A unified information and AI basis for monetary companies
Addepar is a worldwide know-how and information platform that empowers funding professionals to show advanced monetary info into actionable intelligence. Registered funding advisors, household places of work, non-public banks and international establishments depend on Addepar to unify portfolio, market and shopper information and ship a complete portfolio view throughout private and non-private markets.
Knowledge and AI are elementary to this mission. Addepar now manages almost $9 trillion in belongings on its platform, and purchasers depend on safety, high quality and consistency to make knowledgeable, high-stakes choices. To assist this, Addepar moved from a group of older techniques and database instruments to a single information intelligence platform on Databricks operating on AWS. That platform ingests a whole bunch of disparate information feeds, standardizes and enriches them after which delivers the outcomes to purchasers via merchandise, APIs and information sharing.
Constructing on Databricks for scale, governance, and collaboration
Addepar selected Databricks to unify engineering, analytics and AI on a single, ruled information platform. Collaborative notebooks and SQL let inner groups work in a single place, whereas Unity Catalog offers the fine-grained permissions and entry controls {that a} international monetary companies footprint calls for.
The result’s a single supply of fact that engineers, analysts, and now AI techniques can all rely upon.
This resolution has produced a transparent enterprise influence. Since adopting Databricks, Addepar has lowered pipeline prices by 60% versus legacy infrastructure—driving greater than $2 million in infrastructure and information processing financial savings—and achieved a 5x enchancment within the pace of delivering new pipelines and integrations. That acceleration helps onboarding, shopper supply and experimentation, whereas the Databricks and AWS mixture offers Addepar the dimensions and reliability wanted to develop with its purchasers.
Addison: a local AI expertise embedded within the platform
Constructing on its unified information basis, Addepar has launched Addison, a local AI expertise embedded immediately throughout the platform. Addison is designed to offer trusted steering and actionable insights which are grounded in Addepar’s core information and workflows.
Addison goes past a chat-based interface, to:
- Dwell inside Addepar’s core platform, built-in immediately with portfolios, options and workflows.
- Perceive the “nouns and verbs” of finance within the context of Addepar’s information mannequin.
- Mix Q&A, proactive insights (push) and action-oriented workflows right into a single expertise.
- Floor related market information alongside portfolio information, serving to advisors join shopper holdings to present market occasions.
- Run on Addepar’s core calculations engine, referencing the identical portfolio metrics and efficiency calculations used throughout the platform.
For funding professionals, Addison acts like a digital associate:
- Pull: Advisors ask questions like, “Break down this portfolio’s options allocation,” “If charges rise by 50 bps, what’s the projected influence on mounted earnings length?” or “Establish any accounts which have drifted greater than 3% from the goal,” and Addison responds utilizing dwell, ruled information.
- Push: Addison surfaces notifications and occasions, equivalent to rising dangers, alternatives or anomalies in portfolios, with out requiring express prompting.
- Act: Advisors provoke workflows, equivalent to operating a monetary plan,, or exploring various allocations, perceive portfolio traits and behaviors – whereas Addison helps orchestrate the underlying information and steps throughout Addepar instruments and workflows. These capabilities are designed with people within the loop, retaining funding professionals firmly in charge of choices and actions.
The imaginative and prescient is that pure language, workflows and clever brokers change into the first manner customers work together with Addepar. By offloading tedious information manipulation and orchestration to Addison, funding professionals can focus extra time on relationships and strategic choices.
Secure, scalable GenAI for monetary companies
As a result of Addepar’s purchasers function in extremely regulated domains, Addison’s structure have to be secure and scalable in ways in which generic shopper fashions, equivalent to direct calls to public LLMs, can not match. Addepar prioritizes safety, information privateness and governance, and has designed its AI stack accordingly.
By reworking its infrastructure on Databricks, Addepar makes use of Unity Catalog, with permissions and entry controls deeply built-in into its atmosphere. Those self same controls floor in Addison. A mixture of cutting-edge frontier fashions are served and hosted inside Addepar’s atmosphere by way of Databricks Mannequin Serving, and are tracked and managed with MLflow, delivering constant lifecycle administration and auditability.
