Longevity in the technology industry means you’ve no doubt participated in cyclic rites of passage triggered by major technology inflection points that promise to revitalize our industries and reinvent careers. We haven’t really seen one in a while that fundamentally changed our thinking about the art of the possible given the demands of the practical.
If you reflect for a moment, the last major technology inflection points were probably things like mobility, IoT, development operations and the cloud to name but a few. The shifts, however, began to get logically predictable and the flashes of energy, enthusiasm and investment activity always seemed to end up hinged on shallow promises and equally as predictable marketing.
This time however, it’s different. I state that with real conviction and genuinely authentic excitement. What is different about artificial intelligence (AI) — aside from the fact it that has completely absorbed our collective conscience and attention seemingly overnight — is how impactful it will be to efficient business operations and business value. The scope of its impact commands our attention, across all industries, as it forces us to look at business and technology in fundamentally different ways…while balancing what we invest in today with how we expect to conduct our business tomorrow.
As a software architect, I focus primarily on business and technology patterns and planning activities that are aligned to deliver desired business outcomes. My journey started by looking at the AI opportunity landscape in terms of business and technology maturity models, patterns, risk, reward and the path to business value.
Like many, my first real encounter with AI as a consumer was focused on generative AI (genAI) which seemed to take the world by storm almost overnight. In truth, however, it didn’t simply show up on the scene. It was part of a ‘slow moving coup’ that was incremental and foundational like most technology building blocks. The coup started with data at the heart of delivering business value. Let’s follow that journey from the ground up and look at positioning AI in the modern enterprise in manageable, prioritized chunks of capabilities and incremental investment.
Start with data as an AI foundation
Data quality is the first and most critical investment priority for any viable enterprise AI strategy. Data trust is simply not possible without data quality. It is the de facto foundation for reliability in machine learning, generative AI and agentic AI. None of what we do to achieve value from investments in data insights through AI is credible without quality data.
In a study by O’Reilly, 48% of businesses utilize machine learning, data analysis and AI tools to maintain data accuracy, underscoring the importance of solid data foundations for AI initiatives. A contrasting study by Qlik indicates that 21% of enterprises face real challenges with AI due to lack of trusted data for AI applications, highlighting the need for reliable data platforms.
I found that these kinds of challenges routinely emerge for companies that grow by acquisition with multiple, redundant instances of critical core systems and data. As a former enterprise architect for a large durable goods organization, I learned that the unwanted side effect of growth by merger and acquisition is a portfolio of fragmented data stores that reside everywhere and nowhere in OEM-specific ERP and supply chain management platforms.
As renowned technologist and entrepreneur Dave McCrory suggested back in 2010, data has gravity and where it lands in the business makes it a default source of attraction with assumed quality for that business irrespective of its actual accuracy. In this example, data stores were all too often redundant with gross inconsistencies in naming conventions, product nomenclatures and business semantics. The absence of known authoritative sources for something as fundamental as product data meant data fragmentation and data inaccuracies would be continually at odds with the quality of informed business decisions.
To remedy this, significant investments were made in data science and machine learning without a deeper understanding of the how to aggregate and abstract the data first in an aggregated platform. Cloud-based enterprise data platforms like Snowflake, Databricks, AWS Redshift or Azure Data Factory can expose an abstracted semantic model and consumption layer that is business-ready for analytics clients like Power BI and Tableau.
In retrospect, investing first in data science resources to develop machine learning algorithms and models is not only premature but also may amplify problems associated with data quality and data trust. A decision made with AI based on bad data is still the same bad decision without it.
Build on data platform foundations first to enable machine learning
Global spend on data platforms is expected to increase at a compound annual growth rate of 14.9% through 2030 and clearly, data quality and trust are driving that investment. If we revisit our durable goods industry example and consider prioritizing data quality through aggregation in a multi-tier architecture and cloud data platform first, we can achieve the prerequisite needed to build data quality and data trust first.
In parallel, building the organizational constructs around data quality also requires addressing data governance in parallel and supporting roles for data custodians, stewards and a centralized or federated data governance model to support credible and consistent enterprise data catalogs and products. With an enterprise data platform in place that supports a business-ready zone for consumption as well as appropriate governance, a critical machine learning readiness step is also in place. This can address data anomalies, cleansing and data lineage validation with a deeper level of sophistication in terms of algorithms and training-ready models.
By focusing on building the right data platform with the right OEM channel partners, enterprises can not only position their foundational investment in data for success but extend capabilities on the same platform to include best-of-breed, native machine learning features. Enterprises that elect to implement on the Snowflake data cloud, for example, might pursue native machine learning platform options to leverage the strength of the investment they have as opposed to the ones they don’t. Open-source implementations for machine learning invite obvious and hidden costs if your organization is not prepared to manage them.
