Quinnipiac University

QUIP-RS

“There’s Always More Work:” Generative AI as Resource and Ratchet for Working Mothers

A navy blue graphic with the Quinnipiac University logo and text that says “There’s Always More Work”: Generative AI as Resource and Ratchet for Working Mothers  Emily McNeill ’28, MBA ’29 Department of Applied Business Quinnipiac University, Hamden, CT

QUIP-RS

“There’s Always More Work:” Generative AI as Resource and Ratchet for Working Mothers

Emily McNeill ’28, MBA ’29 completed this project as part of the Interdisciplinary Program for Research and Scholarship under the mentorship of Shizhen Jia, assistant professor of business analytics and information systems. 

Overview

This project was completed through Quinnipiac University’s QUIP-RS summer research program under the mentorship of Professor Shizhen Jia. The project included interviews with working mothers and surveys examining the relationship between generative AI use, productivity, work-family balance, and caregiving responsibilities. Data was analyzed, identifying key themes and patterns, and contributing to the development of the study’s findings. The findings suggest that while generative AI often improves efficiency and saves time, working mothers frequently reinvest those gains into additional work and family responsibilities, creating what we describe as a “resource and ratchet” effect.

Student Researcher

Quinnipiac Q logo

Emily McNeill ’28, MBA ’29

Applied Business

School of Business

“There’s Always More Work:” Generative AI as Resource and Ratchet for Working
Mothers

 

Introduction

Generative artificial intelligence (AI) is being adopted at an unusually fast pace across the knowledge economy. By late 2024, nearly 40 percent of U.S. adults aged 18 to 64 had used generative AI tools, 23 percent of employed respondents had used them for work in the prior week, and 9 percent reported daily workplace use (Bick et al., 2024). Recent evidence suggests that generative AI can reduce time spent on routine communication and after-hours work, positioning it not merely as a productivity tool but as a technology that may alter the temporal organization of work (Dillon et al., 2025).

That temporal shift is especially consequential for working mothers, who face persistent disadvantages in wages, hiring, evaluation, and advancement due to stereotypes portraying them as less competent and committed than otherwise equivalent workers (Budig & England, 2001; Correll et al., 2007). The ideal worker norm compounds this disadvantage by rewarding continuous availability and freedom from domestic obligations (Ashforth et al., 2000). COVID-era evidence further suggests that nominal flexibility can reinforce rather than reduce these expectations for working mothers (Zanhour & Sumpter, 2024).

Generative AI may therefore represent both opportunity and risk. Faster drafting, summarization, and documentation may increase temporal control, improve interruption recovery, and reduce administrative burden. At the same time, lower time costs may encourage assumptions that work can be completed more quickly, increasing expectations for responsiveness and output. Time-diary evidence links higher AI exposure to longer workdays and reduced leisure in settings where productivity gains are paired with limited worker bargaining power (Jiang et al., 2025), while policy analyses warn that AI-enabled flexibility may reinforce rather than dismantle gendered caregiving expectations (Collett et al., 2022).

This paper addresses that tension as an information systems (IS) problem. Whereas prior connectivity technologies expanded when and where workers could be reached, generative AI may reshape expectations regarding what can reasonably be completed within a given period of time. Consequently, the effects of AIenabled flexibility depend not only on the technology itself but also on how organizations interpret productivity gains, allocate recovered time, and establish expectations around performance and availability. Building on a framework that conceptualizes AI-enabled flexibility as a double-edged affordance—simultaneously a resource that expands temporal control and a potential ratchet that intensifies availability norms—we report qualitative findings from semi-structured interviews with eleven working mothers employed in knowledge-intensive roles who use generative AI tools. Working mothers represent a critical theoretical case because they are both likely beneficiaries of increased flexibility and particularly vulnerable to intensified availability expectations. Accordingly, we ask: How do working mothers experience AI-enabled flexibility in relation to temporal control, interruption recovery, workload demands, and changing expectations regarding productivity and availability?

Four research questions guide the study: how AI reshapes temporal control and interruption recovery (RQ1); how availability and responsiveness expectations are changing (RQ2); how caregiving and ideal worker expectations condition AI-enabled flexibility (RQ3); and what role organizational boundary protections play in whether AI functions as a resource or source of intensification pressures (RQ4).

