Skip to main content
Banking and Markets

Northern Trust's onboarding agent cuts data preparation time by 83%

Northern Trust and Microsoft announced on November 9, 2026 a client onboarding agent for the Front Office Solutions platform. Northern Trust says it cut data preparation time by 83%, from about 2.5 weeks to 3 days, and reached a 96.5% first-pass F1 score in controlled tests.

  • Nov 18, 2026
  • 3 min read
  • LabForty AI Newsroom
Northern Trust's onboarding agent cuts data preparation time by 83%

Northern Trust and Microsoft announced on November 9, 2026 an expanded collaboration that begins with a new AI-powered client onboarding agent for clients on Northern Trust Asset Servicing's Front Office Solutions platform. The agent was built with Microsoft Frontier Company forward deployed engineers.

Front Office Solutions provides performance and analytics for institutional investors, including foundations, endowments, pension schemes and large family offices. Onboarding can require Northern Trust to consolidate investment and transaction data from several sources and reconstruct performance histories, including for alternative assets. Northern Trust says more than half of asset owners it surveyed, 57%, name consolidating data from multiple internal and external sources as their biggest challenge.

The agent converts a client's historical transaction data into Northern Trust's standardised format. After deployment, Northern Trust measured preparation times for real client onboarding cases, including data transformation, analyst review, exception handling and final validation. It says the time fell by 83%, from a prior baseline of about 2.5 weeks to an average of 3 days on the measured production cases.

A 96.5% F1 first-pass quality score comes from controlled testing across 7 representative cases that used real, anonymised historical client transaction workbooks. Agent output was compared with analyst-prepared reference outputs that were withheld during the run, before analyst feedback or corrections. F1 combines precision and recall.

Human review stays in the workflow. Analysts review outputs, evaluate reasoning, resolve exceptions and improve the system through feedback. The agent runs during implementation, before the client goes live, so errors can be reconciled against the starting data. It is built on Microsoft Azure and Microsoft Foundry and connects to Excel and SharePoint so analysts can review, refine and approve the work.

Microsoft says it improved the F1 score and other measures using only anonymised data, in what the companies call an eyes-off approach, so Northern Trust client data stayed under the firm's control. As of September 30, 2026, Northern Trust reported 19.9 trillion US dollars in assets under custody or administration and 2.0 trillion US dollars in assets under management.

Why this matters

The useful pattern is not the headline time cut alone. The agent drafts a standardised file, an analyst approves it, and the quality number is a first-pass score on 7 anonymised cases, taken before those corrections. Banks copying the design should keep the human checkpoint and treat 83% and 96.5% as Northern Trust's measured results, not a benchmark for every book of alternatives.

Sources: Microsoft announcement.

This article is for informational purposes only and is not investment advice.

This article is not legal advice.

Where every detail matters

Where every detail matters

At LabForty, we develop high-quality websites with a strong focus on detail - from architecture and user experience to business logic.