• Nick Goble, Director at Solution Architecture, Financial Services and Insurance, Domino Data Lab


• 11.12.2025 11:15 am
#AIPredictions2026 #FinancialServicesAI

Prediction 1: The Year AI Goes to Work

The first few years of AI experimentation have passed and companies will shift from learning to measuring. Measuring the cost, benefits, and trade-offs associated with adopting innovation early. There is still plenty of room for exploration, but there will be increased pressure for exploitation to achieve positive ROI. 

Prediction 2: The Reckoning of AI Experiment Technical and Operational Debt

AI experiments generate costs and risks as companies seek to gain first adopter benefits. However, moving these experiments to production will require standardised, collaborative model development and risk management frameworks. 

Innovation is necessarily messy, leaving a trail of tech debt and arbitrary uniqueness that is hard to manage, expensive to fund, and full of hidden risks. In 2026, companies will see their AI experiments mature and move to production where the expected benefits will become more accurately measurable. Once operationalised, any risk with the models or their implementation will quickly become apparent. 

To address cost, complexity, and risk management needs, companies will implement unified model risk development and governance platforms that allow cross-discipline teams to collaborate. Only the most sophisticated companies will do this and they will see reduced cost, better risk management, and increased model throughput. Most other companies will spend the next few years in discovery mode and spend more money to get higher risks and longer model-to-value cycle times. 

Prediction 3: Insurance Industry Prediction: Fraud, Inflation, and Climate Issues Abound

The insurance industry will face three simultaneous threats. AI-powered fraud will explode, uneven building material price increases will contribute to more expensive claims, and climate volatility will drive claim volume and size increases.  

Fraudsters use AI to create fake documents and media, impersonate people through voice and video deepfakes, automate social engineering, and generate false evidence for insurance claims. Companies will need to build and deploy sophisticated detection models at a faster pace than fraudsters. Building material costs will continue to rise unevenly and unpredictably. The combination of inflation and changing tariffs will mean that claims pricing models will need to be updated frequently. Potentially the most complex issue is that climate impact models are hugely computationally intense, the input data is experiencing high variance, and the location precision needed is increasing.  

These issues all share a common challenge. While it may be possible to build models quickly, insurance companies will spend too much time validating the models, which will decrease their effectiveness. Models need to be updated frequently because the drivers of their accuracy change so quickly. This will lead to insurance companies moving to a ‘model factory’ approach to increase model build and refresh cycles.  

Prediction 4: The Collapse of Traditional Credit Risk Models

Credit underwriting and loss models have improved in the last few years to accommodate many more features than simply credit scores. One thing not in current models is the representation of job change due to AI and automation. There are 3 AI and automation scenarios often mentioned; more jobs will be created than destroyed, or the net effects will be about even on the job market, or many jobs will be lost and never regained. Regardless of the scenario you believe, it is clear that job disruption will occur and the impact is not built into current credit models.  

The failure to accurately price risk will compel banks to change their underwriting and loan loss models. This will prove tricky as these models cannot be built on broad speculation. The challenge is that everyone thinks ‘something’ is going to happen, but they don’t have the precision to know ‘what’ is going to happen. Because of this banks will build challenger models for various scenarios and run them simultaneously with production models. Models will also be adjusted to include things like skills, roles, and industry. Instead of improving credit scores, people will seek new certifications to improve their ‘role-risk’ score. This will prove to be highly controversial and require a tremendous amount of brand risk management. 

Prediction 5: Asset Management Prediction: Platform-as-a-Service Emerges

Trading models are built using highly secret individual tech stacks, data, and modeling techniques. Only one of those things is a true advantage yet, trading desks continue to demand independence which leads to extra cost, delays in validating and implementing trading strategies, and lost opportunities. 

Asset management teams will finally realize the secret sauce isn’t in the tech stack, but in the ability of the trading desk team to develop unique insight. Therefore, in 2026, asset management firms will mandate unified model development and risk management platforms that balance independence and speed with cost containment and risk management.  

Prediction 6: 'The Dawn of Ethershoring and Autoshoring

First we had offshoring, then, onshoring. The next wave will be ethershoring and autoshoring. 

In the beginning of offshoring, routine, repeatable tasks were moved. Next more sophisticated things moved, but then there was a backlash and onshoring brought jobs back to America. The market found an optimal balance of on and offshore capabilities and technology evolved to make it irrelevant where work is physically done. For those things requiring a physical presence, there have been huge gains in automation capabilities. 

In 2026, I expect more attempts at employment reduction via AI and Automation’ I’m calling this ethershoring and autoshoring. We’ve already seen a few high-profile attempts at both and a corresponding swing back to getting actual humans to do the work. 

This is a test-and-learn phase as companies try to optimise their workforce between AI, automation, and humans. While the offshore and onshore battles took decades to play out, I predict that the optimal state will be obtained in the next 10 years as AI and robots get better at record-setting rates.