Putting AI to Work: Systems of Intelligence and Actionable Agency
As traditional and generative AI rapidly become pervasive in the corporations, the role of AI is changing from that of a single-purpose application to the center of a new way for people to work, collaborate and create. Leaders in every sector of society, including business, education, and government, should be thinking about the changes that are happening now, and how best to position their organizations for the future.
The good news is, it is happening. I know, because I see it. In my work at Google Cloud’s Office of the Chief Technology Officer, or OCTO, I meet weekly with senior leaders in various industries including digital natives, finance, retail, and healthcare.
My OCTO colleagues are in similar conversations with leaders in other industries and organizations. In the past year, both the uptake of AI and the interest in adapting to new opportunities have accelerated, and now it’s being put into action.
Pervasive AI will create a new System of Intelligence (SoI) that integrates data, technologies, platforms, and practices for the purposes of finding and understanding patterns, extracting insights, promoting efficiency and creativity, and facilitating decision-making. This will illuminate how organizations actually do on a functional basis, through real-time data inputs, allowing people greater awareness and meaningful action.
Here is why: The system of intelligence is designed to work in a way that is different from traditional data systems or systems of record. Rather than requiring users to know how to extract insights from the data, the system of intelligence is designed to identify, ask questions and provide insights in a way that is easy for all users to understand.
Standards and practices of the SoI are still emerging, which gives leaders the rare chance to both learn from and guide the development of a new system of work in the coming year. This is necessary work since imagining that nothing will change with AI is akin to thinking that, at the dawn of television, radio would simply be transposed wholesale, with no particular effect on culture, process, or business models.
It’s helpful to begin this process by considering a few basics and learning from early use cases.
The Core SoI Dynamic
The SoI differs from precursor systems in many ways. Critically, humans and computers interact with each other in natural language, so that people trained in other skills can access large-scale computing with no intermediary, and systems can learn from many types of roles. Add to this a rich and growing set of Large Language Models for training gen AI in various use cases and outputs; databases customized to handle the new workflows; editing tools and integrated development environments, and other aspects; businesses are seeing a vast broadening of the user base, the interactions, and the applications.
SoIs aim to make data actionable by translating it into intelligence, learning, and action. This may come as suggestions to a human operator, or autonomous actions on a trusted basis (see “SOI as a trusted agent,” diagram). This is an extension of the system’s ability to generate novel content, in the form of text, image, video, music, or software code, which has captivated the world’s attention.
This increased development will lead to more linkages between projects, and thus greater observation, insight, and action. These elements of interaction and creation are different from precursor systems of IT in the workplace: Systems of Record and Systems of Engagement, which are more focused on data storage and interaction for internal and external needs. SoIs, on the other hand, are the most future-facing, as they will create actionable insights in any direction from data.
Organizing for a SoI
Sols work primarily through language and learning models, or LLMs, which employ deep learning techniques on large data sets to “understand” structured and unstructured data, visual information, and software code. This understanding is based on extremely complex statistical relationships, and should not be conceived of as an intelligence independent from human operators. It is refined, or “grounded,” for greater accuracy by adding quality data related to a specific domain.
Additional refinement is provided to the grounded LLM through Retrieval Augmented Generation, or RAG, that sources and amalgamates data from various inputs, such as, databases, devices, user interfaces, and external services. In many cases, this will be data unique to the corporation, government, or other operating entity. The SoI can access and process information from other external knowledge sources, which can help to improve the accuracy, timeliness, and coherence of the generated output. Near real-time analysis is thus possible over a range of use cases.
Data selection and data quality are correlated to the desired work output. Patterns and actions are usually determined by human-generated prompts, also influencing results. Learning may be incorporated to improve the SOI’s understanding and predictions. The overall workflow is thus natural-seeming and structured.
The SoI as a Trusted Agent
Autonomous actions by an SoI are closely governed to minimize potentially negative outcomes. For example, in a contact center, the system initially handles customer inquiries, from simple questions about billing to complex technical issues, leaving human operators to handle more challenging, high-value customer problems, ultimately improving efficiency, agent satisfaction, and customer experience.
A future use case could be a more sophisticated travel planning agent that helps plan and book trips: researching destinations and activities based on the traveler’s interests and budget, personalizing itineraries, and booking accommodations. It could also provide real-time updates like flight delays or weather conditions, or troubleshooting problems.
In a critical area like healthcare, a gen AI-powered hospital care agent could flag potential health risks and complications, suggest personalized treatments, monitor progress, and coordinate care among providers. The agent could also send the patient reminders about medication via text message or offer healthy meal plans.
At present, AI-powered care agents are expected to work in conjunction with humans/ existing workflows will need to change, placing a greater emphasis on accurate data, increased room for human autonomy and creative work, automation of rote processes, and more customer feedback in evaluating the work process. Over time, system performance and reliability will be reviewed, as with other medical innovations.
Implications of the SoI: Intelligence
The SoI is designed to improve over time, gaining in accuracy and capability. There may be more autonomous and semi-autonomous actions by trusted agents, much the way modern airlines are now largely flown by computerized systems, and vehicles have more augmented and automated features on their path to more autonomous, safe and predictive driving.
Critically, the SoI’s capability to learn means that, with additional data, more sophisticated models, greater computation, and experience in different contexts, the SoI will “understand” more corporate functions, blending related ones and acting across traditional corporate categories.
This does not take humans out of the loop, but instead frees them to observe and influence their worlds with greater understanding and potential impact. Just as gen AI can now be used to synthesize and allow queries of long documents, the SoI may in the future observe and explain the many interactions of, say, a large multinational, a healthcare system, or other large, complex systems.
In a narrow sense, this might be seen as a kind of generalized intelligence — not, as many science fiction entertainments posit, an autonomous and ungoverned being, but rather complex software that expresses currently siloed corporate functions in a more holistic way, creating new insights and efficiencies. And, rather than displacing humans, it may well challenge them to new times of innovation, organization, and creativity.
Regardless, the SoI is a promising approach to developing corporate artificial general intelligence (AGI), as they have the resources, knowledge and investments in systems that can be leveraged to develop and deploy SOIs. Building and adopting this emerging system will provide them with a competitive advantage through unprecedented productivity and efficiency gains by freeing up human resources.
Every leader today should be thinking about taking advantage of the SoI, the basis for trusted and scalable insights that will deliver disruption for their business and entire industry, if implemented early and effectively.