Building Next Best Action at Kristal.AI

Product Strategy

Executive Summary

As Product Director, Data & AI at Kristal.AI, I led the development of an intelligent Next Best Action (NBA) framework that revolutionized how our wealth management platform engages with mass affluent clients. This initiative addressed the core challenge of delivering personalized, scalable wealth management advice to clients with investible assets between $1-5 million3. The NBA system increased client engagement rates by enabling timely, data-driven recommendations across our hybrid digital-human service model.

Challenge & Problem Statement

Business Context: Kristal.AI serves the underserved mass affluent segment, providing a hybrid approach that is "roughly 80% digital and 20% human"3. Our wealth managers were struggling to serve many clients effectively while maintaining the personalized touch expected in wealth management.

Core Problems Identified:

Wealth managers couldn't deliver personalized recommendations at scale across the client base

Client engagement was declining, with risks of dormancy and churn

Manual processes prevented timely interventions for critical portfolio events

Lack of systematic approach to prioritizing client actions based on their circumstances

Business Impact: The problems manifested as slower AUM growth, reduced client engagement, and increased client dormancy rates.

Product Strategy & Approach

Vision: Empower wealth managers to engage with clients in a "timely, personalized, and scalable manner" by "prioritizing the actions that the client should be taking to maximize their interest".

Strategic Framework: I designed a four-component NBA system that combines AI-driven insights with business rules:

Component Architecture:

1Identify Circumstances: Leverage segmentation and insights engines to assess client situations

2List & Prioritize Actions: Use prioritization engine to rank possible recommendations

3Map Circumstances to Actions: Connect client portfolio signals to specific action types

4Deliver Recommendations: Send prioritized NBA through our recommendation engine

Product Methodology: Rather than building a monolithic system, I opted for a modular approach that could integrate with our existing wealth management infrastructure while allowing for rapid iteration and testing.

Technical Implementation

Data Platform Architecture:

Built a comprehensive data foundation incorporating:

Client portfolio analytics and segmentation engine

Real-time insights engine for circumstance detection

Predictive models for client behavior forecasting

Business rules engine for compliance and suitability

AI & Machine Learning Components:

Implemented predictive models to "forecast customer behavior and preferences"

Developed propensity scoring for action effectiveness

Created dynamic prioritization algorithms combining business value and client likelihood

Action Library: Designed a comprehensive library of 19 pre-built action types covering the full client lifecycle:

Priority	Model Name	Circumstance	Action
1	Recently KYC Approved	People who recently completed the KYC	Know your target return to match your goals
2	Invest in Model Allocation	People who have tried GTC or who have are KYC approved but have not invested yet.	Try Customized portfolio as per your target returns
3	High Potential Investors	People who have high appetite to invest but have not invested with us yet	Consider products in your community
4	Sell List Holders - Minor Investors	People who have less than 10% of their portfolio in Sell List	Sell the products in the Sell List
5	Sell List- Major investors	People who have more than 10% portfolio in Sell List	Review portfolio and offload the product(s) gradually
6	Negative Cash holders	People who have negative account balance	Negative cash balance in your account, take one of the following actions
7	Upcoming Calls	People who have upcoming call in next 7 days	Talk to your advisor to avoid missing the call date
8	Cash Holders	People who are holding a significant amount of cash	Invest in money market to earn higher yields
9	Model Portfolio Deviators	People who deviated from model portfolio a lot	Re-balance to align with model portfolio
10	Likely To Dormant Customer	People who are likely to stop using the platform in near future	Timely opportunity to invest in a high yield fund
11	Dormant	People who have stopped using the platform	--
12	High Risk Exposure	Portfolio is exposed to higher risk than accepted in intended	Re-balance your portfolio to lower than Risk Exposure
13	Retirement seekers	Clients who are closer to retirement or might be doing retirement planning	Consider these Fixed Income products to plan a stress free retirement
14	Kids Education Planners	People who have kids and planning their education in near future	Consider these long tenor bonds to funds your child's future
15	Money Market focused	People who have more than 10% portfolio in Money markets	Invest in long-tenor bonds
16	Next Best Product	People who need to find next best product to invest in	Invest in this XYZ as it will improve Sharpe Ratio by YY%
17	Social Validators	People who wants to know what other people like them are doing	--
18	Potential PW client	Clients who are Digital but show PW behavior	You're eligible for PW services. Upgrade your account to engage with an advisor
19	High Product Views	Clients who viewed certain product types multiple times	Top XYZ products for you to consider adding in your portfolio

Channel Integration: Enabled NBA delivery across multiple touchpoints, supporting our hybrid service model with both digital and advisor-mediated interactions.

Product Differentiation Achieved

Market Positioning: The NBA framework differentiated Kristal.AI in the competitive wealth management space by providing "the ease of DIY platforms combined with the added value of smart curation, greater access and more personalised advice".

Technology Integration: Successfully integrated NBA with Kristal.AI's broader technology strategy, including APIs for private markets access and blockchain technology for illiquid asset liquidity.

Scalability Design: Built the system to support Kristal.AI's partnership model, where "hundreds of partners, each bringing perhaps very small international allocations" could benefit from sophisticated NBA capabilities.

Results & Impact

Operational Efficiency:

Automated identification of 19 distinct client circumstances requiring intervention

Enabled systematic prioritization of client outreach based on business impact and urgency

Reduced manual portfolio monitoring time for wealth managers

Client Experience Enhancement:

Delivered personalized recommendations that align with individual client goals and risk profiles

Enabled proactive communication for critical events (e.g., upcoming bond maturities, sell list exposures)

Improved timing of interventions through data-driven trigger mechanisms

Business Growth:

Supported Kristal.AI's expansion to serve "hundreds of partners" in the EAM ecosystem

Enhanced the platform's ability to "improve the chances of clients accepting product recommendations"

Contributed to the firm's mission of making wealth management more accessible to the mass affluent segment

Key Learnings & Strategic Insights

Data-Driven Product Development: The success of NBA reinforced the importance of combining business domain expertise with AI capabilities. The 19 pre-built action types emerged from deep understanding of wealth management workflows rather than pure algorithmic optimization.

Hybrid Human-AI Approach: Rather than replacing wealth managers, NBA augmented their capabilities, aligning with Kristal.AI's hybrid service philosophy3. This approach proved more effective than fully automated solutions.

Continuous Iteration: Implemented feedback loops to continuously improve action prioritization and effectiveness measurement, similar to how "Netflix or Amazon hone their suggestions based on past usage".

Partnership Enablement: Designed NBA to make partners "feel in some ways more powerful, more capable, even superhuman for their own clients," supporting Kristal.AI's B2B2C strategy.

View all projects