Our analysis yields three observations. First, participants consistently described AI as a resource that improved efficiency, workflow control, interruption recovery, and reduced stress. Second, recovered time was typically reabsorbed into additional work rather than returned to workers as discretionary time, converting efficiency gains into expanded output. Third, evidence for AI-specific expectation escalation was limited and uneven. While some participants perceived growing assumptions that work should be completed more quickly, most attributed availability pressures to pre-existing workplace cultures and role expectations rather than to AI itself. Rather than demonstrating an established availability ratchet, the findings identify conditions that may contribute to expectation escalation over time, including workload reabsorption and selfrecalibration of performance standards.

Theoretical background

The motherhood penalty and the ideal worker norm

Mothers face persistent wage penalties net of human-capital controls (Budig & England, 2001), and evaluators rate equally qualified mothers as less competent and committed, applying stricter standards and lower salary recommendations (Correll et al., 2007). These biases align with the ideal worker norm of uninterrupted availability and work- first devotion (Ashforth et al., 2000), whose roots lie in feminist organizational theory: ostensibly neutral organizational structures are built around an unencumbered, implicitly masculine worker (Acker, 1990). Consistent with that tradition, we treat gender as socially constructed and relational rather than categorical: the motherhood penalty arises not from category membership but from gendered expectations that assign caregiving, and its visibility costs, to women. COVID-era remote-work evidence shows that flexibility can entrench, rather than relax, these expectations for working mothers (Zanhour & Sumpter, 2024).

In the IS literature, gender and technology research emphasizes that women's experiences with information technologies are shaped by both gendered structures and variation within women's experiences (Mennega & de Villiers, 2021; Trauth, 2023). Thus, the motherhood penalty provides a theoretical context for examining how AI-related changes in work may interact with gendered expectations; the present study does not directly measure the motherhood penalty through wages, promotion, hiring, or performance evaluation. The experiences described in this study should therefore not be assumed to represent working mothers as a homogeneous group, as the present study does not collect sufficient demographic and caregiving information to examine how race, age,  employment classification, or caregiving support may shape these dynamics.

Flexibility and work-family boundary management

Boundary theories clarify that the effects of flexibility depend less on nominal discretion than on perceived control over boundaries: border theory conceptualizes workers as managing the permeability between domains (Clark, 2000), boundary theory emphasizes micro role transitions (Ashforth et al., 2000), and cross-level research shows that individual preferences and organizational climates jointly shape boundary strategies (Kossek & Lautsch, 2012). Crucially, flexibility can trigger stigma and self-exploitation when organizations treat visible time investment and responsiveness as signals of commitment (Chung, 2022). More broadly, whether technology-enabled time savings become flexibility or intensified work depends on how organizations allocate workload, interpret performance, and establish norms around responsiveness and availability. Thus, the availability of a flexible technology does not by itself determine whether workers experience greater control or greater work intensity.

Generative AI and the temporal reorganization of work

Generative AI compresses routine cognitive work and may expand the range of tasks that can plausibly be completed within a given period. Field evidence shows task-level productivity gains (Brynjolfsson et al., 2025) and altered work-time patterns in large deployments (Dillon et al., 2025). AI exposure has also been linked to work-time expansion in some contexts (Jiang et al., 2025), while early labor-market effects elsewhere remain small on average (Humlum & Vestergaard, 2025). Because women are disproportionately concentrated in highly exposed jobs (Gmyrek et al., 2025) and policy analyses warn that AI-enabled flexibility may reinforce gendered caregiving expectations (Collett et al., 2022), distributional consequences are central rather than incidental. At the same time, these consequences are unlikely to be uniform across workers or sectors, making organizational context important to understanding when AI-enabled efficiency produces flexibility versus intensified expectations.

What generative AI changes: from reachability to plausible completion

Generative AI extends dynamics associated with earlier connectivity technologies rather than representing an entirely separate phenomenon. Email, smartphones, and enterprise messaging expanded reachability, extending when and where workers could be contacted, and boundary scholarship accordingly theorized the permeability of work-home borders to communication (Ashforth et al., 2000; Clark, 2000). Generative AI operates on an additional margin: by compressing completion time for the work itself, it may change what can plausibly be expected to be completed, and by when. Smartphones and email expanded reachability; generative AI can make rapid completion more plausible and therefore potentially alter expectations surrounding task turnaround. The boundary problem shifts accordingly: where delay could once be legitimated by the time a task required, AI may weaken that justification when faster completion is perceived as possible; the contested resource becomes the evaluative meaning of time, that is, what slowness signifies once “with AI this should be quick” is an available inference.

