2025 is not the year of the “Agentic Agent.” It is the year of the human-versus-tech-led Agent-based development and implementation model.
In other words, after more than 40 years in high-tech and procurement, the industry’s discussion of agentic AI and digital labor is the same as its discussion of ERP, Service-Oriented Architecture, and SaaS—all of which resulted in a high generational initiative failure rate estimated to be as high as 80%.
As you watch the Salesforce CEO video, I want you to keep that in mind.
CONTROLLED LEARNING
Agentic AI’s ability to learn autonomously and adaptively makes it powerful for solving complex problems. However, ensuring safe, ethical, and effective learning requires robust design and oversight.
How Learning is Controlled
Iterative Testing: Continuous monitoring and updates refine the AI’s capabilities while mitigating risks.
Human Oversight: Developers and domain experts guide the AI’s learning process, especially during the early stages.
Constraints and Safeguards: Rules and ethical frameworks are embedded to limit undesirable behaviors.
CO-PILOT VERSUS AGENTIC AI
Microsoft Co-Pilot is an advanced assistive AI that enhances productivity by leveraging natural language processing and machine learning within predefined contexts. It does not qualify as agentic AI, as it lacks autonomy, goal-setting abilities, and proactive decision-making. It is a tool designed to augment human capabilities, not replace or operate independently of them.
Microsoft Co-Pilot is not an agentic AI. It is better classified as a tool-based AI or an assistive AI. Here’s why:
What Is Agentic AI?
Agentic AI refers to systems that can:
Operate autonomously without direct human intervention for extended tasks.
Pursue goals based on its programming or learned behavior.
Make independent decisions within a defined scope of objectives.
Agentic AI systems are designed to act more like “agents” in a system, proactively performing tasks, adapting to changing conditions, and sometimes even setting sub-goals to achieve broader objectives.
Why Co-Pilot Is Not Agentic AI
Assistive, Not Autonomous:
Co-Pilot is built to assist users in completing specific tasks within Microsoft applications, such as summarizing emails, generating Excel formulas, or drafting text in Word.
It requires user input or explicit instructions to function (e.g., a user asks Co-Pilot to draft a report or create a slide).
No Goal-Setting Capability:
Co-Pilot cannot independently set goals or pursue objectives. It executes tasks as per user commands and does not exhibit initiative or self-driven actions.
Limited Scope of Actions:
Its functionality is tightly integrated into Microsoft tools and focuses on enhancing productivity and efficiency rather than making complex decisions or adapting autonomously to new environments.
Human-in-the-Loop Design:
Co-Pilot is designed with a human-in-the-loop approach, meaning the user retains ultimate control over actions, decisions, and outcomes.
Comparison: Co-Pilot vs. Agentic AI
Feature
Microsoft Co-Pilot
Agentic AI
Autonomy
No
Yes
Goal-Setting Ability
No
Yes
Adaptability to New Contexts
Limited to predefined tools
High (in defined scope)
User Dependency
High (human commands needed)
Low
Proactivity
Minimal
High
MY TAKEAWAY
It all starts with using a Tech-Led equation-based development and implementation model that reduces technology to a largely irrelevant shiny bauble rather than a valuable adjunct to an agent-based development and implementation model.
Technology doesn’t solve problems; people do, who then use technology to deliver tangible results: https://youtu.be/49BS-MkGoak
To be successful with any technology, AI or otherwise, the Agent-based model requires effective communication and collaboration between all stakeholders (Agents), internal and external to the enterprise. In short, you need a proper relational governance model rather than a “wow, new tech, plug-and-play strategy.”
The above image for this post was generated by AI – does the picture match the post content?
Procurement Insights
What Are Salesforce’s CEO Marc Benioff’s “Digital Labor” Boundaries?
Posted on January 15, 2025
0
I will also tag Matthew Berman in the videos above (Satya Nadella) and below, where the Salesforce CEO discusses “digital labor.”
2025 is not the year of the “Agentic Agent.” It is the year of the human-versus-tech-led Agent-based development and implementation model.
In other words, after more than 40 years in high-tech and procurement, the industry’s discussion of agentic AI and digital labor is the same as its discussion of ERP, Service-Oriented Architecture, and SaaS—all of which resulted in a high generational initiative failure rate estimated to be as high as 80%.
As you watch the Salesforce CEO video, I want you to keep that in mind.
CONTROLLED LEARNING
Agentic AI’s ability to learn autonomously and adaptively makes it powerful for solving complex problems. However, ensuring safe, ethical, and effective learning requires robust design and oversight.
How Learning is Controlled
Iterative Testing: Continuous monitoring and updates refine the AI’s capabilities while mitigating risks.
Human Oversight: Developers and domain experts guide the AI’s learning process, especially during the early stages.
Constraints and Safeguards: Rules and ethical frameworks are embedded to limit undesirable behaviors.
CO-PILOT VERSUS AGENTIC AI
Microsoft Co-Pilot is an advanced assistive AI that enhances productivity by leveraging natural language processing and machine learning within predefined contexts. It does not qualify as agentic AI, as it lacks autonomy, goal-setting abilities, and proactive decision-making. It is a tool designed to augment human capabilities, not replace or operate independently of them.
Microsoft Co-Pilot is not an agentic AI. It is better classified as a tool-based AI or an assistive AI. Here’s why:
What Is Agentic AI?
Agentic AI refers to systems that can:
Agentic AI systems are designed to act more like “agents” in a system, proactively performing tasks, adapting to changing conditions, and sometimes even setting sub-goals to achieve broader objectives.
Why Co-Pilot Is Not Agentic AI
Comparison: Co-Pilot vs. Agentic AI
MY TAKEAWAY
It all starts with using a Tech-Led equation-based development and implementation model that reduces technology to a largely irrelevant shiny bauble rather than a valuable adjunct to an agent-based development and implementation model.
Technology doesn’t solve problems; people do, who then use technology to deliver tangible results: https://youtu.be/49BS-MkGoak
To be successful with any technology, AI or otherwise, the Agent-based model requires effective communication and collaboration between all stakeholders (Agents), internal and external to the enterprise. In short, you need a proper relational governance model rather than a “wow, new tech, plug-and-play strategy.”
The above image for this post was generated by AI – does the picture match the post content?
30
Share this:
Related