AI is redrawing the boundaries of customer success.

The most intractable problem faced by the Customer Success department may well lie outside the department itself.

For example, the features promised by the sales team are long overdue for launch, technical obstacles arise during implementation and delivery, and the customer's internal business processes fail to be adjusted as expected. When the renewal period comes, the customer considers that the product fails to meet their procurement expectations, and all the pressure eventually falls on the customer success team.

Such scenarios are not unfamiliar in the enterprise software industry.

Customer Success Managers (CSMs) can coordinate resources and push for problem solving, but they do not necessarily have the authority to decide on product R&D plans, technical solutions and delivery input. No matter how standardized the service process is executed, it cannot guarantee that customers will achieve the expected business outcomes.

From 2022 to 2025, the Cui Niu Hui China SaaS Conference has held four consecutive closed-door meetings on customer success, focusing on discussions around issues such as customer value, service efficiency and organizational collaboration. When sorting out these topics again now, you will find a consistent contradiction:

The operational responsibilities undertaken by customer success are getting heavier and heavier, but the key factors affecting customer success are scattered across the entire enterprise.

Entering the AI era, this contradiction has become more prominent. In the past, vendors mainly helped customers get the software up and running and make better use of it; now, as Agents begin to undertake actual business tasks, customers also need to verify whether the tasks can be completed, whether the results are accurate, and whether the input can generate returns.

When software starts to work on behalf of people, who should be responsible for customer success?

01 The problems of customer success often lie outside the department

Renewal rate, customer churn rate, and Net Revenue Retention (NRR/NDR) are important indicators for SaaS enterprises to measure customer operation status.

The customer success team usually needs to take corresponding responsibilities for these indicators, but the factors affecting the results are far more complex than daily services.

Customers do not renew their subscriptions, which may be due to product functions that do not meet their needs, poor implementation effects, or changes in the customer's own business environment.

Customers do not make additional purchases, which may not necessarily be caused by inadequate services, but may also be affected by budget, product capabilities and business development stages.

If you do not analyze the reasons and only attribute the results to the insufficient work of the customer success team, it will be difficult to find truly effective improvement methods.

The difference between customer success and traditional customer service is also reflected here.

Customer service usually focuses on problem response, fault handling and service quality; customer success also needs to pay attention to whether the product is truly integrated into the business process, whether customers obtain the expected value, and whether the cooperation between the two parties can be sustained.

These tasks are inseparable from sales, product, R&D and implementation delivery teams.

The commitments made by the sales team during the signing stage determine customers' expectations for the product. If the CSM does not get involved until the contract is signed, what they take over may already be a target that is difficult to fulfill.

When the project enters the implementation phase, the smooth launch of the system depends on technical capabilities, data preparation and project management of both parties.

Even if the customer success team identifies problems that involve product improvement and R&D investment, joint decision-making by other departments is required.

Therefore, customer success can be led by one department, but cannot be completed independently by a single department.

Enterprises need to align sales commitments, implementation delivery and continuous operation around the same customer goal, clarify who will make decisions on key issues, who is responsible for solving them, and how to conduct acceptance. Otherwise, the customer success team will easily become the recipient of all problems, without the resources needed to solve them.

The measurement methods for customer success also need to change accordingly.

System launch, active user engagement, and timely service response can reflect product application and service processes, but cannot fully prove that customers have obtained value. Customers ultimately care about whether operating costs are reduced, work efficiency is improved, and whether the investment generates reasonable returns.

This requires the customer success team to have a deeper understanding of the business, while taking into account the vendor's own service costs.

For key customers, it is sometimes inevitable to invest more engineering and service resources, but the investment also needs to have boundaries. If every renewal relies on a large amount of additional development and manual support, even if the customer is retained, it may not be a healthy business.

Customer success requires calculating not only the value account for customers, but also the operation account for the vendor itself. Whether both accounts can be settled accurately tests an enterprise's product capabilities, resource allocation and operation management level.

02 AI is redrawing the boundary of customer success

In the traditional SaaS era, the important work of customer success is to help customers complete product launch, promote employees to use the product, and expand the application scope through continuous operation.