Assuming the data platform roadmap aligns with required technical capabilities, this may help address downstream issues related to organic competencies versus bigger investments in acquiring competencies. The same would be true for a host of other similar cloud data platforms (Databricks, Azure Data Factory, AWS Redshift). In all cases, the foundational data platform roadmap matters in terms of near-term gains and long-term vision.
Is there a platform-based path to genAI that aligns well to an enterprise data platform (EDP) and use case opportunities? Ensure your foundational investments in a data platform have a viable path from machine learning to genAI and support use cases that will give your organization the right technology and investment runway over time. GenAI investments span a spectrum of use case opportunities that include but are not limited to human resources, procurement, finance, customer service, sales, marketing, corporate and supply chain. The business value and operational efficiencies for some use cases may be obvious but the financial barriers to entry for many in terms of short-term and long-haul investment are significant.
The market adoption rates favor low barrier-to-entry use cases and minimized risk (complexity, security and infrastructure) in terms of startup investment and time to implement from a platform perspective. Additionally, rapid return on effort through proof points is a must with readily observable business benefits. The consistent winner in market adoption and investment is genAI for customer experience. Consider:
- Approximately 15% of contact centers have integrated genAI capabilities to enhance customer authentication processes, streamline solution options and automate call summaries and follow-up recommendations.
- According to Reuters, Lyft achieved an 87% reduction in average customer service resolution time, effectively handling thousands of requests daily.
- According to CRN, approximately 53% of enterprises identify customer service chatbots as their top genAI priority. Use case runners-up include software development and code generation (e.g., Github Copilot), sales and marketing (Salesforce), R&D and finance (SAP).
When we combine investments in an EDP with a logical, product roadmap-supported path to AI that minimally includes machine learning and genAI, we can lay out a logical progression for investments and proof points that leverage the native capabilities of a data platform and have a clear path to support for use cases and capabilities that align to viable outcomes and business value.
Shiny objects and practical innovation
What about agentic AI? ‘Shiny object syndrome’ is something we all encounter and try to avoid but the allure of new technology knows no boundaries. For architects and software engineers alike, AI presents no shortage of opportunities to be driven to distraction by the art of the possible.
Yet there are some emerging technologies in the AI arena that are major contenders for the kind of disruption that can help enterprises extract immediate efficiencies. We can choose to ignore emerging technologies like agentic AI in terms of investment, but current market momentum and innovation remind us ignorance is not bliss.
Agentic AI essentially implements software agents that can act on the behalf of an application, platform or end user to reason and iteratively plan to solve complex, multi-step problems via task execution and learning. These autonomous or semi-autonomous agents can even operate in an ecosystem of agents in what is referred to as an ‘agentic mesh’. The process implications are staggering when we consider what this means in terms of operational efficiencies for core systems that include supply chain management, ERP, HCM, finance, sales and marketing.
Consider a simple use case example like email marketing where an agent can devise a plan that executes tasks across enterprise systems to access structured and unstructured data, transactional systems, APIs and document management systems. The agent can break down tasks needed to acquire business information context, interact with a large language model (LLM) to reason and plan and then coordinate and execute the tasks.
For example, you could request the agent to generate a tailored email campaign to achieve sales of $100,000 USD per month. Inputs to the tasks could be the location of products and performance metrics and a CRM system for customer contact information. The agent’s ability to understand and decompose the tasks, interact with systems, ‘remember’ actions taken and then take more optimal, alternative actions based on what is learned is compelling and powerful.
This is especially true when you consider complex workflow use cases where an agent could interact with other agents in other ecosystems to manage and learn more broadly and deeply over time and execute at scale. Despite only gaining real traction in 2024, Deloitte predicts that by 2025, 25% of companies employing GenAI will initiate agentic AI pilot programs, or proofs of concept with this figure expected to rise by 50% by 2027. Further, UIPath reveals that 77% of IT executives are prepared to invest in agentic AI within the current year while 37% of IT executives surveyed report that their organizations are already using agentic AI solutions.
A compelling and inspiring case study for practical innovation in the enterprise
Outcome-driven investments that deliver business value in complex business environments are rare with new technologies. Georgia-Pacific, a leading manufacturer and distributor of pulp and paper products, has been at the forefront of integrating artificial intelligence into its operations to enhance efficiency, safety and decision-making processes. Georgia-Pacific (GP) is a shining example of risk and reward for enterprises that want to adopt AI to tackle complex critical systems like ERP and complex functions like order management.
