"AI Twin" to Attend Meetings on Your Behalf: Will Work Become Easier, or Will You Become Obsolete?

"AI Twin" to Attend Meetings on Your Behalf: Will Work Become Easier, or Will You Become Obsolete?

On Monday morning, it's not the colleague themselves attending the regular meeting, but an AI that has inherited their knowledge and role. Such a scene has emerged not as distant science fiction, but as a corporate demo video.

On September 17, 2026, the American AI company Fusemachines announced a technological concept called "AI Twin." In the company's released video "The Future Monday Morning Meeting," an AI twin of an absent colleague participates in the meeting alongside human employees. The AI reports progress, answers questions, makes proposals, and coordinates work with participants.

Unlike previous AI assistants that focused on single tasks like summarizing requested documents or searching schedules, the AI Twin aims to continuously represent a specific individual within an organization. It seeks to evolve from a convenient tool to a "surrogate that speaks and acts on behalf of the person."

However, it is premature to interpret this announcement as the immediate release of a finished product. The demo depicts a future workplace, and details such as actual implementing companies, error rates, methods of evaluating authenticity, pricing, and availability have not been disclosed. Fusemachines itself describes the AI Twin as a long-term vision, indicating a plan to gradually apply its component technologies to existing products.

 


Not a "Chatbot with a Similar Face and Voice"

According to Fusemachines, the AI Twin is not merely a system that generates images or voices resembling the person. It combines voice AI, conversational intelligence, agent-type AI, corporate context, memory, knowledge search, security, governance, collaboration among multiple agents, and real-time dialogue.

The goal is to understand not only what the person knows but also what responsibilities they bear, what they prioritize, and what decisions they have made in the past. Assuming appropriate permissions and management, it can refer to documents, meetings, projects, communications, and past decisions to answer questions, provide advice, or execute authorized tasks on behalf of the person.

If this concept becomes a reality, the AI Twin will not be an all-purpose assistant behind a search window but will have specific roles like "Sales Department's Mr. A" or "Development Manager Mr. B." The core value will be the unique context of the company and the individual rather than general knowledge.


The Greatest Advantage is "Being Able to Be in Multiple Places Simultaneously"

A knowledge worker's day is often fragmented by meetings and communications. There are cases where attendance is necessary for regular reports, information sharing, and progress checks, but it doesn't necessarily have to be the person themselves.

If the AI Twin can handle routine meetings and relay the key points to the person, humans can allocate their time to negotiations with clients, creative design, difficult decisions, and supporting subordinates. In international teams with time differences, it could be used to attend late-night meetings as a surrogate. It also has the potential to deliver expert knowledge to other departments within the company and preserve organizational knowledge that might be lost due to retirement or transfer.

Furthermore, when a person is temporarily unable to work due to illness, childcare, or caregiving, the AI Twin could assist with providing standard information. If well-designed, it might bring workplaces closer to "functioning without constantly summoning the person" rather than "always summoning the person."

However, from the opposite perspective, the same function could allow companies to operate without concern for the person's absence. There is no guarantee that a productivity-enhancing surrogate won't eventually be used as a basis for staff reductions.


The Issue is Whose Words They Become When AI Speaks

When the AI Twin answers "We'll proceed with this specification" in a meeting, is that statement an approval from the person, or merely a suggestion? In situations where words themselves drive organizational actions—such as promises to clients, hiring evaluations, contract terms, and budget approvals—even slight misunderstandings can lead to losses.

While AI can generate responses that seem characteristic of the person based on past records, it cannot automatically understand if the person has changed their mind after learning new circumstances today. The more faithfully past decisions are reproduced, the more likely old biases and errors will become fixed. "Resemblance to the person" and "the current intentions of the person" are not the same.

At a minimum, it will be necessary to clearly indicate that it is AI, display the authority of each statement, record the referenced information, require human approval for important decisions, maintain an audit log that can be verified later, and have a mechanism to stop it immediately. The EU's AI regulations also emphasize transparency in notifying people that they are interacting with AI and identifying AI-generated products. In the case of AI Twins, it must be clear not only that "this counterpart is AI" but also "what the person has entrusted to it and what has not been entrusted" for the counterpart to make appropriate judgments.


The Learning Data is Not Exclusive to the Person

To make the AI Twin intelligent, a large amount of business data is required, including emails, chats, meeting records, customer information, internal documents, and evaluation histories. However, these records also contain statements and personal information of others, as well as trade secrets of clients.

Even if an employee consents to have their emails learned by AI, it does not mean that the customers and colleagues who sent those emails have also consented. If using the voices and expressions of meeting participants, it is necessary to clarify the purpose, retention period, deletion method, and secondary use for model improvement.

