Skip to content


Generative AI

Gen AI for images

Gen AI for texts

Gen AI for data visualization

WSJ Global Food Forum 2023

View from the venture capital
Shiru - use generative AI for optimizing and identifying plant proteins (can replace palm oil with natural enhanced fat substitute)
Generative AI as a tool → more important is the scalability to bring these products to market (business model, GTM strategy)
Some companies come up with innovative business models that reduce their capex
Business model and resilience to make decisions in this tough macro environment
From farm to fork
Rising food prices but lower supply (higher production cost, adverse weather)

Generative AI in Financial Services

My take

Generative AI
AI that can generate new content/ data e.g. ChatGPT can generate human-like text
Create personalised content e.g. customised itinerary, shopping recommendations
Create synthetic data when there is a shortage of training data for training machine learning models
Important points
There are 3 ways organizations can use to build generative AI models
Construct: Build a custom large language model (LLM) using proprietary datasets.
Configure: Use existing LLMs and customize them for specific use cases. → for DiMuto proprietary AI applications e.g. Trade Health AI
Adopt: Purchase a complete generative AI application for a specific use case with little customization. → for general applications e.g. answering FAQ
Generative AI is used as a component of a product/ software, and not the product itself.
No moat for companies in this highly competitive and open-source AI landscape.
We have no moat, and neither does OpenAI. - Google
Generative AI will complement traditional AI.
Generative AI needs to serve a use case e.g. expand revenue, lower costs → automating manual tasks in the agriculture industry e.g. looking through trade documents and images
AI is poised to offer personalised consumer experiences, make operations more cost-efficient, enable dynamic forecasting and reporting.
Generative AI can give personalised answers and factor context into decision making. → making recommendations on how companies can make their operations more sustainable


Generative AI is being adopted in financial services to complement traditional AI.
Why? It's versatile, enhances customer experience, and can be integrated in various applications through custom-building, configuring existing models, or adopting ready-to-use applications.
Trends: Its adoption follows the Gartner hype cycle and mainly benefits incumbent firms by improving existing products. It's accelerating trends like embedded finance and open banking.
Implementing generative AI: Companies should have clear goals, assess readiness, select vendors wisely, adopt an agile approach, address data privacy, manage costs, and solve AI limitations such as the "black box" issue.
What’s next: Monitoring the evolving AI space, regulatory changes, and tech stack dynamics is crucial for future success.

