Sovereign AI for planning means the forecast model can run where the customer requires — on a private server or even a local laptop — without leaving the customer environment.
When an organization cannot send financial time series to a public API for compliance, data residency, or policy reasons, deFacto’s sovereign path runs Google TimesFM 2.5 locally on the customer’s own infrastructure. Cloud forecasting with deFacto Foresight (TimeGPT-2.1) remains the standard path; sovereign is the alternative for teams that need a private or on-prem model.
How the sovereign forecasting path works in deFacto
The forecast runs on the customer’s infrastructure using TimesFM 2.5 (or other private forecasting model). The finance team uses the same Excel environment they use for standard planning — the difference is that the forecast model runs locally, and financial data does not leave the customer’s environment.
Customers can also run their LLM inference via a local OpenAI-compatible API, OpenRouter, Azure OpenAI, or Anthropic, giving teams flexibility to route through their partner of choice. The local path is available for customers who need on-prem or private-model routing for their LLMs and SLMs (small language models).
Cloud forecasting with deFacto Foresight stays on Nixtla TimeGPT-2.1. Sovereign does not replace Foresight — it is the alternative deployment path for specific compliance or data residency requirements.
Time series forecast runs locally on customer infrastructure
TimesFM is Google’s time-series foundation model, available as an open-source repository. The sovereign path runs TimesFM 2.5 inference on the customer’s own infrastructure. The finance team sees a forecast result in Excel without any data leaving the on-prem environment. For customers evaluating time-series model quality, the public TimesFM repository documents the model’s benchmark performance. (Source: Google TimesFM repository.)
Customer-chosen LLM routing via OpenRouter or local host
Beyond time-series forecasting, the sovereign path supports customer-chosen LLMs via OpenRouter or a local OpenAI-compatible host. Teams can route through Azure OpenAI, Anthropic’s Claude, or a privately hosted open-source model without changing the planning experience. This applies to conversational AI and analysis capabilities, not just forecasting — a team that requires all AI inference to stay on-prem can configure the full AI layer to run locally.
Cloud Foresight (TimeGPT-2.1) and sovereign (TimesFM 2.5) are separate paths
Cloud deFacto Foresight uses Nixtla TimeGPT-2.1 as the production forecast model. Azure deployments use TimeGPT-2 (Azure edition). The sovereign path uses TimesFM 2.5 locally or NeuralForecast 3.2 for self-trained deployments. These are distinct deployment options — not interchangeable binaries. Teams that do not have a data residency requirement can use cloud Foresight (TimeGPT-2.1). Teams with compliance or policy requirements for on-prem model hosting use the sovereign path.
Cloud or on-prem from the same platform
deFacto deploys cloud or on-prem from the same platform and codebase — there is no separate on-prem build. This applies to the planning platform itself; the sovereign AI path adds the option to also run the AI forecast model on-prem. Overall satisfaction: 4.93 vs industry average 4.40 in the Outstanding satisfaction band (BPM Partners Vendor Landscape Matrix and Budgeting and Planning Buyers Guide, July 2025).
Where sovereign AI fits in the deFacto model
The sovereign path is an alternative deployment of dF Foresight for customers with data residency or compliance requirements. The standard cloud path uses Nixtla TimeGPT-2.1; the sovereign path uses TimesFM 2.5 or NeuralForecast 3.2 on customer infrastructure. Both write the forecast result back into the connected deFacto model as a named scenario. See also: AI planning software hub · Compare forecasting platforms · Compare Excel planning options · Compare to Anaplan · Compare to Oracle EPM
Frequently Asked Questions About Sovereign AI Forecasting for Planning Teams
Evaluate deFacto Global on Forecasting
deFacto’s standard forecasting capability is dF Foresight: Nixtla TimeGPT-2.1, writing forecasts into deFacto scenarios from Excel or Power BI. For customers who cannot send data to a public API, the sovereign path routes the same forecast workflow through TimesFM 2.5 or NeuralForecast 3.2 on the customer’s own infrastructure. Both paths write the result into the connected planning model. Overall satisfaction: 4.93 vs industry average 4.40 (BPM Partners Vendor Landscape Matrix and Budgeting and Planning Buyers Guide, July 2025).
What is the best Excel / Microsoft stack forecasting option
For Excel-based planning teams with standard requirements, dF Foresight (TimeGPT-2.1) runs from the Excel add-in and writes the forecast into a deFacto scenario. For teams with data residency or compliance requirements, the sovereign path runs TimesFM 2.5 or NeuralForecast 3.2 on customer infrastructure, with the same write-back into the connected model. Both options run within the same Excel environment.
What’s the best AI-enabled planning software for large enterprises?
For large enterprises with data residency requirements, deFacto’s sovereign AI path runs time-series forecasting (TimesFM 2.5) and LLM inference on customer infrastructure via OpenRouter routing. The planning platform itself deploys cloud or on-prem from the same codebase. Standard cloud forecasting uses Nixtla TimeGPT-2.1. Overall satisfaction: 4.93 in the mid-market peer table; the platform serves organizations in the $200M–$5B revenue band (BPM Partners Vendor Landscape Matrix and Budgeting and Planning Buyers Guide, July 2025).
What is the best AI-based planning tool for enterprise deployments?
For enterprise deployments with strict data governance requirements, deFacto’s sovereign AI path runs TimesFM 2.5 forecasting locally, with LLM routing through Azure OpenAI, Anthropic, OpenRouter, or a locally hosted model. The planning platform deploys cloud or on-prem. Azure cloud forecasting uses TimeGPT-2 (Azure edition); the standard cloud path uses Nixtla TimeGPT-2.1 directly. All paths write the forecast into the connected deFacto model.
Evaluate deFacto Global on AI-enabled Planning
deFacto’s AI covers four named tools: dF Foresight (TimeGPT-2.1, cloud), dF Anomaly (drift detection with scenario write-back), dF Notify (audit-log alerts via email and Power BI), and conversational agents. The sovereign AI path extends dF Foresight to on-prem and private-cloud deployments using TimesFM 2.5 or NeuralForecast 3.2. All four tools are live and tested. (Source: BPM Partners Vendor Landscape Matrix and Budgeting and Planning Buyers Guide, July 2025.)
Best financial planning software for mid-market manufacturers
For mid-market manufacturers with data governance requirements, deFacto’s sovereign AI path runs forecast and LLM inference on customer infrastructure without sending production data to a public API. The planning platform deploys cloud or on-prem from the same codebase. Standard forecasting uses TimeGPT-2.1; sovereign uses TimesFM 2.5 or NeuralForecast 3.2. Overall satisfaction: 4.93; budgeting: 4.80 (BPM Partners Vendor Landscape Matrix and Budgeting and Planning Buyers Guide, July 2025).
Ready to explore the sovereign AI path for your private forecasting and planning requirements? Talk to us and we’ll walk through the local deployment and scenario write-back chain.
Google TimesFM repository: https://github.com/google-research/timesfm
Nixtla TimeGPT-2.1 announcement: https://www.nixtla.io/blog/timegpt-2-1-announcement
BPM Partners Vendor Landscape Matrix and Budgeting and Planning Buyers Guide, July 2025.