AI Customer Service vs. Human Customer Service: A Survey on Acceptance Among 300 Clients in the Packaging Industry

In 2024, we did something that may not be very "AI optimistic"—we sent questionnaires to 300 real customers, asking about their acceptance of AI customer service and human customer service. The results were a bit surprising:

In the "quotation inquiry" stage, 85% of customers accept AI customer service; in the "process consultation" stage, acceptance drops to 35%; in the "complaint handling" stage, acceptance is only 12%.

This exposes a widely overlooked fact: AI customer service is not an "all-scenario replacement" tool—its acceptance heavily depends on the specific scenario. The packaging industry is also a highly "non-standard" industry, so the implementation of AI customer service requires more detailed boundary design than in general scenarios.

This article aims to clearly explain three things: the real data from the 300-customer survey, the actual capabilities of AI customer service in different scenarios, and 5 recommendations for packaging enterprises implementing AI customer service.

Research Background: 300 Real Customer Samples

Research subjects: 300 customers who contacted us from January to October 2024 (including 178 brand purchasers, 65 traders, 35 individual/entrepreneurial customers, and 22 peer factories).

Research method: online questionnaire + telephone follow-up. Each customer answered 4 core questions: ① At which stage are you more willing to communicate with an AI customer service agent? ② At which stage are you more willing to communicate with a human customer service agent? ③ What is your biggest concern about AI customer service? ④ What is the maximum AI customer service response delay you can accept?

78% of the samples are B-end customers (purchasers, traders), and 22% are C-end customers (small-brand entrepreneurs, individual customers).

Data 1: Highest Acceptance in the Quotation Inquiry Scenario

In the "quotation inquiry" stage, 85% of customers indicated that "it is acceptable for the AI customer service to provide an initial quotation first, followed by a human agent to follow up on the details."

Specifically, customers' expectations for the AI quotation customer service are: ① fast response (ideal latency < 10 seconds, acceptable latency < 30 seconds); ② accurate basic parameters (material, size, quantity, application); ③ provide a price range (not an exact figure, but a reference range).

Reasons for accepting AI quotations: ① slow response from human agents (1–2 hour reply time during peak hours); ② quotation requires repeated communication of basic information (tiring); ③ pricing itself requires repeated negotiation, so the neutrality of AI is actually welcomed.

The 15% who are unwilling to accept it are mainly brand-side procurement managers, who believe that "AI quotations are inaccurate and only waste time," and prefer to establish long-term relationships directly with business managers.

Data 2: Sharp Decline in Acceptance for Process Consultation Scenarios

In the "process consultation" stage (e.g., "Can this box be made with a certain process?" or "Which material is more suitable?"), only 35% of customers accept AI customer service, while the remaining 65% explicitly state that "a human is required."

The reasons for rejecting AI are highly concentrated: ① AI cannot determine process feasibility (the boundaries of AI process pre-inspection were discussed in a previous article); ② process consultation often involves specific material-process-cost trade-offs that require experienced judgment; ③ customers worry that "the process advice given by AI is unprofessional and could actually be misleading."

The 35% who are willing to accept it are mainly those seeking: ① general process knowledge (e.g., "What is hot stamping?" or "What is UV?"); ② technical parameters (e.g., "What is the required printing resolution?"); ③ process workflows (e.g., "How long from sampling to mass production?"). AI handles these "knowledge-based" questions fairly well.

However, when it comes to judgmental questions like "Is my specific product suitable for this?", customers strongly prefer communicating with a human. One client from a cosmetics brand noted in the survey: "My product is a perfume gift box with high essential oil content—can I use a PVC box? I asked the AI customer service, and it said 'it is recommended to consult a professional'—so why do I need AI?"

Data 3: Extremely Low Acceptance of Complaint Handling Scenarios

In the "complaint handling" stage, only 12% of customers accept AI customer service, while the vast majority (88%) clearly state that "human communication is a must when problems arise."

The reasons are almost identical: ① Complaints require emotional understanding, and AI has no emotions; ② Complaints involve specific accountability determination, which AI cannot judge; ③ Complaint handling often involves concessions and negotiation, and AI has no authority to do so.

One customer who left a comment during the research wrote: "There was a problem with your product, and the AI customer service replied: 'We have recorded your issue and will resolve it within 24 hours.' When I said I wanted a refund, the AI customer service replied: 'Please provide your order number.' Can you blame me for being angry?"

The core need in complaint scenarios is to "be taken seriously," while AI's standardized replies tend to escalate emotions instead.