Framed in sociotechnical terms, generative AI affords faster drafting, summarization, and task recovery, but whether those affordances become flexibility for workers or intensification of demands depends on norms, organizational practices, governance, and evaluative systems. This interpretive contingency is central to IS work on the organizational embedding of digital technologies (Pang et al., 2024), AI-related work intensification and monitoring (Jiang et al., 2025), and technostress more broadly (Tarafdar et al., 2019). The dual-pathway framework developed next distinguishes the resource effects that are evident in the present data from the more tentative possibility that efficiency gains may contribute to escalating expectations.

AI-enabled flexibility as resource and

Integrating these streams, we conceptualize AI-enabled flexibility tools as creating two analytically distinct pathways. In the resource pathway, AI reduces the time and cognitive burden of routine communication, coordination, drafting, and documentation, increasing temporal control, easing recovery from caregiving interruptions, and making high performance more structurally compatible with caregiving.

In the ratchet pathway, the same efficiency gains may contribute to expanded definitions of acceptable responsiveness, output, and availability. This pathway may involve several distinct processes that should not be treated as equivalent: time savings may be reabsorbed into additional work; workers may recalibrate their own standards for how quickly work should be completed; and, potentially, managers may escalate explicit expectations for turnaround and sustained output. We term this the availability ratchet: a proposed process through which technology-enabled efficiency gains may eventually be reinterpreted as raising the minimum acceptable standard of responsiveness and output, converting time savings into new obligations.

For mothers, the double bind is acute: the caregiving demands that make AI-enabled flexibility useful are also the demands that may make intensified expectations more difficult to meet. The interviews provide evidence of some conditions that could contribute to such a process, particularly workload reabsorption and self-calibration, but provide more limited evidence of explicit AI-related managerial escalation. The present study does not establish that these processes occur in a particular temporal sequence or that they necessarily consolidate into a ratchet.

The theoretical claim is therefore not that generative AI has already redefined the organizational baseline of acceptable responsiveness and productivity. Rather, we propose that AI creates conditions under which efficiency gains, workload reabsorption, self-calibration, and eventual expectation escalation may become connected. Whether these conditions develop into an availability ratchet may depend on organizational context, including how AI-assisted output is observed and evaluated. The contrast between institutionally scheduled work and roles with more observable, time-sensitive deliverables is therefore treated as a proposition for future research rather than an established boundary condition. This study provides a first qualitative, theory-building examination of these dynamics.

Method

Design, sampling, and participants

We conducted semi-structured interviews with eleven working mothers in knowledge-intensive roles in organizations where generative AI is used in daily work. Participants were recruited through convenience sampling, drawn in part from the authors' professional networks. Three inclusion criteria were applied: (a) self-identification as a working mother with caregiving responsibilities, with recruitment prioritizing mothers of young children; (b) employment in a knowledge- intensive role, operationalized as work consisting primarily of cognitive, administrative, analytical, or communicative tasks; and (c) use of generative AI tools in one's work. Each participant received a $10 Amazon gift card.

The sample spanned K-12 and higher education, healthcare, insurance, software and technology, and nonprofit sectors, and included managerial and individual-contributor roles (Table 1). Participants used general-purpose generative AI tools such as ChatGPT, Claude, and school-approved platforms, through both organizationally sanctioned and individually adopted channels. Work arrangement, employment classification, and demographic characteristics relevant to the motherhood penalty (race, age, partner status, caregiving support) were not systematically collected. We return to these omissions in the limitations.

Because participants were recruited partly through professional networks and all participants were active users of generative AI, the sample may overrepresent positive experiences with AI use and underrepresent the perspectives of reluctant, infrequent, or non-users. This consideration should be taken into account when interpreting the findings.

The study received an IRB exemption determination at the authors' institution prior to data collection. Informed consent followed a standardized script covering confidentiality, de-identification, recording permission, and the right to pause, decline questions, or withdraw without consequence.

Interview protocol

Interviews were conducted in May 2026, lasted approximately 30 to 60 minutes, and were audio-recorded with participant permission and transcribed. The standardized semi-structured protocol comprised eighteen questions in a warm-up block and six topical blocks aligned with the framework: AI use and workflow integration; the resource pathway (temporal control, interruption recovery); the ratchet pathway (availability and responsiveness expectations); caregiving, motherhood, and the ideal worker norm; organizational boundary protections; and reflective, model-refinement questions. The protocol included neutral purpose framing stating explicitly that questions about AI's benefits and about pressure or expectations “matter equally to this research,” reducing demand characteristics in either direction.

