AI’s promise and power have led many organizations to race up the adoption curve, especially when it comes to customer relationship management (CRM) applications. Many are tempted to use AI itself to clean up their customer data in advance of deploying downstream AI applications that will yield meaningful productivity and revenue upsides. But is that really a good idea?
That’s what banks are asking. And the answer, at least according to a recent study by Forrester, is a resounding “not so fast.”
Forrester warns that unleashing AI on disorganized CRM data without a strategy or governance practices could create data chaos. Companies should instead organize their data and formalize the governance through their data management practices before trying to build AI into CRM systems.
Getting the sequencing right has implications that go far beyond mere timing. Companies that get too far ahead of the AI curve may set themselves up for bad results later on, because they’ve trained their systems on incomplete or inadequately organized data, according to the study, The Journey To AI-Powered CRM.
The report is based on a survey of 773 global decision-makers, including C-level executives, vice presidents, and directors. Some 18% of the respondents work in the finance and insurance industry.
A key takeaway is that most companies do in fact appear to be going too fast.
Solve the Puzzle of Core Deposit & New Client Growth
In this strategy-centered webinar, Crack the Code of Core Deposit & Client Growth, learn how to create sustainable deposit and client growth. Watch Now.
Read More about Solve the Puzzle of Core Deposit & New Client Growth
This FI Built Two Branches Without Adding Consumer Lending Employees.
Heartland wanted to expand. Being short-staffed made it hard. Here’s how deploying a new technology helped them build two new branches anyway.
Read More about This FI Built Two Branches Without Adding Consumer Lending Employees.
Here’s the problem: Two thirds of respondents in Forrester’s survey say their firms do not have a formal company-wide data strategy, and more than a quarter say they have no formal strategy at all. Yet more than half of respondents say their companies are already using AI to complete online commerce transactions and to conduct marketing outreach. And more than 60% say they plan, within the next year, to develop AI tools for “hyper-personalization for one-to-one engagement using first-, second-, and third-party data.”
A Rising of the Tides:
More than three-fifths of executives say they plan to roll out AI tools to hyperpersonalize customer experiences in the next year — specifically for one-to-one engagement.
“We found that many are embracing AI-powered CRM without ensuring they have the necessary data infrastructure, making them more vulnerable to undesirable outcomes,” the report said.
AI’s potential within banking is enormous. For context, a McKinsey & Co report says banking is among the sectors that will gain the most from the new technology’ application – potentially by as much as $340 billion annually including cost savings and upsides.
But poorly trained AI-driven CRM could lead to unchecked inaccuracies due to AI-generated inaccuracies, or “hallucinations,” as well as leaks of sensitive customer data. Without being properly monitored, AI could create unintended discriminatory behavior such as systematically avoiding underwriting loans to specific populations, for example.
Forrester’s report, which was commissioned by Salesforce, offers five recommendations on how to manage AI in CRMs.
Read more:
- How to Find the Right CRM System and Make the Most of It
- Building a CRM from Scratch? Why That’s a Big Mistake for Banks
1. No Garbage In, No Garbage Out: Start with Clean and Unified Data
Siloed data is a major liability for banks seeking to deploy AI CRM solutions. A banking company may store client data relating to checking, mortgage, and business accounts on three different systems. Without a unified 360-degree view, AI systems may be trained on outdated and incomplete information. Core systems are backed up by several layers of redundancy, and data extracted from them must be anonymized in keeping with privacy regulations – meaning that knocking down silo walls can take years and cost tens of millions of dollars.
But feeding quality data to an AI model is the sine qua non for getting worthwhile responses. “Without high-quality, well-structured, and clean data, AI algorithms will struggle to deliver meaningful insights and outcomes,” reads the report.
2. AI Outputs Require Governance – Just Like Every Other Data Output
Data governance describes internal standards on gathering, storing, and disposing of data to ensure it is accurate and consistent, protected from privacy breaches, and meets regulatory standards. Everything that emerges from an AI-driven CRM – from chatbot responses to highly tailored customer solutions – should be held to the same standards, Forrester says.
“We found that many are embracing AI-powered CRM without ensuring they have the necessary data infrastructure, making them more vulnerable to undesirable outcomes.”
— Salesforce report
Proper governance will also make data easier to audit, which in turn helps track mistakes before they get repeated and propagated through the system. Forrester further recommends that companies establish new compliance and/or brand management positions whose job it is to shape company policy when using AI models.
3. You Probably Need an External Partner; Make Sure They’re Trusted
Many credit unions and smaller banks rely on in-house CRM solutions. Such systems are harder to customize because they rely on employees who have other responsibilities and are sometimes deployed for specific functions rather than across the entire institution. In fact, respondents said they are not fully confident in their employees’ ability to use AI. (This mirrors a broader trend in banks, one third of which prohibit employees from using tools such as ChatGPT, according to American Banker.)
Adding AI to CRM systems makes it even more important to seek an external partner that can create and manage large language models. Ninety-six percent of respondents said trust was the principal factor in choosing a vendor for AI solutions, specifically around security and data protection.
4. Teaching Matters Because Everyone Is on a Learning Curve
Companies need to invest in training employees on the use of AI, Forrester notes. Even at firms with company-wide data strategies, more than one third of respondents were unable to correctly distinguish between generative AI and predictive AI. At firms without company-wide data protocols, the figure rose to almost 60%.
5. Don’t to Capture Cost Savings; Instead Rethink Collaboration and Productivity
Deploying AI in CRM solutions presents an opportunity to rethink the evolution of a company’s workforce, Forrester says. Rather than downsizing operations, companies can find ways for customer service generalists to transition into roles such as prompt engineers. This can help banks retain employees who already know the organizational culture and best practices.
Brian Ellsworth is a Washington-based journalist and communications advisor. He is a former Reuters correspondent who covered Latin America and the Caribbean for 20 years.