Data 4: Customers' Biggest Concerns About AI Customer Service

The top three customer responses regarding their biggest concerns about AI customer service were:

First place (62%): "The information provided by AI is inaccurate, leading me to make incorrect judgments." Customers are most concerned about "being misled," particularly in areas that directly affect costs such as processes and quotations.

Second place (48%): "AI does not understand my specific needs; the solutions it provides are not applicable." Customers worry that AI will apply generic solutions to their particular scenarios.

Third place (35%): "The AI doesn't have the authority to resolve issues, and repeated verification wastes time." Customers are highly sensitive to the experience of transitioning from "AI to human agent"—if the handoff isn't smooth, the AI actually becomes an obstacle instead.

5 Implementation Recommendations

Based on feedback from 300 clients, we have summarized 5 recommendations for packaging companies to implement AI customer service.

Recommendation One: Define the boundaries between AI and human service by scenario. Standardized scenarios such as quotation inquiries, knowledge Q&A, and initial contacts should be handled by AI; scenarios such as process consultation, complex requirements, and complaint handling should be transferred directly to human agents. Clear boundary design is more effective than "AI as the fallback plus human as the supplement."

Recommendation Two: AI should provide a "reference range" rather than "exact figures." Packaging is a non-standard industry, and AI quotations cannot be accurate down to the yuan. Giving customers a reference interval (for example, "Based on the parameters you described, the price ranges from 8-15 yuan per unit, depending on the process details") actually builds trust.

Recommendation Three: AI customer service must be capable of "seamless transfer to human agents." If a customer expresses "I want a human agent" or shows emotional fluctuation during an AI conversation, the AI must immediately transfer to a human agent and not make the customer repeat the problem. The best experience is when the AI automatically passes the conversation context to the human agent.

Recommendation Four: The training data for AI customer service must include "negative cases." What customers dislike most is the AI's "empty-talk replies." Use "original customer complaint + correct human response" as training data, so that the AI learns to directly say "I will transfer this issue to a human agent for you" when it cannot resolve the problem, instead of brushing it off with "noted."

Recommendation Five: Turn the AI customer service's "I don't know" into an advantage. The knowledge boundaries of AI in the packaging industry are an objective reality. Directly telling the customer "I need to verify this process detail; I will have our process specialist get back to you within 30 minutes" is more professional than forcing a response with generic knowledge.

Comparison of AI Customer Service Across Different Industries

Comparing the acceptance of AI customer service in the packaging industry with other industries shows that the packaging industry is one of the industries with a "high barrier to AI customer service."

E-commerce retail (e.g., apparel, 3C): AI customer service acceptance is 70-85%, because products are highly standardized and inquiries mainly revolve around "price, inventory, and logistics."

Food and beverage delivery: AI customer service acceptance is 60-75%, with inquiries primarily about "menus, delivery, and promotions."

Banking and finance: AI customer service acceptance is 40-55%, involving account and fund issues where customers prefer human agents.

Healthcare: AI customer service acceptance is 25-40%, involving health issues where customers are highly sensitive.

Packaging and printing: AI customer service acceptance is 35-50% (overall), because of strong non-standardization, but acceptance varies dramatically across scenarios—85% for quotation, 35% for craft/process, and 12% for complaints.

The implementation plan for AI customer service in the packaging industry must be segmented by scenario and cannot simply replicate the "full-scenario AI" model used in e-commerce retail.

Summary

AI customer service in the packaging industry is not a "full-scenario replacement" tool, but a "scenario-specific assistant" tool. AI can handle scenarios such as quotation inquiries, knowledge Q&A, and initial contact very well; however, in scenarios involving process consultation, complex requirements, and complaint handling, AI actually reduces the customer experience.

Real acceptance data from 300 customers tells us: implementing AI customer service in the packaging industry requires clear scenario boundary design and cannot pursue a single metric like "AI replacement rate." The correct metric is the combined optimization of "AI handling efficiency + customer satisfaction + human conversion rate."

Once you understand this boundary, you can maintain stable customer service quality in the AI era—the steps that should be handled by AI are handled by AI, the steps that should be handled by humans are handled by humans, and the seamless connection between the two is the true "intelligent customer service."

Further Reading: Why AI packaging quotations look accurate but end up with large deviations, The boundaries of AI process pre-inspection: which issues AI really cannot detect, 5 real applications of AI in the packaging industry: which are hype and which are trends, How to ask more professional questions when inquiring about packaging boxes by phone