Data and analysis

The data consist of interview transcripts, organized question-by-question against the protocol for each participant. Transcripts were reviewed in two ways: participants validated the transcript of their own interview, providing a member check, and a second researcher reviewed the full set for completeness and fidelity. This transcript verification process supported data quality but did not constitute analytical member checking of themes or interpretations. Several questions in some interviews have no recorded response, and we treat missing responses as missing rather than as negative evidence.

Analysis followed an abductive, iterative coding approach informed by the study's preliminary dual-pathway framework and guided by the Gioia methodology (Gioia et al., 2013). Transcript statements were coded into first-order concepts retaining participant language, abstracted into second-order themes reflecting researcher interpretation, and aggregated into six broad dimensions. The initial framework informed both the interview protocol and sensitized coding, while the analysis also refined the framework through distinctions that emerged from the data, particularly among workload reabsorption, self- calibration, pre-existing availability pressure, and explicit expectation escalation. Coding was performed by the first author and reviewed by the second, with disagreements resolved through consensus; no quantitative intercoder reliability statistic was computed. Figure 1 below presents the resulting data structure.

Findings

Coding produced six aggregate dimensions (Figure 1): AI as a workplace resource; recovered time becomes more work; the persistence of the always-on culture; motherhood and the ideal worker norm; weak organizational boundary protections; and AI as both opportunity and pressure. The first dimension corresponds to the resource pathway; the second, third, and sixth illuminate how and where the ratchet pathway is, and is not yet, emerging; the fourth and fifth establish the gendered and institutional conditions under which these dynamics unfold. We present each in turn.

To keep the pathways analytically distinct, we differentiate four related but separate dynamics that the interviews surface: (1) always-on availability pressure that predates AI, rooted in communication technology, workplace culture, and role responsibility (Section 4.3); (2) AI-enabled time savings that improve efficiency, workflow control, and interruption recovery (Section 4.1); (3) workload reabsorption, in which saved time is converted into additional work without any explicit demand (Section 4.2); and (4) AI-specific expectation escalation, in which managers, peers, or workers themselves come to expect faster output because AI exists (Section 4.6). The evidence is strong and consistent for the first three dynamics but limited and uneven for the fourth, a calibration we carry through the findings and discussion. Short quotations are verbatim participant phrases from the transcripts, as preserved in first-order coding (Figure 1). The six dimensions map onto the research questions as follows: Sections 4.1 and 4.2 address RQ1; Sections 4.3 and 4.6 address RQ2; Section 4.4 addresses RQ3; and Section 4.5 addresses RQ4.

Table 1. Participant overview

ID

Role Sector Reported AI use Children (as reported)
P1 STEAM teacher / technology support

K-12 education

Extensive (“a lot”)

Older children
P2

Executive director, science & instructional technology

K-12 education (district)

Frequent

Older children
P3

Human resources manager

Healthcare

Several times per week

Elementary-school age
P4 Project manager

Software

Daily

Two children under eight

 
P5

High school English teacher

K-12 education

Several times per week

Teenagers

 
P6

Financial analyst

Insurance

Daily

Preschool-age child

 
P7

Marketing director

Nonprofit

Almost daily

Two school-age children

 
P8

Project manager

Technology

Several times per week

Two middle-school children
P9

Elementary school principal

K-12 education

A few times per week

Older children
P10

Financial analyst

Healthcare

 Almost daily Young children
P11

University administrator

Higher education

Several times per week

Teenagers
AI as a workplace resource

Evidence for the resource pathway was strong and consistent across all eleven interviews. Participants described AI as materially reducing the time cost of routine cognitive work, including drafting communications, summarizing meetings and reports, preparing lesson plans and presentations, and organizing or analyzing data. The most vivid account came from P2, a district technology executive, who called AI “a lifesaver,” described using an AI assistant to process roughly thirty spreadsheets of student data, saving an estimated five hours, and said she “can't imagine” doing her job without AI.

