex-Central Bank examiner (co-author)
    External reviewers
    A model-risk committee chair from a GCC bank and a SAMA-experienced supervisor
    Download the whitepaper
  • EconomicsAvailable — gated PDFAugust 2025

    The Real Cost of GenAI at Enterprise Scale: a 12-Month FinOps View from Three GCC Deployments

    38 pages · 12 data exhibits · methodology appendix

    Twelve months of bill-of-materials data from three GCC GenAI deployments — total dirham cost, cost per million tokens, GPU utilisation curves, and the line items that swallowed the budget. Hyperscaler and on-prem stacks compared at the same workload volumes.

    Authors
    Tareq Ibrahim (Principal Platform Engineer)
    External reviewers
    A FinOps lead at a UAE-listed company and a senior platform engineer at a regional bank
    Download the whitepaper

Research standards

What earns the gate.

The standards every Brocode whitepaper must clear before it goes to print. The page is the contract.

  • Methodology appendix on every paper

    Sample size, sampling frame, data sources, model versions, and the cut-off date — disclosed in plain text in the appendix.

  • Named authors and external reviewers

    Every paper carries the byline of a Brocode practitioner and at least one external co-author or reviewer drawn from a regional bank, regulator, university, or operator.

  • Public erratum log

    When we get something wrong, we publish the correction at /insights/whitepapers/errata with the date and the prior text. Errata are not deleted.

  • Open data and code where licensing permits

    Evaluation harnesses, anonymised survey data, and benchmark code are published on GitHub and Hugging Face under permissive licences.

  • No Brocode product pitch in any whitepaper

    The back page carries the author bios and a single neutral line ("Brocode advises GCC enterprises on AI; this paper does not promote our services") and nothing else.

  • Version history and reissue cadence

    Every paper carries a version number. Material updates are reissued with a fresh date; the prior version remains downloadable.

Erratum log: the live erratum register is published at /insights/whitepapers/errata. Past corrections are not deleted; if a finding was revised, the date and the prior text remain visible.

Open data and code

Reproducers on GitHub. Datasets on Hugging Face.

Where licensing permits, the data and the code behind each whitepaper are published openly under permissive licences.

  • Khaleeji Benchmark v2. The open evaluation suite for Arabic LLMs that backs the Arabic-LLM whitepaper. MIT licence. Pull requests welcome.
  • GCC Enterprise AI Benchmark — anonymised data. The 218-respondent dataset, with PII removed and respondent identifiers replaced by stable hashes. CC-BY-NC-4.0.
  • Sovereign-cloud TCO models. Open spreadsheets (Google Sheet and Excel) modelling 36-month total cost of ownership for AWS UAE North, Azure UAE North, OCI Abu Dhabi, G42 Cloud, and on-prem appliances.
  • Model-risk evidence pack templates. Specimen model cards, datasheets, fairness test packs, and explainability suites used in the model-risk whitepaper.

Cited by

Where Brocode Research has been read.

A partial list of publications and institutions that have cited Brocode Research, with explicit permission. Citing peers is the proof we trust most.

  • A CBUAE technical note (2025)
  • An MBZUAI working paper (2025)
  • Three tier-1 GCC bank board materials (with permission)
  • A Mubadala portfolio investment memo (with permission)
  • A regional procurement framework circular
  • A Khaleej Times technology feature (2025)

Authors and reviewers

The bylines on Brocode Research.

Each paper carries a named Brocode practitioner and at least one external co-author or reviewer.

Brocode authors

Yasmin Al Marzooqi — Head of Arabic NLP. Aisha Al Hosani — Head of AI Risk. Khaled Al Otaibi — Principal Architect, sovereign cloud. Omar Haddad — Principal Architect, benchmark methodology. Layla Mansoor — Principal ML Engineer, GenAI in banking. Tareq Ibrahim — Principal Platform Engineer, FinOps.

External co-authors and reviewers

Two MBZUAI faculty members. A KAUST researcher. A former Federal CIO. An ex-CBUAE supervisor. The Head of Data at a tier-1 GCC bank. A SAMA-experienced model-risk committee chair. A FinOps lead at a UAE-listed company. A sovereign-cloud architect at an operator. Full bios at /about/research-team — written consent renewed annually.

Subscribe

The Brocode Research quarterly.

One email, four times a year. We send the new flagship paper, two short papers, and a one-line note on what is in the pipeline. No second send, no drip, no sales hand-off without a clear inbound trigger from you.

Contact a research author

Journalists, academics, and CIOs welcome — directly to the author.

We keep WhatsApp lighter on this page: research readers prefer email. The link below routes a short message to the author of the paper you mention.

Message a research author on WhatsApp

Adjacent reading

Related capabilities and stories