A Breakthrough Node in Organizational Evolution: Injecting the Initial Power of "Role" into AI

In Tanka's vision, we are currently in a species-level era of civilizational transition: shifting from "humans driving machines" to "humans defining intent, AI executing everything." We can delegate long, tedious execution to AI, retaining only the high-density human intent and responsibility. Almost all of "how to do it" can be computed by models and executed by agents; AI is becoming the primary labor force within enterprises. This is our imagination of the ultimate form of an AI-Native organization.

Tanka's ultimate vision in management science is to achieve deep symbiosis between a very few "carbon-based humans" and countless dynamically scalable "silicon-based agents" through a Manageless engine. When human productivity can be replicated without limits, the "company of the few" will possess universal scalability.

I. Leverage: The First Principle of Future Organizations

How can fewer people leverage countless agents? To achieve this leap, what can we do?

• One conventional path is to perfect the screening and development of the "few": This follows the lineage of modern management science, selecting super-individuals capable of mastering the AI era, and building management mechanisms based on an understanding of "carbon-based humans" themselves. We can attempt to let AI assist in talent screening, assessment, and incentives, placing the right people in the right positions to accelerate the business flywheel.

• The other path is to proactively "construct agents": Here, construction means building a digital twin for every "carbon-based human" that replicates their capabilities and intent. Through this digital twin, tasks are assigned, progress is tracked, and work is directly completed. When we say "human productivity can be replicated a hundred or a thousand times," what is truly being replicated is time, and a capability structure that understands our true intent and is stable and transferable.

As Tanka's HRBP, I am undoubtedly more familiar with the first route, but this also became my limitation. When Mr. Chen constantly pushed us to think:

"What exactly is the first principle of future organizations?"

At that time, I was still trying to distinguish the necessary elements of "managers" versus "employees" from a first-principles perspective. Looking back now, it seems laughable.

"Will managers still exist in an organization of the few?"

Even if a small number of managers still exist, the management authority they hold will no longer come from personal resume experience, span of control, or information asymmetry, because these have all been democratized by AI. The ultimate potential energy differential of a manager may lie solely in the right to allocate resources.

In this sense, using AI to empower traditional management science—a method serving managers—is merely a compromise route in the transition from traditional to intelligent organizations, not a thorough organizational evolutionary breakthrough.

So, when we shift our perspective from "augmenting management tools" to "reconstructing organizational forms," we see that the second path is faster, more imaginative, and more inspiring. The future "few humans" and "many agents" is not an absolute number, but a relative relationship. As long as the "leverage" of an enterprise's employees is high enough and can continue to rise in non-linear dimensions: one person's productivity could potentially be replicated a hundred, a thousand, or ten thousand times, reaching heights we can hardly imagine today. In this sense, it is already an enterprise that can be called AI-Native.

AI Leverage is the first principle of future organizations.

Make AI Work As You—this is true organizational evolution.

II. Injecting Initial Momentum: Role as the First Touchpoint for "Intent Understanding"

For AI to truly "Work as you," it cannot just learn from scattered user behavior trails; it must understand the value function behind the behavior.

We once envisioned a scenario in an internal weekly meeting:

"Can we read all the information about Steve Jobs on the internet and mold a Steve Jobs digital human who can 'work' for employees?"

We quickly discovered that the real difficulty isn't restoring a "portrait of Steve Jobs," but constructing a Steve Jobs digital human that "can work."

The key difference lies in whether this agent possesses a clear and stable will and intent in a work scenario. Therefore, the first and most critical step is to understand the user's intent.

Let's consider: what do we usually rely on to distinguish a person's intent in traditional organizations? Most of the time, its responsibilities are represented by "Job Descriptions" (JD). The JD is a product of the industrial age; static JDs with fixed boundaries will be eliminated. However, within the context of business operations, to provide AI with a first-principle coordinate for "understanding human intent" in an enterprise context, this remains the best touchpoint for projecting human intelligence into AI systems—only it is no longer static, but replaced by a fluid Role.

In Tanka, we start from Role to complete memory initialization and work initiation:

• Role is not a "Job Title,"

• It is a capability and role positioning that can migrate and flow,

• It is the basic unit by which a digital twin can be replicated and scheduled.

III. Scenario Experiments: Role Driving Conversion to Frontend Productivity

Once this underlying technical logic is clarified, Role is no longer an abstract concept; it directly determines whether Tanka can truly "deliver" in critical scenarios:

Scenario 1: The Digital Twin of Departing Employees — Gone, yet Remaining

When a key member leaves an organization, what business managers truly worry about is often not which files they took (this can be solved by compliance reviews, IT audits, etc.), but those invisible, intangible implicit knowledge assets:

• The knowledge and experience they possess.

