Traditional, correlational, Machine Learning approaches are often not sufficiently trusted or capable enough to address some of the most crucial questions across the customer journey, credit portfolio, marketing, model risk, insurance claims, insurance policies, AI regulations as well as centralised planning & strategy
Traditional machine learning approaches often fail to address critical business questions
Spurious correlations lead to bad decisions
And are often perceived as “black boxes”
Do these questions sound familiar?
Financial Services run on causal questions
- What are the best actions (interventions) to retain or up-sell my customers?
- How do I best price my products to optimize for market share or profitability?
- How does my competitor’s behavior impact churn and customer lifetime value?
- How do interest rates, inflation, FX rates, and other external drivers impact the creditworthiness of my customers?
- What are the causal drivers of credit risk?
- What are the best actions (interventions) for loans that are defaulted or about to default?
- How should I rebalance my portfolio to optimize risk-adjusted returns?
- Based on my budget, what is the optimal allocation of investment across channels (e.g.: media, digital, and mail)?
- How well did my last marketing campaign perform? Why?
- What are the causal drivers of campaign performance?
- What are the best interventions to improve campaign performance?
- Has the model accounted for confounders?
- Does the model include any spurious correlations?
- Is the model affected by Simpson’s Paradox? (Does the full population relationship hold for subpopulations?)
- How would the model behave in unseen scenarios that I haven’t observed before?
- Is my model biased towards gender, race, and other protected classes?
- Is my model fair, and can it withstand the scrutiny of any external audit?
- What are the root causes of claims costs?
- What is the causal impact of inflation on claim costs?
- Does a partnered network (e.g. garages) help reduce claim costs? Which claims are they most effective at reducing?
- What is causing the claim in our policy network? Can we reduce these with physical features or training?
- What is the causal impact of claims experience on renewal?
- Would auto-approving all claims vs the cost to investigate fraud be beneficial?
- What is the optimal price for our policy?
- What happens if I offer a rewards programme?
- What is our policies’ true price elasticity?
- What policy would be most relevant to a potential client?
- How can I fairly approve Life & Health policies?
- What is the causal impact of market conditions on demand for policy?
- How do interest rates and inflation dynamics affect the risk profile of the entire portfolio?
- How will the competitive, market, and macro dynamics impact my PnL?
- How will decisions in pricing, marketing or risk management affect the PnL and the risk profile of the entire portfolio?
the first operating system for decision making powered by Causal AI, to address all those causal questions
To move beyond traditional ML and into a world where you can provide actionable recommendations by leveraging state of the Causal AI tools and methods.Learn more
Seamlessly surface recommendations to your business partners as expressive, tailored and interactive applications focused on decision-making.Learn more
Trusted by leading organisations
Causal AI at Scotiabank
Watch the talk from the Causal AI Conference 2022
A large European bank with over €400bn in aggregate deposits wanted sees a return of €12mn using decisionOS to better understand and manage deposit risk.
North American pension plan improved beneficiary satisfaction and increased retention by 17% using decisionOS powered by Causal AI
A leading Mobile App company sees a projected 15x ROI through a reduction of 5% in annual marketing spend using decisionOS to optimise marketing allocation
Proven value in weeks
Two to three hours
3Proof of Concept
Three to four weeks