Agentic Enterprise and AI

From: aidotengineer

Presented by Hbert Mishtella of Novartis, this article explores the concept of the Agentic Enterprise, focusing on how AI agents can impact organizations and what leadership needs to know to prepare [00:00:15]. The discussion covers workflows, organizational networks, challenges, and individual employee perspectives on the impact and future potential of AI and agents [00:00:37].

The Rise of AI Agents in the Enterprise

It is a provocative, yet realistic, statement that in three years, organizations might be run by AI agents given the current pace of AI development [00:00:53].

What is an AI Agent?

An AI agent is defined as a Large Language Model (LLM) based application that exhibits a certain degree of autonomy in its behavior [00:01:11]. Its core function involves using a language model to “glue together” various specific tasks [00:01:22]. AI agents execute tasks through a step-by-step planning process, utilizing different tools [00:01:29]. The use of these tools can be dynamically adjusted based on the outcome of previous steps [00:01:37]. This capability introduces a new level of functionality that was previously almost unique to humans [00:01:47].

Key Capabilities of AI Agents

Organizations should focus on evolving their digital assets for the effective use of AI agents [00:02:01], aiming for every digital asset to become either a tool for an AI agent or an agent itself [00:02:14].

The key capabilities that make AI agents transformative include:

• Reasoning for multi-step planning: Enabling complex task execution [00:02:26].
• Adaptability for dynamic work: Adjusting to changing conditions [00:02:28].
• Persistent memory: Retaining context over time [00:02:32].
• Contextual knowledge (RAG): Utilizing Retrieval Augmented Generation (RAG) to access specific external knowledge not available in the language model [00:02:35].
• Interactive workspaces: Collaborating iteratively with an AI agent in sandboxes or canvases to refine tasks like coding or art [00:02:49].
• Computer-using agents: Navigating graphical user interfaces (GUIs) to automate steps previously done only by humans or through Robotic Process Automation (RPA) [00:03:05].

Transforming Workflows with AI Agents

Workflows are the actual execution of processes within a company [00:03:32]. Technology can be applied in two ways:

• Automation: Substituting a specific step with technology [00:04:19].
• Augmentation: Performing a step quicker or better [00:04:26].

While machine learning (ML) could previously automate or augment some tasks [00:04:31], the difference today lies in agentic workflows [00:04:45].

Agentic Workflows

Agentic workflows are possible now for two main reasons:

1. AI agents can tackle a broader set of tasks, operating on cognitive steps that are precisely represented in natural language [00:05:00].
2. Language models and AI agents can act as a “glue” between tasks, composing them together without human intervention [00:05:20]. Human feedback is only needed when the agent requires it [00:05:32].

This allows for the automation and delegation of not just one, but multiple steps in a process [00:05:43].

First Takeaway for Leadership

To start building agents, organizations must first identify their workflows [00:05:57]. Crucially, they need to understand the exact context around these workflows [00:06:04]. This context, including data, systems, and transformation steps, is often unrecorded and resides only in employees’ heads [00:06:17]. This “buried” knowledge, distributed across teams and systems, presents a significant challenge that companies must overcome to effectively leverage AI agents [00:06:54].

Organizational Perspective: The Network of Personas

Relying solely on traditional roles and job titles is insufficient for applying AI agents [00:07:51]. To effectively implement an AI agent, one needs to understand the specific tools to be used, the planning required, and the conditions for changing behavior [00:08:00].

A more helpful approach is to consider a network perspective, defining different archetypes or personas for each contributor to a workflow [00:08:36]. Examples of personas include:

• Silent Achiever: High performance, lower communication [00:08:57].
• Individual Contributor: Strong in both performance and communication [00:09:04].
• Connector: Links non-adjacent teams within the organization [00:09:09].
• Multiplier: Enhances the work of other team members [00:09:16].
• Knowledge Hub: Provides domain expertise or necessary information [00:09:21].

Defining these personas allows for a better understanding of how AI agents might impact workflows and specific employees [00:09:39]. This detailed view is necessary for a global organizational understanding, identifying similarities and differences across teams, and optimizing for the development of agents [00:10:11].

Second Takeaway for Leadership

To help organizations and people adapt to AI, looking at employee personas is a helpful first step [00:10:44].

Impact on Roles and Emerging Patterns

The introduction of AI agents will change workflows dramatically. For example, a “silent achiever” with a coding assistant might become productive enough to handle work previously done by multiple people in several steps [00:11:18]. An agent can magnify the impact of a “multiplier” persona, extending their influence beyond adjacent team members to the entire team, leading to higher performance and better communication [00:11:56]. A “knowledge hub” persona is highly prone to substitution by an AI agent, which can dynamically support all team members [00:12:36].

