AI - and generative AI - is becoming a larger focus for global organizations and Enterprise Architecture is emerging as one of the critical success factors in running AI at scale.
The Search for Value
It has been easy to get AI pilots up and running and there are a thousand things we can do with generative AI but there is a balance between AI projects we can do and the ones we should do. One of the largest questions on the table in AI and generative AI is where to invest and what are the considerations that drive success. Three key areas of value have emerged: individual productivity, group automation, and the creation of new products and services.
Unlocking Individual Productivity
At the heart of generative AI's value proposition is its ability to augment human capabilities. "Every little task you do 20% faster. Soon it will be part of the standard kit." This increased productivity can lead to a shift in organizational belief systems, as enterprises recognize the potential to achieve more with the same resources.
For example, in areas like document analysis and code assistance, generative AI can streamline processes and eliminate unnecessary steps. "Leverage AI to enhance human decision-making rather than replacing humans entirely." By doing so, enterprises can create a more efficient and effective workforce.
However, it's important to remember that the individual productivity benefits of AI may not be immediately apparent. "It took 15 years for the productivity effect of the PC to show up in the national accounts." – yet if companies had not invested in the PC they would be left behind today. This highlights the need for organizations to act now even if the short-term productivity benefits cannot be measured easily when implementing AI initiatives.
Driving Group Automation
Beyond individual productivity, generative AI can also drive significant benefits at the group level. "This is where $ and cents really add up. You can take 25-40-70% cost out." This can have a direct impact on the bottom line, as demonstrated by companies that have successfully automated processes like revenue operations.
To realize these benefits, enterprises should focus on mapping out their end-to-end processes and identifying opportunities for automation. "Working with third-party providers can accelerate the process" as they can deliver the automation as part of their ongoing services to enterprises.
While cost savings are important, it's also essential to consider other metrics such as error rates, customer satisfaction, and time to market. By measuring these factors in addition to dollars saved, enterprises can get a more complete picture of the value they are deriving from Aia t the group automation level.
Creating New Products and Services
Perhaps the most exciting potential of generative AI lies in its ability to drive innovation. "By creating new products, experiences, and services, enterprises can tap into new markets and generate additional revenue."
One example from the retail industry is the idea of providing personalized styling services to every customer. "This is a new product that could significantly enhance the customer experience and drive top-line growth."
To maximize the value of generative AI in this area, enterprises should consider metrics beyond cost savings. "Measuring the impact on customer experience, revenue, and time to market can provide valuable insights into the success of AI initiatives."
It's also important to start small and iterate to assess the feasibility and benefits of AI initiatives. While AI may have upfront costs, the long-term benefits can be significant. By focusing on both defensive and offensive strategies, enterprises can identify opportunities to improve margins, bring in new revenue, and stay ahead of the competition.
Organizing Data for AI
As generative AI continues to transform industries, getting a data foundation is essential - and yet data continues to be last unsolved problem in the enterprise - requiring a special focus on data preparation, including data cleanup, architecture changes, and acquisition of new data, to ensure AI initiatives are fueled by high-quality information.
Ensuring Data Quality and Governance
The foundation of any successful AI initiative is high-quality data. "The first thing that needs to be done is to agree on common definitions of data." By establishing data quality standards and implementing governance mechanisms, enterprises can ensure that their data is accurate, consistent, and reliable.
"Break down data silos and create a unified data platform." This involves modernizing data infrastructure and breaking down barriers between departments. Additionally, leveraging AI tools can help identify and correct data errors, further improving data quality.
Defining Data Accessibility and Operating Models
To unlock the full potential of AI, data must be readily accessible to users. "Provide easy access to data through self-service tools and APIs." By integrating data from various sources, enterprises can gain a comprehensive view of their operations.
"Implement robust data governance and security measures." This includes considering factors like where data engineering should be located and whether a centralized or federated model is best suited for the organization.
Driving Data-Driven Decision Making
"Embed data insights into workflows and integrate AI-powered analytics into decision-making processes." By using AI for predictive analytics, enterprises can forecast future trends and make informed decisions.
"Track the results of data-driven initiatives to assess their effectiveness." By focusing on key questions that are asked at the board meetings, such as "what moves the needle" and "where to invest" – enterprise technology leaders can ensure that their AI efforts are aligned with their strategic goals.
Architecting for AI
The success of AI initiatives hinges on effective governance and organizational structure. As AI models become increasingly sophisticated, questions arise about ownership, scaling ML expertise, and ensuring ethical and responsible AI development. The challenge is to create an environment that fosters innovation while mitigating risks and ensuring accountability. The ability to evaluate and monitor AI performance is key. Organizations must develop robust metrics and frameworks to assess the accuracy, security, and ROI of their AI solutions. This is particularly critical in the context of LLMs, where the complexity and potential biases of these models require careful oversight.
Selecting and Integrating the Right Technology
Choosing the right AI models is crucial for success. "Evaluate different AI models based on their capabilities, cost, and ease of use". Integrating AI into existing systems requires a strategic approach, considering factors like scalability and flexibility. Specifically, Model Accuracy is an evolving science and picking the right language model for the long term is almost a non-starter - but equally implementing a model-switchable backend to accommodate the latest models is also hard.
"The process of assembling your own stack is way harder". It often involves a combination of factors such as building a foundation for the AI platform. While cost is important, other considerations like latency and performance can also impact user experience.
Staying up-to-date with the latest AI models is essential. "We don't want to waste time looking for the latest models," and yet as model performance increases and cost decreases rapidly, staying on top of the newly developed models remains critical. By carefully monitoring the market and validating model performance, enterprises can make informed decisions about which models to use.
Scaling and Funding AI Initiatives
Scaling AI initiatives can be challenging, especially in large organizations. "Start small and iterate," - by gradually expanding the scope of AI initiatives, enterprises can mitigate risks and ensure that they are effectively leveraging AI.
"Measure and learn from results" - Continuously monitoring the performance of AI initiatives and making adjustments as needed is essential for success. Exploring partnerships and outsourcing can also help enterprises leverage expertise and potentially reduce costs.
Funding AI initiatives is another important consideration. "We decided to earmark a certain percentage of our IT budget" to generative AI instead of asking for an increase of budget. By allocating a specific portion of the budget, enterprises can ensure that AI remains a priority.
Driving Human-AI Collaboration
To fully realize the potential of AI, it is crucial to "design AI systems that augment human capabilities." By providing clear explanations and transparency, organizations can build trust and ensure that AI is used ethically and responsibly. Addressing ethical concerns and biases is essential to prevent unintended consequences. As organizations manage the transition to AI, it is vital to provide adequate training and support to employees. The ultimate goal is to create "autonomous lines of business," where AI handles routine tasks and frees up humans to focus on higher-level work. Imagine a future where AI powers contact centers, bank reconciliation, tax accounting, and HR helpdesks. This is the vision that technology leaders should strive for, where AI acts as a powerful tool to enhance human capabilities and drive innovation.
Executive Technology Board (c) | North America & Europe