Published on
Data is Addepar’s foundation. One year ago, our CEO Eric Poirier introduced the Addepar Way as a standard for how we operate. The first tenet states that “Data is our compass,” reflecting Addepar’s commitment to data-driven decisions.
However, that data must be high-quality in order to provide an accurate bedrock for decision-making. Our clients require the same level of data fidelity. Now, as Addepar continues to embed AI across the platform, data quality becomes even more important for our clients.
Data quality has always mattered and AI raises the stakes
Data quality is not a new issue in computing. In fact, it predates the first computers. It arguably dates back to the 1820s when Charles Babbage documented that his Difference Engine could not possibly overcome incorrect input. The term “garbage in garbage out” (GIGO) sums up the problem quite succinctly.1 This term comes from when a computer was only calculating A-B. Now, generative AI (GenAI) is able to not only add and subtract, but create content in all forms based on the data it was trained with and the data supplied to it. Agentic AI can go one step further and take actions across systems and tools. Even though this shifts how humans interact with computers, this shift does not detract from the importance of data quality. Now more than ever high quality data is necessary to produce high quality computed outcomes. In fact, researchers from Carnegie Mellon found that in 499 publicly available harm-causing GenAI incidents, nearly one-quarter stemmed directly from incorrect or low quality data.2
This is not only an issue for GenAI and agentic AI. Researchers from the University of Potsdam and the University of Amsterdam observed that when they intentionally polluted training and testing data across the six major dimensions of data quality, there was a precipitous drop in model performance when polluting feature accuracy.3 Feature accuracy refers to whether or not the data in the dataset is accurate. Target accuracy refers to whether or not the classification, or answer, supplied for training the model is accurate. In other words, when the researchers purposefully manipulated data quality in the classic sense (i.e., the researchers made the data incorrect), machine learning algorithm performance was severely degraded. Target accuracy is the only dimension that eclipsed the impact of feature accuracy on model performance.
The emerging GenAI divide
GenAI presents the potential for organizations to undergo significant change and realize massive efficiency gains, but not every organization is set up for success. Researchers from MIT’s NANDA Project have coined the phrase “the GenAI Divide” to refer to the disparate outcomes of AI programs from enterprise to enterprise.4 They found that 95% of integrated AI pilots never make it to full production. Companies they interviewed are willing to adopt AI tools, but find difficulty doing so with disconnected solutions that live away from their existing workflows and data.
The firms that overcome this divide share a common strength: they adopt solutions that work where their data already lives and that integrate directly into the workflows teams use every day. Connectivity is key to these companies. Agentic AI has even greater potential. McKinsey found that, if leveraged correctly, mid-sized asset managers could capture between 25 and 40 percent of total cost base in AI-enabled efficiencies.5 However, to realize these efficiencies, the authors stress the importance of unified data platforms and robust governance strategies.
EY found that 75% of wealth managers have created GenAI teams to enhance their firms’ workflows; Accenture has that number at 78%.6,7 Furthermore, our research, conducted with clients and prospects and focused on workflows, found that 63% of respondents expect a high impact of AI on their daily processes.8 Accenture found that 96% of their respondents foresee GenAI “revolutioniz[ing] client servicing and investment management”; our research study found that 63% of respondents believe that AI will have a high impact on workflows.7,8 However, data quality remains a concern. In EY’s study, 67% of wealth and asset management firms surveyed cited data accuracy as an issue. Accenture found something similar for 77% of the financial advisors they surveyed.6,7 AWS found that over 50% of companies they surveyed across all industries do not believe that their data is ready for AI use.9
Finance faces unique exposure when AI makes mistakes
Financial institutions are also wary of possible productivity loss due to incorrect or flawed AI implementations. In fact, according to the Federal Reserve Bank of Richmond, banks’ investments in AI have actually been found to increase operational cost.10 The Roosevelt Institute warns about a variety of potential risks of GenAI in the financial sector.11 Rather than slowing adoption, these concerns are pushing firms to invest in AI that is grounded in accurate data and deeply connected to established workflows. As such, high-profile AI implementations continue, and firms are hungry to reap the benefits GenAI poses. Citi started a pilot with 5,000 employees, UBS has embarked on a three-year AI investment journey, and there are many others deploying similar high-profile initiatives.12,13,14
Why unified data unlocks real AI value
MIT’s NANDA Project highlighted several contributors to the GenAI Divide, including the finding that firms are most successful with GenAI when they use solutions that integrate deeply with their existing data and workflows and continuously adapt over time.4 This principle directly informs Addepar’s approach to AI. We’re embedding intelligence throughout the platform in ways that help firms surface insights faster from the trusted data they already manage in Addepar and act on those insights within their existing workflows. As capabilities continue to expand, we will build on this data foundation to further enhance productivity and decision-making.
Addepar aggregates and normalizes over $8 trillion in assets across multiple currencies and complex ownership structures, forming a trusted data foundation for reporting and analysis. McKinsey notes that to truly “unlock alpha” from GenAI, firms must integrate their data and systems with the AI solutions they select.15 Addepar provides this integration and access from day one.
Better data wins every time
In 8VC’s the AI Wave, they report on “first-generation smart enterprises” that “have made the successful transition to becoming significantly more assistive or automated. LLMs enabled the logical evolution of the product, building on best-in-class data integrations and workflow consolidation.”16 Specifically, “Addepar has built on its position as the operating system of finance to pioneer such capabilities as complex workflow automation, AI usability assistants and risk and opportunity intelligence.”
AI is only as strong as the data behind it. The most advanced intelligence begins with trusted data, like the unified data in Addepar. Connect with us to discover how Addepar can accelerate your firm’s workflows with the high-quality data you already rely on.
References
Garbage in, garbage out (GIGO), EBSCO Knowledge Advantage, 2024.
A closer look at the existing risks of generative AI: Mapping the who, what, and how of real-world incidents, Carnegie Mellon University, 2025.
The effects of data quality on machine learning performance on tabular data, Information Systems, 2025.
The GenAI divide, state of AI in business 2025, NANDA MIT, 2025.
How AI could reshape the economics of the asset management industry, McKinsey & Company, 2025.
Five priorities for winning with GenAI in wealth and asset management, EY, 2024.
Using generative AI to power growth for wealth managers, Accenture, 2025.
Addepar Workflows Research [Unpublished internal report], Addepar, 2025.
Data and generative AI: A window into your organisation’s soul?, AWS Cloud Enterprise Strategy Blog, 2024.
Does AI cause higher operational losses at banks?, Federal Reserve Bank of Richmond, 2025.
The risks of generative AI agents to financial services, Roosevelt Institute, 2024.
AI agents arrive at Citi, The Wall Street Journal, 2025.
UBS’s Rob Karofsky on using AI to boost US wealth arm: ‘60% of our efforts are focused on productivity,’ Financial News, 2025.
Banks accelerate AI deployments as agentic tools gain traction, CIO Dive, 2025.
Charting a path to the data- and AI-driven enterprise of 2030, McKinsey & Company, 2024.
The AI wave, 8VC, 2025.