The emergence of AI Agent has changed the working mode of some enterprise software. Tasks that were previously completed by employees operating the system can now be handed over to Agents for execution.

Customers not only pay attention to whether the product is easy to use, but also judge whether AI can correctly understand business requirements, call relevant systems, and complete tasks stably.

This change puts higher requirements on delivery.

An Agent that can complete tasks in the demonstration environment may fail to operate stably after entering the enterprise production environment due to inconsistent data calibers, insufficient system permissions, and complex business rules.

Even if there is no problem with the model itself, the interfaces, approval processes and exception handling between different internal systems of the enterprise may also affect the final results.

Especially in complex enterprises, many business rules exist in the experience of employees, and are not fully recorded in systems or documents.

To enable Agents to undertake relevant work, vendors need to deeply understand the real business and convert these implicit rules into executable and verifiable system capabilities.

AI can enable software to have a certain function faster, but it cannot guarantee that it works stably in a complex enterprise environment.

As a result, the customer success team needs to cooperate more closely with R&D, data and implementation delivery personnel. After the product is launched, it is also necessary to continuously pay attention to the Agent's task completion status, exception handling and business effects. Some traditional customer operation work has also begun to extend to technical delivery and continuous operation management.

At the same time, AI is changing the internal division of labor of the customer success team.

Customer information sorting, ticket classification, meeting minutes, service report generation, and part of customer risk identification can already use AI to improve efficiency. Information that previously required repeated manual processing can now be aggregated and analyzed more quickly.

However, the improvement of information processing efficiency does not mean that business judgment can be fully automated.

For example, the system detects a decline in customer usage, which may be caused by business contraction, poor product experience, or the intervention of competitors. AI can sort out relevant clues, but it still requires professional personnel to judge the cause, coordinate resources, and decide on follow-up actions.

As repetitive work is gradually reduced, customer success personnel need to devote more energy to business diagnosis, coordination of complex problems and value verification.

Being familiar with products and good at communication is still important, and the understanding of customer business and technical implementation will also become more critical.

This does not require every CSM to become an engineer. Facing complex AI projects, enterprises need a clearer professional division of labor and establish a working mechanism that can coordinate different professional capabilities.

03 Can FDE solve the problem of large-scale delivery?

In this context, FDE (Forward Deployed Engineer) is becoming a role worthy of attention.

FDEs go deep into the customer site, combine software engineering capabilities with actual business needs, and solve complex problems that standardized products cannot handle directly.

For enterprise AI projects that need to connect multiple systems and process non-standardized processes, this working method has practical value.

But FDE also faces a commercial problem, that is, the fact that engineers can deliver one project does not mean that the enterprise has the ability to deliver in batches.

There are differences in the system architecture, data environment and business processes of different customers. If each project needs to re-sort out requirements, develop interfaces, and design workflows, the number of customers increases, and the engineering team and delivery costs may also grow simultaneously.

Even if AI Coding reduces the cost of code development, business understanding, system integration, test acceptance and subsequent maintenance still require investment.

Therefore, AI vendors must clarify which customer requirements are worthy of customization, which capabilities should be productized, which work can be reused through tools, and which complex problems still require expert intervention. These choices directly affect delivery efficiency and profit levels.

For the customer success team, project launch does not mean the end of the work. After the Agent is put into use, the customer's business rules, model capabilities and related systems may all change, and it is still necessary to continuously monitor the effect, handle exceptions and optimize applications.

FDEs need to make AI really work, and customer success needs to make customers continue to benefit. The business experience accumulated by FDEs on site needs to promote product improvement and precipitation of delivery capabilities; the application feedback and business objectives mastered by the customer success team should also become an important basis for technical iteration.

AI reduces part of the cost of software development, but further pushes enterprise competition to business understanding, engineering delivery and continuous operation. Customer success is stepping out of the customer success department, and is becoming a business responsibility that the entire organization needs to undertake together.

This article is from the WeChat official account "Neuters" (ID: Neuters), author: Neuters, published with authorization from 36Kr.