Under the leadership of Michael Carrol, VP of transformation and Rob Norris, director of innovation, the team “created a more seamless order management process that helped navigate the complexities of a myriad of individual orders received every minute, hour and day with unparalleled precision”. Carroll’s thought leadership in the concept of causal AI enabled near real-time pattern and anomaly detection and proactively managed order errors and discrepancies using cause-and-effect AI solutions to intervene when a system “can’t perform an autonomous decision and provide a recommendation based on causal analysis patterns”.
This semi-autonomous, human-in-the-loop solution preserved both human and machine-mediated workflows to achieve high business value outcomes in intensely complex business processes. Their work and accomplishments were featured in an article by Fortune Magazine entitled “What Georgia Pacific is Doing with Causal AI is Remarkable” in April of 2024 and it built a foundation for success. What differentiated the work? Incremental investment, a deep understanding of the business process, systems and data stores and a tempered balance between human and machine-mediated problem-solving. That balance between fully autonomous and semi-autonomous and groundbreaking work in causal AI patterns will forge a path to innovation and broader adoption.
Key considerations in managing the balancing act between risk and reward
Early investment in data foundations is critical to the continued success of downstream investment in AI at an enterprise level. Focus on enabling enterprise data platforms that prioritize data quality first to establish trustworthy data products. If sensible investments are made in data platform enablement with an OEM roadmap that has a solid track record of AI feature set introduction, IT professionals can reduce redundancy and technical debt significantly by making deliberate and purposeful choices.
Additionally, the human investment in data governance early through data stewards and data product owners under a centralized or federated model can ensure consistency and semantic context for all business domains that need trusted data to make business-critical decisions. Absent quality data, AI will only replicate quality issues broadly in the enterprise and introduce widespread risk in decision quality.
Learn from the Georgia Pacific playbook and tackle machine learning with incremental investment and steps to manage data nuances through analysis and anomaly detection that take data quality a level up. Leverage the native capabilities of OEM-based enterprise data platforms when it makes sense to exploit native machine learning capabilities with a path to greater sophistication over time.
GenAI has a broad base of enterprise use cases. If you are contemplating modest investments that advance the use of the technology with observable return on effort, consider customer experience as a common point of entry with lots of case studies and business benefit data from large enterprises that have already taken the plunge. Lowest common denominator solutions for ChatGPT interfaces to data lake houses and warehouses may provide an even lower barrier to entry proof points that could complement an investment in an existing enterprise data platform.
Agentic AI is here to stay and will gain tremendous momentum in 2024. It is also poised to make even greater advances in 2025 through mega-vendors like Salesforce’s Agent Force and SAP’s agentic Joule implementation. Both focus on semi and fully autonomous intelligent workflow automation and natively support some of the more complex business processes we encounter as IT professionals. The advantage of templated agents for Salesforce CRM, Marketing Cloud & Commerce Cloud is clear and if you are an SAP customer, having a fully supported pathway to agentic AI for your ERP implementation could provide a running start to what is certain to be the game-changer of 2025.
Smart investments are the most important balancing act we must support between near-term gains, quick wins and long-term vision for the AI-enabled enterprise. Tempering that balance between opportunistic near-term investment in the right technologies and platforms that provide a solid foundation for growth and business value is the most business-critical decision of our time. There are inflection points and then there are tectonic shifts. Like all IT inflection points, however, foundational and incremental investments with clear and purposeful intent will pave the way to discovering a longer path to greater business value over time as the AI story unfolds.
Dion Eusepi is a technology industry veteran focused on practical innovation in the architectural design, development and delivery of enterprise data and AI-ML platforms and intelligent ecosystem solutions for hybrid cloud environments, multi-tier data pipeline aggregation architectures and infrastructure, for on-premises, cloud and edge compute environments. Dion has had the privilege of contributing to multi-industry Fortune 100 and 500 companies including Ford Motor Company, General Motors, Stanley Black & Decker, IBM and Salesforce. His work includes comprehensive platform solutions for cloud, data, integration and AI-led enablement strategy and spans core ERP, CRM and HCM systems, SaaS and digital channel integration, ML ops, IIOT and I4.0 edge compute data distribution that connect broad, deep PLM eco-systems.
This article was made possible by our partnership with the IASA Chief Architect Forum. The CAF’s purpose is to test, challenge and support the art and science of Business Technology Architecture and its evolution over time as well as grow the influence and leadership of chief architects both inside and outside the profession. The CAF is a leadership community of the IASA, the leading non-profit professional association for business technology architects.
Read More from This Article: Prioritizing AI investments: Balancing short-term gains with long-term vision
Source: News