Skip to main content
Assistant in development Join the launch list

NATIONAL ACCOUNTS · ASSISTANT IN DEVELOPMENT

From the 2008 SNA
to the 2025 SNA

Which recommendations changed, and what do they mean for the accounts you compile? sna2025.ai is a multilingual assistant in development that will help national accountants compare the two editions and trace each answer to the source.

An initiative for professionals who compile the accounts.

THE REASON FOR THE PROJECT

The questions start
when compilation begins

The System of National Accounts (SNA) is the framework behind national accounts. The 2008 SNA updated the 1993 edition and was adopted in stages in 2008 and 2009. Many established compilation methods are based on it. The UN Statistical Commission adopted the 2025 SNA in 2025. The joint implementation strategy encourages a 2029–2030 target window, with timing depending on national circumstances.

Updating established compilation systems takes time and care. Even after new methods are in place, workshops, review and ongoing monitoring remain essential: questions keep arising in a field as demanding as national accounting. We're building sna2025.ai to make specialist support easier to reach at each step, helping professionals work through those questions and verify the reasoning against the standard.

FROM QUESTION TO EXPLANATION

How the assistant will support compilation

  1. Clarify what you need

    Ask in ordinary language. Clear questions would proceed directly; otherwise, the agent would suggest interpretations and help you refine the question, confirming what you need before answering.

  2. Build a verifiable answer

    The agent would find relevant passages in the 2008 SNA and 2025 SNA and explain the answer, with edition, chapter, section and paragraph references, plus page numbers where reliable.

  3. Explore, refine and reuse

    Follow up for more detail, a simpler explanation or a graph. Planned exports include text, Word and PDF, with source references carried through.

THE TECHNOLOGY BEHIND THE TOOL

Built around sources.
Shaped by expertise.

The 2008 SNA and 2025 SNA form the primary source corpus for the planned knowledge base and retrieval-augmented generation (RAG) pipeline.

01 / Model selection

We plan to start with leading proprietary LLMs such as OpenAI’s GPT-6 Astra, prioritising answer quality, and experiment with open-weight alternatives such as gpt-oss-120b.

02 / Domain training & fine-tuning

The 2008 SNA and 2025 SNA will be our main reference material. We’ll evaluate fine-tuning on expert-reviewed examples drawn from these sources to adapt suitable models to national accounts terminology and questions.

03 / RAG & traceability

For each question, RAG would retrieve relevant passages and produce an answer with edition, chapter and paragraph references, plus page numbers where reliable. Traceable sources would support verification, auditability and responsible AI use.

Planned answer formats include text, tables, knowledge graphs, diagrams and images, adapted to what the user needs.

Help give the answers more context

SNA authors and practitioners are welcome to contribute explanatory notes, workshop material and practical examples. Shared with permission, these could strengthen the knowledge base, evaluation and fine-tuning.

Contribute material or expertise

A PERSONAL INITIATIVE

Who is building
sna2025.ai?

I’m Iliass Taleb, a Senior National Accounts Statistician at the High Commission for Planning, Morocco’s national statistical office.

Working with the 2008 SNA has taught me how difficult it can be to find the right guidance, or even frame a question. With the 2025 SNA, there is more to work through. I started sna2025.ai to make that work easier, with explanations you can trace to the source.

This project is a personal initiative, currently financed by me.

STAY IN THE LOOP

Be notified when it’s ready.

Join the launch list to hear when sna2025.ai is ready to try.

Join the launch list