Beyond raw efficiency, participants emphasized workflow control and interruption recovery, the mechanism most directly relevant to caregiving because the interruptions in question are often caregiving itself. P3, an HR manager, reported that when family responsibilities interrupt her, AI “makes it easier to return to interrupted work,” shortening the path back into a task by generating drafts and summaries to resume from. Similar experiences were reported across the sample, with participants describing faster recovery from interruptions, improved ability to stay on top of deadlines, and greater capacity to complete work within existing time constraints. P1, a teacher, explained that AI lets her complete more work during school hours, reducing what she takes home, a direct conversion of efficiency into protected family time.

Participants also reported reduced stress because AI “makes large tasks less overwhelming” and “reduces time spent on repetitive work.” Several participants described feeling less stressed and better able to manage demanding workloads. Notably, some framed the benefit in explicitly maternal terms: P2 emphasized that AI is exceptionally beneficial “particularly for working mothers,” and extended its use to family logistics such as trip planning and budgeting.

Recovered time becomes more work

If the resource pathway describes what AI gives, this dimension describes where the gains go: across effectively the entire sample, saved time was reabsorbed into additional work rather than returned to the worker as personal time. Participants consistently reported redirecting recovered time toward additional projects, support activities, strategic work, grading, analysis, and other role-related responsibilities. As P3 explained, saved time is “quickly absorbed by other responsibilities” because HR work is never fully finished. Similarly, P8 noted that there is always something else requiring attention, so the time is “rarely truly free.” Participants' own words were blunt: “workload doesn't decrease”; “time savings rarely become personal time”; “AI increases capacity, but not free time.” What becomes “more work,” in these accounts, is thus not externally assigned overtime but an expansion of scope within the role: more projects, more support given to others, and deeper versions of existing tasks. P1's case is the partial exception that proves the rule: her gains stayed within the workday but were redirected to tasks that would otherwise go unfinished, with the benefit realized at home only as less work brought there. The pattern is theoretically consequential: no participant described a managerial decision to claim the recovered time; reabsorption was experienced as the natural consequence of unbounded workloads (“there's always more work,” as the first-order coding summarized it). Work was commonly described as expanding to fill the time AI made available, without participants reporting explicit managerial demands for that additional work.

The persistence of the always-on culture

Pressure to remain available was nearly universal, but participants largely described it as predating AI, rooted in communication technology, workplace culture, and role responsibility. P1 reported frequent pressure to respond outside work hours, saying “phones make it hard to disconnect,” and noted that her boss sometimes responds to messages as late as 2:00 AM. “I check email after my kids go to bed,” P3 reported, and “I feel guilty if messages go unanswered.” P2, in her first leadership year, regularly checks communications on weekends out of a sense of responsibility, while P9 similarly linked responsiveness to the demands of the principalship.

Notably, participants located the source of this pressure away from AI. P1, asked directly whether AI had shifted response-time expectations, said it had not, blaming instead the constant accessibility phones give work communication. P2 stressed that her weekend responsiveness is self-imposed, flowing from responsibility felt in a new leadership role, while P9 tied the pressure to the principalship itself. The always-on culture was narrated as older than and separate from AI. Participants also connected responsiveness to evaluation, describing availability as a signal of commitment. Several believed leaders are expected to remain available after hours and that employees who maintain stronger boundaries may be viewed as less dedicated. Responsiveness, in short, already functioned as a measure of commitment before AI entered these workplaces, a normative substrate onto which any AI-driven recalibration would be grafted.

Motherhood and the ideal worker norm

Caregiving shaped both participants' workdays and their impression management. Mothers of younger children described workdays organized around school schedules, pickups, appointments, illnesses, and closures (“my schedule revolves around my kids”; P3, P4, P6, P7, P8, P10); mothers of older children recalled how much harder the balance had been at earlier stages (P1, P2, P9, P11). Ten of the eleven participants answered affirmatively when asked whether becoming a parent required working harder to prove commitment, although P1 reframed the demand as having to work differently rather than simply more, citing shortened lunches and compressed, maximally efficient school days.

Participants were concrete about what proving commitment looks like. P8 feels she must be especially organized and productive so that nobody questions her commitment; P6 described pressure to demonstrate reliability despite caregiving responsibilities; P11 recalled feeling, earlier in her career, that she had to show family responsibilities would not affect her
performance; and P7, believing mothers are judged on availability, compensates by staying deliberately organized and responsive.

Participants perceived the commitment burden as gendered. P3, P7, and P10 each stated that mothers are judged more on their availability or responsiveness than fathers; P5 believed mothers face greater scrutiny than fathers. P2 offered a structural illustration of “women stay home when kids are sick”: many male colleagues accumulated unused sick days and received retirement payouts for them, while many female employees had few or none left because they used them for family caregiving, a concrete mechanism converting caregiving into financial and reputational disadvantage.