Retaining each information and fashions contained in the Addepar ecosystem is important for personally identifiable and shopper‑identifiable information throughout Addepar’s international infrastructure footprint. It helps the corporate meet shopper expectations round threat, compliance and authorized or jurisdictional considerations.
This strategy means Addison is not only an LLM endpoint. It’s an AI system that inherits the identical governance ensures as the remainder of Addepar’s platform, one thing that may be considerably more durable to attain with fragmented instruments or unmanaged exterior APIs.
From LLMs to brokers with Agent Bricks, Basis Mannequin Serving and MLflow
Easy LLM prompts may be highly effective, however making them dependable and repeatable sufficient for manufacturing monetary companies workflows is troublesome. It requires orchestration, validation and iteration to achieve the extent of consistency advisors and buyers want.
Addepar is now adopting Databricks Agent Bricks as the following evolution of its AI journey, beginning with Supervisor Agent that coordinates Genie‑powered analytics behind the scenes. Addison makes use of these Supervisor flows to maneuver from “LLM plus immediate” to trusted, agentic workflows, the place the system can execute sequences of actions on behalf of advisors with their oversight. What was beforehand a disjoint, handbook strategy of wiring collectively prompts, instruments and validation logic is now centralized and simplified by Agent Bricks, together with early use of multi‑agent Genie workflows to energy inner Slackbots and advisor experiences.
Addison leverages LLMs served from Databricks Basis Mannequin APIs, which give entry to state-of-the-art fashions from a wide range of mannequin suppliers via ruled serving endpoints. Manufacturing monetary companies workflows demand transparency, audibility, and fine-grained analysis of AI accuracy. Addepar leverages Databricks Managed MLFlow to energy traceability and granular insights into particular person agent workflows. Addepar additionally now makes use of MLFlow to develop, consider, and iterate on Addison’s efficiency and conduct.
For Addepar, all of this implies it might outline agent workflows, equivalent to multi-step portfolio analyses, planning flows or automated perception technology, check them rigorously, and deploy them with governance, all on the identical platform that powers its core information. This can be a uniquely Databricks worth proposition: unified information, governance and agent orchestration in a single place.
Collaboration and information sharing as a power multiplier
Databricks has additionally modified how Addepar collaborates internally and with purchasers. Beforehand, various kinds of customers inside Addepar and at shopper organizations typically labored in a transactional manner utilizing spreadsheets, extracts and one-off API exchanges. Collaboration was restricted and disjointed.
With Databricks Notebooks and Unity Catalog, Addepar can now share information, code and SQL in a single atmosphere with the precise entry controls. Groups can work on information and fashions in the identical place, and that collaboration extends to AI. They will share fashions, configurations and prompts with constant context. For purchasers, with the ability to view the identical information concurrently builds belief, reduces miscommunication throughout onboarding and ongoing operations, and helps a extra correct and clear view of portfolios.
A partnership targeted on outcomes
Addepar offers the foundational information platform for the funding ecosystem, bringing collectively advanced portfolio, market and shopper information to energy the workflows funding professionals depend on day-after-day. To assist the dimensions, safety and innovation the platform requires, Addepar works carefully with know-how companions like Databricks and AWS, whose capabilities assist energy key components of its information infrastructure. These partnerships are constructed round open trade and shared success somewhat than a easy vendor transaction.
As Databricks continues to advance its information and AI capabilities, Addepar expects Addison to change into the first manner many customers expertise the platform. By combining a unified, ruled information basis with GenAI and brokers, Addepar helps funding professionals reduce via complexity throughout portfolios, information and workflows to make extra knowledgeable choices and ship higher outcomes for the purchasers they serve.
Attend Databricks AI Days in a metropolis close to you to discover ways to take management of your information and construct AI brokers that drive enterprise influence.