The key is to understand the archetypes or personas of employees across different workflows to project and plan for the development of various agents [00:13:29]. Identifying repeating patterns in workflows (e.g., issues always on the first step) indicates what kind of agents need to be built for optimization [00:13:55].

Emerging Patterns

• Human vs. AI Tasks: Tasks where ambiguity is high, or subjectivity is desired, are better for humans [00:14:42]. Repetitive, tedious tasks where perfect, consistent execution is needed are ideal for agents [00:15:10].
• Intelligence and Domain Knowledge as Commodity: These attributes will become cheap and easily available, no longer providing a competitive edge for employees or companies [00:15:32].
• Shift in Employee Focus: Employees and companies will need to pivot towards multi-disciplinary know-how or much deeper specialization [00:16:06]. Senior engineers might work directly with product and domain experts, developing quicker with assistants [00:16:24].
• Agents as Super Connectors: Agents can enhance communication and impact across teams [00:16:43].
• Increased Value of Context Understanding: This will gain more value compared to intelligence or domain expertise [00:16:48].
• AI Supercharges “Doers”: People who perform tasks can do even more and quicker, potentially leading to a “100 times engineer” [00:17:04].
• Polarization and New Specializations: A polarization in positions and tasks is expected, leading to very new specializations [00:17:27].

New Roles in AI-Mature Organizations

New roles might appear, such as:

• Workflow Minor [00:17:48]
• Human AI Orchestrator [00:17:55]

Enabling AI Adoption: A Phased Approach

For executives, democratizing AI adoption across the organization is crucial [00:18:15]. This typically follows a step-by-step progression:

1. General Assistants: Provide access to general AI assistants (likely already in place) [00:18:32].
2. Individual Assistants: Offer personalized assistants per employee with memory, context understanding, and personal databases [00:18:39]. This allows employees to work more closely with their specific context without repetition [00:18:48].
3. Employee-Built Agents: Enable employees to quickly build their own agents for optimized workflows using coding assistants, no-code, and low-code tools [00:19:10]. This requires employees to develop “cognitive self-awareness” – understanding the specific mental and system steps they perform to achieve a goal [00:19:45].
4. Digital Twins: Implement agents as “digital twins” or “time travelers” to retain organizational knowledge [00:20:24]. This means an agent can represent an employee’s knowledge and expertise, allowing for reconstruction of past processes and answering questions even after an employee leaves [00:20:42].
5. Multi-Employee Agents: Develop agents that serve not just one employee but entire teams or workflows [00:21:06].
6. Swarm of Agents: Progress to a swarm of agents, where software is already available [00:21:12].

This progression is enabled by the maturing software stack around LLMs and AI agents [00:21:20].

Challenges and Opportunities

Challenges

• Adoption: A significant cultural question is whether organizations are ready for AI agents to not just receive delegated tasks, but to drive interactions and even delegate subtasks to humans [00:22:25].
• Ethics of Autonomy:
  • Defining explicit values for AI agents to uphold as “value keepers” is a primary challenge, as humans themselves don’t always agree on values [00:23:25].
  • Resolving conflicts or interactions between humans and AI agents presents a new, interesting challenge [00:23:54].
  • As machines gain more freedom, they will require increasingly defined moral standards [00:24:16].

Opportunities

• Simulation: Simulating synthetic reality before real-life activities in marketing, labs, or other areas [00:24:53].
• Proactive Insights: Proactively understanding regulatory loopholes or identifying mergers and acquisitions (M&A) opportunities [00:25:06].

Executive Summary: Key Takeaways for Leadership

• Outdated Strategy: If your organizational strategy still makes sense when “AI agent” is swapped with “machine learning” or “data”, it’s likely outdated [00:26:31].
• New Capabilities: Agents offer distinct new capabilities [00:26:56].
• Workflow Understanding: Detailed workflow understanding is critical to derive significant benefits from agents [00:27:00].
• Democratization: Democratizing agents is essential to fully accelerate the AI agent revolution [00:27:05].
• Human Aspect: Looking at the network of personas (the human aspect) is highly important [00:27:13].
• Rapid Transformation: Organizations must be ready for rapid transformation, as entire org charts might change drastically [00:27:19].
• Mindset Shift: A mindset shift is vital for leadership and organizational strategies for AI integration, redefining interaction with technology that might, at times, lead humans [00:27:31].
• Ethics: Ethics remains a crucial part of AI implementation and societal impact [00:27:47].