Neither Evolution Nor Revolution
Financial services are extensively using AI, including both traditional AI and emerging generative AI.
Generative AI is flexible and can perform tasks traditional AI cannot.
Generative AI won't replace traditional AI as they serve different purposes.
Generative AI and traditional AI will coexist and complement each other within financial services organizations.
Generative AI Is Essential for Financial Services
Generative AI fills gaps unaddressed by traditional AI, improving customer experiences and organizational profitability.
It simplifies complexity and can tackle bespoke requests and queries.
Generative AI can help make financial services more emotionally resonant and attuned to customers' needs.
Generative AI as a Fundamental Building Block
Generative AI represents a platform shift, allowing interaction with technology in natural language.
It is expected to be embedded in traditional AI and other technologies, enhancing their capabilities.
Generative AI will be integrated into various financial applications and can be distributed through interfaces beyond chatbots.
Three Models for Leveraging Generative AI
Construct: Build a custom large language model (LLM) using proprietary datasets.
Configure: Use existing LLMs and customize them for specific use cases.
Adopt: Purchase a complete generative AI application for a specific use case with little customization.
The choice between these models depends on organizational goals, budget, data availability, time constraints, data sensitivity, and internal skills.
Generative AI is used within financial services
Financial services companies and fintechs are increasingly adopting generative AI.
The list of applications is constantly growing and not all investments and applications are disclosed.
Companies may disclose their use of generative AI to showcase innovation, demonstrate customer value, or as a competitive strategy.
Generative AI use cases in financial services can be domain-agnostic or domain-specific.
Domain-agnostic applications can be used across industries, such as customer support or marketing. Example: Chime partnering with a startup to train an internal language model for faster product launches.
Domain-specific applications are tailored for financial services, like risk modeling or fraud detection. Example: SlopeGPT by Slope Pay for custom risk underwriting, and Navan's "Ava" virtual assistant for expense reporting.
Teams implementing generative AI in financial services may use different approaches - constructing in-house solutions, configuring existing tools, or adopting third-party services.
Over time, most generative AI use cases are expected to fall into the "Adopt" category due to the technical complexity of building AI/ML solutions in-house.
Generative AI Hype
The adoption of generative AI in financial services follows the Gartner hype cycle, with an initial surge of interest, followed by disillusionment, and eventually integration and productivity.
When evaluating generative AI announcements in financial services, differentiate between hype (vague intent), hope (specific plans but not yet fully realized), and happening (in production with measurable impacts).
Financial services companies have found success in integrating generative AI into specific workflows or tools they are already using.
Generative AI benefits incumbents
Generative AI is often hailed as a disruptive technology, but in the context of financial services, it is largely a sustaining technology.
Incumbents, which include large financial services companies and established fintechs, can use generative AI to enhance existing products, benefiting from their existing distribution and customer knowledge.
Generative AI, as defined by Clay Christensen, is primarily a sustaining technology as it improves the performance of established products, while disruptive technologies result in initially worse product performance but have features that new customers value such as being cheaper or simpler.
Incumbents have an advantage in the generative AI “arms race” as they can establish partnerships, hire skilled teams, and leverage proprietary data to build technology using generative AI.
Disruptive technologies are usually ignored by incumbents as they generally result in simpler, cheaper products with lower margins, and don't appeal to the most profitable customers.
Generative AI can still be disruptive in certain instances within the financial services industry where it's not in the best interest of incumbents to offer a generative AI product.
Generative AI is accelerating trends within financial services
Generative AI is also accelerating pre-existing trends in financial services like embedded finance and open banking.
Embedded finance benefits from generative AI in aspects like fraud detection, risk management, personalized user engagement, and customer support.
Open banking is enhanced by generative AI by creating more utility from open banking APIs, enabling better leveraging of real, personal data for services like balance checking, transaction finding, budgeting discussions, and payments.
Advice for companies looking to leverage generative AI
Determine the purpose and goals for integrating generative AI in your organization.
Assess your organization’s readiness in terms of goals, budget, data sophistication, time, and skills.
Select vendors and partners carefully based on your use case, industry, scale, and sophistication level.
Adopt an agile approach: start small, scale what works, adapt or discontinue what doesn't, and learn from every experiment.
Encourage creativity and experimentation throughout the organization and create a culture that values new ideas.
Engage with others in the field for shared learning experiences and collaborations.
Be cautious of data privacy, ensure you comply with regulations and devise strategies to protect sensitive data.
Tackle the "data context" issue by preparing your data through pre-processing and fine-tuning on proprietary data.
Address the issue of AI "hallucinations" by creating guardrails for AI model output and employing safety-oriented approaches.
Solve the "black box" issue by using AI models that provide clear reasoning for their recommendations.
Be mindful of costs, especially in terms of computing power, and find solutions like query concatenation and caching.
What’s next
Monitor the evolution of the generative AI space, including regulations, development of domain-specific models, and the competitive dynamics of the generative AI tech stack.
Be prepared to address regulatory actions as governments might impose new rules on AI systems for safety, transparency, and ethics.
Pay attention to the development of domain-specific models that are custom-trained and may offer cost-effective solutions in specific contexts.
Evaluate which layer of the generative AI tech stack is the most valuable and defensible for your specific use case and industry.

Want to print your doc?
This is not the way.
Try clicking the ⋯ next to your doc name or using a keyboard shortcut (
) instead.