• The decision-making and thinking logic of how they weigh pros and cons in complex situations.

• The implicit role they played in the team (Leader / Coordinator / Conflict Buffer / Quality Gatekeeper...).

In Tanka, if a person has formed a relatively clear Role and knowledge memory trajectory during their tenure, we can:

• During Offboarding Handover: One-click generation of the work list they need to hand over, with AI automatically reproducing all key information, ensuring the handover doesn't stop at formalities.

• Upon Offboarding Completion: Immediately form a digital twin based on their complete work memory. Authorized personnel within the enterprise can continue to ask the digital twin questions:

  • "Why was it designed this way at the time?"

  • "What is the progress of Project XXX?"

  • "Is there a final delivery document for the XXX item you were following up on?"

• During External Recruitment: One-click generation of a persona introduction and job responsibility description suitable for external recruitment based on the departing employee's Role.

• Over a Longer Cycle: The experience of all departing employees will be further structured and precipitated as enterprise-level memory assets.

In this scenario, what we are trying to solve is preventing knowledge and human judgment from suddenly breaking off with a person's departure—letting the digital employee stand the last watch.

Scenario 2: The First 2 Hours of Onboarding — Synchronizing Corporate Context

In an AI-Native company, the first question for a new hire is no longer:

"Who should I ask for materials?"

But will become:

"Can my AI assistant help me understand the relevant context of this company over the past 2 years within two hours?"

In Tanka's design, what happens within these two hours is:

1. The system understands the newcomer's initial Role based on onboarding information and preset expectations.

2. Based on this Role, Tanka automatically links for their AI assistant:

   • Historical projects and decision records highly relevant to that Role.

   • Key collaborators, typical conflicts, and resolution paths.

   • The organization's implicit expectations and hidden rules for this Role.

3. The newcomer does not need to start accumulating cognition of this enterprise from zero. Instead, through the Role perspective, their personal assistant acquires a compressed understanding of the context from past years in a short time.

   • On Day 1, the assistant will push relevant key materials, onboarding kits, and learning maps, and help supervise personal learning progress.

   • The newcomer can ask the assistant at any time to obtain historical context for relevant projects.

   • Subsequent newcomers taking over the same module can refer to the judgment paths of their predecessors when facing similar scenarios, rather than just blindly rummaging through a pile of historical documents.

Essentially, we are not building a new enterprise knowledge base, but an "memory synchronization system from the perspective of Role," allowing dormant memory to be directly consumed in a scenario-based manner.

IV. Initialization Configuration of Role

Facing the future, Tanka has to deal with a "brand new Role"—it is constantly flowing, composite, and has blurred boundaries. This also brings difficulty to the definition of Role. We have made the following attempts:

1. Extracting Role from Full Behavior

We did something that seems simple but is actually very fundamental: using AI to conversely "read and understand" every carbon-based human's Role and Purpose.

• No longer relying on job descriptions written by HR, but using AI to distill the responsibilities actually undertaken, capabilities frequently relied upon, and typical decision scenarios from real communication records, project history, and deliverables.

• This extraction process is continuous, not a one-time fill-in at onboarding, and is constantly flowing.

2. Adding a Clear Purpose for Everyone

• Purpose is not "the mission of my department," but "my value proposition in the organization during this period": What do I want to do? What results do I want to create for this system?

• Tanka helps everyone constantly reiterate and clarify their Purpose, making Purpose a visible, conversational object.

3. "Rolling Memory" Starting from Onboarding

• For a person newly joining the organization, Tanka no longer just connects their account and permissions, but starts filling in memory around their Role from day one: collaborators, common tools, responsible modules, decision preferences...

• Role is the anchor of this memory pipeline—memory is not "piled up in a warehouse," but bound to "who this person is becoming."

This is what we call the "initial momentum" of injecting Role into AI: not asking humans to fill out forms to accommodate a ready-made system, but letting the system proactively understand the true role of every carbon-based human.

Summary: From Backend to Frontend, Role as the "Neural Skeleton" of the AI-Native Organization

Initialization with Role as the entry point starts the first heartbeat of the AI-Native organization; it constitutes the foundation for us to push backend memory to frontend productivity.

In the past, organizational memory was "backend data" dead on the hard drive; now, through Role initialization, memory becomes "frontend productivity" ready at any time.

We initialize memory starting from employee roles and capabilities, making AI truly an efficient extension of humans, rather than just a search box. In an era where AI can undertake the vast majority of execution, what we need is a system: to build the skeleton of a digital twin for every carbon-based human, allowing the right subject to turn to the right page of memory at the right moment.

This is our thinking on why and how we design Tanka Role. Of course, our exploration is still on the road.