Consistent with ideal worker pressures, most participants described strategically managing the visibility of family responsibilities: scheduling meetings around family obligations without explaining why (P3); “I block time without explaining” (P4, P7); omitting details when rescheduling (P8); “I don't disclose childcare interruptions” (P6, P10); and, earlier in their careers, avoiding mention of family obligations altogether (P9, P11). Only P5 reported never hiding family responsibilities. These concealment strategies were described as ongoing practice, evidence that the ideal worker norm remains operative.

Weak organizational boundary protections

No participant reported a formal policy governing after-hours availability, response times, or protected offline time. Where boundary-protective behavior existed, it was individual and discretionary: P2, as a manager, avoids emailing principals after roughly 4:00 or 5:00 PM and uses scheduled sending so messages arrive at 7:00 AM, while noting her own weekend work is self-imposed. Participants repeatedly emphasized that culture overrides formal rules (“culture matters more than policy,” “employees follow what leaders do,” “written rules don't hold”): P3 said cultural expectations matter more than written policies; P7 and P9 said employees attend to what leaders do; P8, P10, and P11 likewise located real expectations in culture and manager behavior rather than policy. Asked what they would change, the most common answers were clearer expectations or stronger boundaries around after-hours communication (P3, P6, P7, P8, P10) and leaders modeling healthier work-life boundaries (P11). The vacuum is consequential: if availability expectations ratchet upward, no formal mechanism in any of these organizations would arrest the process.

AI as both opportunity and pressure: early signs of a possible ratchet

Evidence for the fourth dynamic, AI-specific expectation escalation, was the weakest, and we present it as early signs of a possible ratchet rather than an established pattern. Only one participant reported a concrete escalation attributable to AI: P6, a financial analyst, said leadership now expects faster turnaround on reports, such that “faster turnaround has become the norm,” and that AI lowers stress on repetitive tasks while increasing productivity expectations. More commonly, participants described a diffuse, unattributed recalibration, an “assumption that work should be faster now” and “higher expectations because AI exists”: P3 sensed an increasing assumption that work can be completed more quickly; P7 felt a growing assumption that some tasks should take less time than
before; P10 perceived a growing assumption that reports can be completed faster; P4 believed faster communication has become the norm; P8 thought people expect projects to move somewhat faster; P11 felt more pressure to be efficient. The negative cases are analytically informative rather than mere exceptions. P1 and P9 reported no change in expectations and P5 reported none officially; all three work in K-12 schools, where the tempo of work is set institutionally by bell schedules, school calendars, and instructional contact time rather than by managed deliverable turnaround. Their AI use is also configured differently: individually adopted for self-directed preparation (lesson plans, rubrics, announcements, parent letters) whose speed is not visible to or measured by a supervisor. The contrast with the affirmative cases is sharp: the clearest escalation reports came from roles whose AI-assisted outputs are deliverables with manager- or client-facing turnaround: financial reports (P6, P10) and project pace (P4, P8). One possible interpretation is that expectation escalation may be more likely where AI compresses production whose timing is organizationally observable and evaluated than in contexts where work rhythm is institutionally scheduled and AI output primarily supports private preparation. We offer this interpretation, grounded in the sample's sectoral pattern, for testing in the subsequent field study.

Two further observations refine the ratchet construct, while remaining consistent with its still-emergent status. First, the recalibration was frequently self-directed: “pressure shifts to myself,” “personal standards increase.” P3 reported that AI lowers her stress but may raise the expectations she places on herself, and proposed that future research ask employees whether AI changes self-imposed expectations. Second, the ratchet can operate laterally and downward as well as from above: P2, herself a manager, acknowledged holding others to a higher standard because AI is available, recalling thinking, “Why don't you just use AI to do it?” This is the ratchet seen from the enactor's side, escalation framed not as a new demand but as a natural inference from technological capability.

Finally, asked to step back, participants captured the predicted duality: P3 said AI gives room and takes it away at once, saving time while creating pressure to accomplish more; P6 and P10 answered both, P10 noting pressure to do more because of AI; P7 said AI can make people feel they should get more done. The rest judged AI as mostly giving more room. The modal experience is genuine resource gain accompanied by concerns that expectations and output standards may gradually rise over time: opportunity and pressure at once.

