By George Trujillo, Principal Data Strategist, DataStax
I recently spoke with a senior executive who had just joined a new organization. He was trying to gather new data insights, but he was frustrated with the time it took. (Does that sound familiar?) After explaining his management team about data hops, flows, integrations, and processing across various ingestion software, databases, and analytics platforms, they discuss the current data his architecture and technology his stack. I was shocked by the complexity of It was clear that things had to change if organizations were to be able to execute quickly in real time.
Data is a critical component when it comes to making accurate and timely recommendations and decisions in real time, especially when organizations seek to implement real-time artificial intelligence. Real-time AI involves processing data to make decisions within a specific timeframe. The time frame window can be specified in minutes, seconds, or milliseconds based on your use case. Real-time AI combines streaming data and machine learning algorithms to make fast, automated decisions. Examples include recommendations, fraud detection, security monitoring, and chatbots.
A lot has to happen behind the scenes to be successful and deliver tangible business results. The underlying architecture should include event streaming technology, high-performance databases, and machine learning feature stores. All of these must work together in a real-time ecosystem to support the speed and scale required to realize the business benefits of real-time AI.
It’s not easy. Most modern data architectures are designed for batch processing with analytical and machine learning models running in data warehouses and data lakes. Real-time AI requires a different mindset, a different process, and faster execution speeds. In this article, we share insights for aligning vision and leadership as well as reducing the complexity of making data actionable to deliver real-time AI solutions.
Real-time AI North Star
Time and time again I see senior management aligned perfectly with the mission while the team fights a nuanced but intense battle of attrition between different technologies, silos, and beliefs on how to execute the vision. I’ve seen
A clear vision for executing a real-time AI strategy is a critical step in aligning executives and line-of-business leaders on how real-time AI will drive business value for the organization.
An action plan should be created from a shared vision that provides transparency. This includes a well-defined list of methodologies, technology stacks, scope, processes, cross-functional impacts, resources, and measurements. Collaborate and work together to achieve operational goals.
Machine learning models (algorithms that comb through data to recognize patterns and make decisions) rely on the quality and reliability of the data created and maintained by application developers, data engineers, SREs, and data stewards. depends. How well these teams work together determines the speed with which we can deliver real-time AI solutions. As real-time permeates the organization, some questions begin to arise.
- As machine learning models evolve, how can cross-functional teams support the speed of change, agility, and data quality of real-time AI?
- What level of alerting, observability, and profiling can you expect to ensure business trust in your data?
- How do analysts and data scientists find, access, and understand the context of real-time data?
- How are data, process, and model drifts managed for reliability?
- Downstream teams can create strategy drift without a well-defined and governed execution strategy. Is your strategy consistent, evolving, or starting to drift?
- Real-time AI is a science project until business benefits are realized. What metrics are used to understand the business impact of real-time AI?
As the range increases so does the need for a wider range of adjustments
The growth of real-time AI in organizations impacts execution strategies. New projects and initiatives tend to run at the edge of the organization, whether it’s adding intelligent devices to streamline operations, improving real-time product recommendations, or pioneering new business models in real-time.—by subject matter experts, evangelists, and other innovative individuals.
The edge is away from the center of gravity of your business. Away from fixed interests, vested interests, and traditional thinking.
Because the edge has less inertia, it can easily foster innovation, new thinking, and fresh approaches compared to an organization’s traditional business units, institutional thinking, and existing infrastructure. Business transformation occurs when innovation at the edge can move to core business lines such as operations, e-commerce, customer service, marketing, human resources, inventory, and shipping/receiving.
Real-time AI initiatives are science projects until they demonstrate business value. Tangible business benefits such as increased revenue, reduced costs in operational efficiency, and better decision-making should be shared with the business.
Scaling AI from the edge to core business units requires a continuous effort to manage risk and change, demonstrate value and strategy, and strengthen the culture around data and real-time AI. Without metrics and results demonstrating the business value achieved by AI at its current level, AI should not move deeply into the core of an organization. Business outcomes are the currency for AI to grow within an organization.
real-time data platform
Here, we present the current state of most data ecosystems compared to the real-time data stacks required to drive real-time AI success.
data stacks
Leaders face the challenge of executing a unified and shared vision across these environments. Real-time data does not exist in silos. It flows in two directions throughout the data ecosystem. Data used to train ML models can reside in memory caches, operational data stores, or analytical databases. Data needs to be sent back to the source to provide instructions to the device or recommendations to the mobile app. An integrated data ecosystem makes this possible in real time.
data stacks
Within the real-time data ecosystem, the heart of real-time decision-making consists of real-time streaming data, ML feature stores, and ML feature engines. Reducing complexity here is important.
data stacks
We highlighted how data for real-time decision-making flows bi-directionally between data sources, streaming data, databases, analytical data platforms, and the cloud. Machine learning capabilities include data used to train machine learning models and data used as inference data when the models run in production. A real-time model that makes real-time decisions requires an ecosystem that supports the speed and agility to update existing models and operationalize new models across the data dimensions listed below.
data stacks
The real-time data ecosystem includes two core components: a data ingestion platform that receives real-time messages and event streams, and an operational data store that aggregates and persists real-time events, operational data, and machine learning capability data. increase. These two fundamental cores must be aligned for agility across edge, on-premises, hybrid cloud, and multi-vendor cloud.
The complexity of disparate data platforms doesn’t support the speed and agility data needs to support real-time AI. Changing standards, new data, and evolving customer conditions can quickly make machine learning models obsolete. Data pipelines flow through memory caches, dashboards, event streams, databases, and analytics platforms where new data conditions need to be updated, changed, or injected. The complexity of the entire data ecosystem impacts the speed at which these updates can be performed accurately.
A unified multi-purpose data ingestion platform and operational data store greatly reduces the number of technology languages your team must speak and the complexity of manipulating real-time data flows across your ecosystem. A unified stack also improves the ability to scale real-time AI across the organization. As mentioned earlier, reducing complexity also improves the cohesion of the various teams supporting the real-time data ecosystem.
New real-time AI initiatives should consider the right data technology stack through the lens of what it takes to support evolving machine learning models running in real-time. This does not necessarily require ripping and replacing existing systems. Minimize disruption by running new data through a refreshed, agile real-time data ecosystem, slowly migrating from the data platform to the real-time AI stack as needed.
summary
Moving real-time AI from the edge of innovation to the heart of business will be one of the biggest challenges for organizations in 2023. A shared vision driven by leadership and an integrated real-time data stack is a key enabler of innovation. with real-time AI. Using real-time AI to grow communities around innovation can strengthen the whole rather than the parts. This is the only way AI can deliver tangible business results.
Learn how DataStax enables real-time AI.
About George Trujillo:
George is a Principal Data Strategist at DataStax. Previously, he has built high-performance teams for data value-driven initiatives at organizations such as Charles Schwab, Overstock, and VMware. George works with CDOs and data executives to continuously evolve real-time data strategies for enterprise data ecosystems.
