Agentic AI & Process: Better Together

One of the best places to implement AI practically and successfully is in external or internal processes, including front and back-office processes used in every day and game-changing modes. Processes are often the basis of organizational actions that cross internal and external boundaries. These processes often employ resources that could benefit from AI's automation or assistance, especially where knowledge, decision-making, and agile optimization based on changing or emerging goals are required. Examples include staying in touch with customer sentiment during interactions with your organization, the state of your supply chain, prioritizing limited supplies, or internal processes that must be tuned to stay operationally optimized while responding to change. Typically, organizations start with automation and then human assistance. AI-enabled processes can also take intelligent and agile action and are often the first place unusual events and patterns are sensed, requiring tactical adjustments and pointing to potential strategy changes. Processes are a great place to start with AI, whether in mining current or past results, sensing behavior shifts that point to new opportunities or threats, making better decisions, or assisting scarce resources while optimizing outcomes.

AI can safely start inside the context of a process while sitting outside the process, watching and guiding results in either static or emergent processes/cases. Organizations must consider how best to leverage AI with processes inside or outside a process. Traditionally, processes are flow-directed, and AI services the tasks or resources of the process. Still, there are significant benefits to leveraging AI in a goal-directed process where AI responds to sensed change outside a process, managing the process's shape, sequence, and resources for organizational outcomes within governance constraints.

Three Process Types: Headed to Hybrid

Understanding three types of processes is essential before entering the experimentation phase or deciding to institutionalize and process activity. See Figure 1 for the typical process types used today that will likely be used in a hybrid fashion in the future.

                                Figure 1: Process Types Used Today

It is important to know what each process type brings to the party as fixed processes evolve to goal-driven and event-sensitive processes. Below, I created a quick and dirty comparison of the types of processes and their attributes. As organizations journey into the AI Agentic world, leveraging the strengths of other process types will come in handy while allowing the leverage of legacy processes and computing assets.

 AI & Process Evolution

AI can quickly start with actions and migrate to and then move to add recognition and decision tasks in emergent processes with feedback loops as fast as real-time. Putting control outside the process/process snippets will lead to Agentic AI. See Figure 2 for the journey to Agentic AI.

                              Figure 2: Journey to Agentic AI

While AI promises and has delivered significant benefits, particularly in automation and people assistance, it is vital to manage the risks involved with any new and emergent technology. Some proven approaches allow organizations to experiment and implement new technologies while gleaning solid benefits that can be optimized over time. The two major phases are experimentation and institutionalizing. Experimentation aims to manage risk while gleaning benefits, growing institutional learning, establishing skill sets, and creating additional pathways to benefits. Institutionalization seeks to expand an organization's skill base, extend better/best practices, and create an organizational strength to leverage in the battle for AI advantage. Savvy organizations will repeat this experimentation/institutionalization cycle continuously.

Agentic AI introduces another area for experimentation that combines the strengths of agents with the strengths of processes. Over time, some or all the control of processes will be relinquished to outside the processes to Agentic AI. See Figure 3 for the four vectors of agents' abilities growing in influence in and around processes.

Trying Agentic AI

                      Figure 3: Agentic AI Ability Vectors



Institutionalizing Agentic AI After Initial Successes

                         Figure 4: Agentic AI in Context.

 
After the experimental phase of operationalizing AI, organizations must invest additional efforts to take advantage of AI while fully institutionalizing AI incrementally. Depending on the use cases chosen, this investment could include capital expenditures for software and services and budgeted internal efforts—understanding where selected use cases play in AI's overall maturity and direction. Refer to Figure 4 Agentic AI in Context. Savvy organizations will not only focus on use cases that provide benefits in the short term but also consider the use cases in an overall AI architecture and direction that matches AI maturity. This writing encourages using AI within business processes or customer journeys. Still, AI can be used in isolation of processes as bots/agents acting independently as solo automation or monitoring. The sweet spot of AI today is at the business level in internal or external processes, but AI can be used in the technical infrastructure or the B2B levels.

Net; Net:

As you can infer from the detail in this paper, processes are essential for safely operationalizing AI initially while incrementally growing the benefits and effects of AI now and in the foreseeable future. I suggest starting with process-driven AI to augment the traditional data-driven AI that has been successful. Searching data for opportunities is good, but goal-driven processes that link to business goals ensure a more direct contribution to the bottom line and the goals of an organization while staying congruent with its values. While data is critical, a dynamic process powered by AI is the closest to executive success. The emphasis on goal-directed Agentic AI approaches coupled with governance boundaries (constraints) is one of the most important ways to keep AI on point. The importance of goal management will increase over time. See my writing on goal cycles here: 

Since processes are more closely connected to goals and point to the required data sources to attain current and emerging goals, pure data-driven approaches are likely to drive costs high as all data could be significant with little goal guidance.