Back to the library

Key question New

AI-processed · Author-approved

Should AI assistants push suggestions automatically to service agents or wait until employees ask for help?

Push or Pull: AI Advice Initiation in Service Work Under Customer Incivility
Marc Christopher Grau · International Conference on Information Systems (ICIS) · 2026
Illustration for: Should AI assistants push suggestions automatically to service agents or wait until employees ask for help?

Loads Spotify's player; Spotify may set cookies. This device remembers your choice.

Key findings

  • Reactive advice significantly lowered negative agency compared to proactive advice, which indirectly strengthened employees' felt responsibility for the task.
  • For polite customer interactions, reactive and proactive advice resulted in similar overall task performance.
  • Under customer incivility, proactive advice significantly outperformed reactive advice, mediated by increased reliance on AI recommendations.
  • The choice of AI advice delivery involves an inherent trade-off: proactive assistance protects performance under hostile interactions, while reactive assistance preserves employee agency and ownership.

What it means for you

CIO / IT Executive

Configure the customer service AI copilot platform to feature a dynamic toggle that automatically switches from reactive to proactive mode when sentiment analysis detects customer incivility.

IT Manager

Review team support queues to identify high-hostility customer channels, and instruct agents to manually trigger AI advice (reactive mode) during standard calls while relying on pushed AI recommendations (proactive mode) during abusive interactions.

Business Strategist

Update standard operating procedures for customer escalation paths to mandate proactive AI assistance specifically for Tier-2 or abusive support workflows to protect task performance.

Researcher

Design a follow-up experimental study investigating the longitudinal psychological burnout effects of prolonged exposure to proactive AI advice under chronic customer incivility.

Policymaker

Draft workplace technology guidelines that require employers to provide frontline customer service workers with control over AI intervention modes to safeguard employee agency and psychological well-being.

Summary

This study investigates how the initiation mode of an AI copilot—proactive versus reactive—affects employee agency, felt responsibility, and task performance. The author conducted a 3 (no advice, reactive, proactive) by 2 (civil vs. uncivil customer) between-subjects experiment with 176 participants with customer-facing experience handling a simulated customer phone interaction.

Problem

Organizations frequently implement AI copilot advice delivery as an arbitrary system default rather than a deliberate design decision. While manually requesting AI help adds cognitive and attentional strain during stressful customer interactions, automatically pushing advice may erode workers' sense of control and personal ownership over the task.

Transcript
Host: Welcome to A.I.S. Insights, powered by Living Knowledge. I'm Anna Ivy Summers.

Expert: And I'm Alex Ian Sutherland.

Host: Today, we examine a study by Marc Christopher Grau presented at the International Conference on Information Systems, or ICIS, in 2026. The study is titled "Push or Pull: AI Advice Initiation in Service Work Under Customer Incivility." Alex, this study investigates how AI copilots deliver suggestions to customer service agents, specifically whether advice should be pushed automatically or pulled on demand.

Expert: Exactly. When organizations deploy generative AI assistants in voice-based service environments, how advice is initiated is often treated as an arbitrary vendor default rather than an active design choice. Systems typically either wait for an agent to formulate and submit a request, which is reactive, or they display suggestions automatically during the call, which is proactive.

Host: And that difference creates a fundamental dilemma for operational workflows.

Expert: It does. Voice service work requires following structured procedures while simultaneously managing live interpersonal dynamics. If an agent has to stop, formulate a query, and type it into a panel while speaking with a customer, that creates attentional friction. Pushing advice automatically reduces that friction, but having a system take the initiative can weaken the employee's perceived control and personal ownership over the task.

Host: To evaluate that trade-off, how was the study set up?

Expert: The author conducted a between-subjects experiment with 176 participants who had prior customer-facing experience. Each participant handled a simulated voice call with a customer reporting a lost debit card. The design crossed three advice conditions—no advice, reactive advice, and proactive advice—with two customer behavior conditions: a civil customer who was polite, and an uncivil customer who was rude and confrontational.

Host: The task required authenticating the caller and explaining replacement procedures. What did the results show regarding the employee's sense of agency?

Expert: On the authorship side, reactive advice significantly reduced negative agency—the felt sense of not being in control of one's actions—compared to proactive advice. That lower negative agency was statistically associated with an indirect increase in experienced responsibility for the work. This relationship held regardless of whether the customer was civil or uncivil.

Host: So retaining initiation agency helped agents feel more personally responsible for the call. But what happened to task performance?

Expert: That depended directly on customer civility. When dealing with a civil customer, reactive and proactive advice produced similar task performance scores. However, under customer incivility, proactive advice significantly outperformed reactive advice. In the uncivil condition, proactive advice yielded a procedural score of 6.41 out of 10, compared to 5.25 for reactive advice.

Host: Why did that performance gap appear under hostile calls?

Expert: The statistical analysis showed this was mediated by reliance on the AI advice. Under stressful, uncivil interactions, participants in the proactive condition relied more heavily on the recommendations. In contrast, participants in the reactive condition requested advice in only 12 out of 61 calls, averaging just three queries when they did. The behavioral barrier to asking for help during an uncivil interaction was simply too high.

Host: What are the practical implications for service managers and system designers?

Expert: First, organizations should treat advice initiation as an explicit design variable rather than accepting a vendor default. Second, the choice involves an inherent trade-off. Proactive advice provides instrumental performance support during high-demand, stressful customer interactions. However, reactive advice preserves employee agency and felt ownership.

Host: Which means matching the initiation mode to the deployment environment.

Expert: Precisely. In environments where customer incivility is frequent and procedural compliance is paramount, proactive advice helps protect performance. Where employee discretion, empathy, and ownership are central to service quality, reactive advice is more defensible. The findings also highlight the potential value of adaptive systems that switch from reactive to proactive assistance when customer hostility is detected.

Host: A helpful framework showing that how assistance enters a workflow is just as important as the assistance itself. Alex, thank you for sharing these insights.

Expert: My pleasure, Anna.

Host: And thank you for listening to A.I.S. Insights. Until next time.
Artificial Intelligence & Automation Information Systems & Management Human-Computer Interaction Data Analytics & Big Data Digital Transformation & Strategy Cybersecurity & Privacy Cloud & Infrastructure E-Commerce & Digital Platforms Operations & Supply Chain Sustainability & Green IS

Artificial Intelligence, AI Copilots, Customer Incivility, Customer Service, Human Agency, Advice Initiation