Discussion

A pre-ratchet stage: incipient and internalized

The central theoretical refinement concerns the form and calibration of the availability ratchet. The framework anticipated managerial escalation of responsiveness expectations justified by AI capability. The interviews do not show that escalation as an established pattern; instead, they suggest what may be an earlier stage of the process. Expectations of speed rise diffusely, without explicit mandate or attribution, and workers, particularly mothers already accustomed to proving commitment, may convert technological capability into self-imposed obligation (P3, P2). AI-specific expectation escalation remained limited and uneven, observed concretely in only one case.

We therefore tentatively propose that, where an availability ratchet develops, it may begin as an internalized process in which workers absorb time savings into additional work and raise their own expectations in response to AI. Because this study captures experiences at a single point in time, we cannot determine whether these processes eventually become explicit managerial expectations. Instead, the findings identify conditions that may contribute to such a process and that warrant future study. This sequencing matters for theory and intervention alike: if it holds, by the time a ratchet is explicit it may merely formalize a baseline that workers have already established through their own behavior.

The resource pathway is real, and its gains are captured

The data provide unambiguous support for the resource pathway, including its most caregiving-relevant mechanism, interruption recovery: for mothers whose days are punctured by school calls, illnesses, and appointments, resuming interrupted work quickly is not a marginal convenience but a structural enabler of performance (Sections 4.1, 4.4). Yet the near-universal reabsorption of recovered time implies that, under current governance, the pathway's gains accrue largely to the organization as output rather than to workers as personal time. P1's experience of completing work at school so less follows her home shows worker-captured gains are possible; the rest of the sample shows why they are rare: absent formal boundary protections, no mechanism earmarks efficiency gains for the worker.

A normative substrate awaiting recalibration: the IS contribution

The findings on the always-on culture and the ideal worker norm clarify why the distributional stakes of AI deployment are gendered. Responsiveness already functions as a commitment signal in these organizations; mothers already report being judged on availability more than fathers and often conceal caregiving responsibilities to manage those judgments. AI introduces a new interpretive resource: if AI is assumed to make work faster, then delays and protected offline time may become more difficult to justify. While this inference was evident among some participants, it was not yet institutionalized and was absent in several workplaces. This extends boundary management theory (Clark, 2000; Kossek & Lautsch, 2012) by suggesting that intelligent tools can alter the evaluative meaning of boundary-protecting behavior.

For IS research, the contribution is twofold. First, the findings locate the boundary-management challenge of generative AI in expectation recalibration rather than reachability: unlike email and smartphones, which expanded when workers could be contacted, AI changes assumptions about how quickly work can be completed. Second, the findings demonstrate interpretive contingency. The same AI affordances were experienced as flexibility in some settings and as emerging intensification pressures in others, with no formal governance steering the outcome. Whether AI functions as a resource or contributes to a ratchet therefore depends on organizational interpretation and governance rather than the technology alone.

Implications for practice

First, because recovered time is reabsorbed by default, organizations intending AI as a flexibility resource must actively earmark gains, for example through workload review at deployment. Second, the governance vacuum suggests pairing AI rollout with explicit availability norms (response-time expectations, protected offline hours, after-hours guidance), precisely what participants most frequently requested. Third, because leader behavior outweighs written policy, protections are unlikely to hold unless modeled by leadership; P2's scheduled-send practice offers a template.

Taken together, these recommendations suggest that AI governance should extend beyond technical oversight of AI systems to include the governance of expectations surrounding AI-enabled productivity, availability, and workload allocation.

Limitations and future research

Several limitations bound these findings. First, the sample comprises eleven working mothers; the design's comparison groups (fathers, non-mothers, managers) are not yet represented, so claims about gendered differences rest on participants' perceptions, and the sample falls short of the broader program's saturation target. Second, some questions in some interviews lack recorded responses, which we treat as missing rather than as negative evidence. Third, the sample is concentrated in education and adjacent public-sector settings; the sectoral pattern noted in Section 4.6 is a lead worth pursuing but limits generalization. Fourth, the design is cross-sectional and retrospective; a possible incipient ratchet is precisely the kind of process that requires longitudinal observation to confirm or disconfirm, and the present data cannot rule out that the diffuse expectations participants perceived reflect general workplace intensification rather than AI specifically. Fifth, the convenience sample was drawn partly from the authors' professional networks, participants self-selected into a study about AI and work, and most were relatively enthusiastic AI users. Sixth, race, age, partner status, caregiving support, work arrangement, and employment classification were not collected; the resulting inability to analyze gender intersectionally is the most consequential of these omissions, and intersectional extension is a priority for the field study, since the costs of an availability ratchet are unlikely to fall evenly among mothers.

