Connect with the author: olly.downs@curinos.com
As organizations rush to deploy agentic AI and autonomous decisioning, one prerequisite is often underestimated: unified, curated and well-described data. Without it, AI simply accelerates siloed thinking—at scale.
Decision intelligence depends on understanding customers, actions and outcomes across time. When data remains locked in product silos, even sophisticated AI agents optimize locally while destroying value globally.
Retail banking provides a powerful example. When banks analyze deposit data in isolation, they tend to prioritize acquiring “mass affluent” customers—those with higher initial balances and short-term appeal. It feels rational… until you connect the dots.
When deposit and lending relationships are unified at the customer level and outcomes are measured over a longer horizon (3–5 years), a very different story emerges. Many so-called “mass market” customers with relatively low short-term deposit payoffs expand into profitable, multiproduct relationships (see chart). With the right targeting, banks can achieve similar long-term value per customer, while dramatically increasing the total acquisition opportunity due to the larger addressable segment.
Average Customer Balance Change
Over Time, by Segment
Source(s): Curinos Deposit Analyzer; Curinos Distribution Analyzer; Curinos Analysis
Note(s): 1. Customer base established as those who are with the bank at three months on book (does not account for attrition in first three months) 2. Average balance indexed to customer balances at three months on book
Based on consumer customers who entered through checking | Mass Market defined as <$10K deposit balances at M3, Mass Affluent >$10K deposit balances at M3 | Proforma view assumes 3MOB starting dep. balance of $10K (MA) and $1K (MM)
This is the hidden cost of data silos—entire growth strategies optimized around an incomplete view of value.
Ironically, AI itself can help solve this—by inferring connections, improving semantic descriptions, flagging quality issues and accelerating data mapping when paired with human governance. But AI only amplifies the foundation it’s built on.
Before deploying agents to make decisions, make sure your data can see the whole customer.