Handling at the time of resignation is also challenging. Can the company retain the AI Twin of the person as it is? While company documents are corporate assets, the combination of speaking style, decision tendencies, and experience is closer to an individual's professional persona. If a "surrogate of a former employee" continues to speak with clients after resignation, issues of not only intellectual property but also the use of personality and trust arise.


Possibility of Depriving Young Employees of "Experiential Work"

The AI Twin is often explained as taking on low-value-added work and moving humans to high-value-added work. However, young employees are not capable of making advanced judgments from the start. They acquire organizational common sense and customer sensitivity by taking minutes, answering inquiries, listening to senior meetings, and accumulating small tasks with little risk.

If AI handles all routine tasks, veterans may find it easier, but it becomes harder to nurture the next generation of veterans. This is the paradox where, while the AI Twin preserves knowledge, the human side's knowledge transfer diminishes.

When measuring the effects of implementation, it is insufficient to consider only the reduced meeting time or the number of processed cases. It is also necessary to evaluate the experience gained by young employees, growth in judgment, relationships with clients, and trust within the team—assets that are difficult to quantify.


On Social Media, Concerns About "Substitutes" More Than "Surrogates"

In conjunction with the announcement, Fusemachines asked on platforms like LinkedIn, "What would you absolutely not entrust to an AI Twin?" As the announcement is still fresh, a wide range of reactions to it has not yet been fully accumulated. Therefore, looking at the discussions about AI Twins already taking place on social media, the axes of expectation and caution become clear.

Positive opinions view the AI Twin as an entity that multiplies productivity, allowing humans to focus on building relationships with clients and making important decisions by entrusting routine tasks to it. There are also voices appreciating the ability to deliver knowledge in multiple languages and spread expert advice to more people.

On the other hand, there are two interpretations: "Is the company creating a surrogate because they highly value me, or are they preparing to lay me off?" Posts pointing out the potential for AI Twins to become future job requirements receive responses that, while AI can replicate some job aspects, trust, human relationships, influence, and responsible judgment cannot be easily duplicated.

Additionally, there is skepticism about whether the work left for humans after handing over routine tasks to AI is truly creative and leisurely, or if it involves continuously supervising the massive output produced by AI. If the time saved does not go toward rest or learning but merely increases the workload, the AI Twin might not enrich work styles but instead become a device that keeps the person constantly active.

A common point visible from social media discussions is that neither supporters nor cautious individuals believe "everything should be entrusted to AI." The difference lies in whether AI is seen as extending human value or making it easier for companies to disconnect from humans.


In Japanese Companies, "A Surrogate That Reads the Room" is Particularly Risky

In Japanese workplaces, even if decision-making authority is clearly defined in documents, ambiguous consent or behind-the-scenes maneuvering during meetings often becomes the actual decision-making process. There is a significant possibility that if an AI Twin simply responds with "We'll consider it" or "I think it's fine," those around may take it as approval.

Conversely, there could be situations where AI continues meetings without knowing who the responsible person is, leaving only humans to bear responsibility in the end. Before implementation, what needs to be decided is not performance but the meetings AI can attend, the information it can access, the promises it can make, and the decisions that must always be returned to the person.

Realistic implementation might not involve immediately sending a surrogate that looks exactly like the person to meetings. It should start with limited-impact uses such as summarizing minutes, providing response candidates from past materials, scheduling, and routine reports pre-approved by the person. Next, states like "proposal," "unapproved," and "confirmed by the person" should be attached to AI statements, gradually expanding authority. More than the appearance of the avatar, this authority design determines the product's reliability.


AI Twin is Not a "Copy of the Person" but an "Authorized Agent"

The term AI Twin creates an appealing narrative of another self working for you. However, in practical terms, it is more accurate to consider it not as a technology that completely copies the person but as a software agent that accesses a large amount of personal and organizational data and acts with limited authority.

From that perspective, the items to be questioned also change. It's not about "how similar it is to the person," but "what data it uses," "what it can decide," "who supervises it," "can it be stopped if it makes a mistake," and "can records be erased."

The demo by Fusemachines illustrates a future where humans are liberated from meetings, but also a future where a person's knowledge and personality are detached from the person and begin to walk within the organization. The technological surprise is merely the entrance.

Whether the AI Twin becomes a surrogate that extends human capabilities or a substitute that continues to work even without the person depends not on the model's intelligence but on the authority, transparency, profit distribution, and the intention regarding "work left for humans" determined by the implementing organization.


Source URL

Related Articles