These limitations map onto the next phases of the research program: a multi-wave field study testing the dual pathways and their moderation by availability norms and boundary protections, and a managerial vignette experiment testing whether identical AI-enabled boundary-protecting behavior is evaluated more negatively when enacted by mothers under productivity-multiplier framings of AI. The finding sharpen those designs: measurement should capture self-imposed as well as perceived-external expectations, and workload reabsorption should be modeled as distinct from responsiveness escalation.

Conclusion

This study asked whether AI-enabled flexibility tools reduce the motherhood penalty by increasing temporal control or intensify it by raising norms of continuous availability, responsiveness, and productivity. Interviews with eleven working mothers in knowledge-intensive roles showed strong, consistent evidence for the resource pathway (efficiency gains, workflow control, faster recovery from caregiving interruptions, reduced stress) alongside near-universal reabsorption of recovered time into additional work. Evidence for AI-specific expectation escalation was limited and uneven, indicating not an established availability ratchet but its possible beginnings: a pre-ratchet stage in which expectations rise diffusely and internally, within an always-on culture and a vacuum of formal boundary protections, before becoming explicit managerial demands.

The study contributes to IS research by locating the boundary-management problem generative AI introduces in expectation recalibration rather than reachability, proposing the pre-ratchet stage as a refinement of the availability ratchet construct, and identifying a candidate boundary condition, namely the organizational observability of AI-compressed output. The planned field study and vignette experiment will test these claims and their distributional consequences for mothers. Whether AI eases or modernizes the motherhood penalty may depend less on the technology than on whether organizations govern the baseline it resets.

References

Acker, J. (1990). Hierarchies, jobs, bodies: A theory of gendered organizations. Gender & Society, 4(2), 139–158. https://doi.org/10.1177/089124390004002002

Ashforth, B. E., Kreiner, G. E., & Fugate, M. (2000). All in a day's work: Boundaries and micro role transitions. Academy of Management Review, 25(3), 472–491. https://doi.org/10.5465/AMR.2000.3363315

Bick, A., Blandin, A., & Deming, D. J. (2024). The rapid adoption of generative AI (NBER Working Paper No. 32966). National Bureau of Economic Research. https://doi.org/10.3386/w32966

Brynjolfsson, E., Li, D., & Raymond, L. R. (2025). Generative AI at work. The Quarterly Journal of Economics, 140(2), 889–942. https://doi.org/10.1093/qje/qjae044

Budig, M. J., & England, P. (2001). The wage penalty for motherhood. American Sociological Review, 66(2), 204–225. https://doi.org/10.2307/2657415

Chung, H. (2022). The flexibility paradox: Why flexible working leads to (self-)exploitation. Policy Press.

Clark, S. C. (2000). Work/family border theory: A new theory of work/family balance. Human Relations, 53(6), 747–770. https://doi.org/10.1177/0018726700536001

Collett, C., Gomes, L. G., & Neff, G. (2022). The effects of AI on the working lives of women. UNESCO Publishing.

Correll, S. J., Benard, S., & Paik, I. (2007). Getting a job: Is there a motherhood penalty? American Journal of Sociology, 112(5), 1297–1338. https://doi.org/10.1086/511799

Dillon, E. W., Jaffe, S., Immorlica, N., & Stanton, C. T. (2025). Shifting work patterns with generative AI (NBER Working Paper No. 33795). National Bureau of Economic Research. https://doi.org/10.3386/w33795

Gmyrek, P., Berg, J., Kamiński, K., Konopczyński, F., Ładna, A., Nafradi, B., Rosłaniec, K., & Troszyński, M. (2025). Generative AI and jobs: A refined global index of occupational exposure (ILO Working Paper No. 140). International Labour Organization. https://doi.org/10.54394/HETP0387

Gioia, D. A., Corley, K. G., & Hamilton, A. L. (2013). Seeking qualitative rigor in inductive research: Notes on the Gioia methodology. Organizational Research Methods, 16(1), 15–31. https://doi.org/10.1177/1094428112452151

Humlum, A., & Vestergaard, E. (2025). Large language models, small labor market effects (NBER Working Paper No. 33777). National Bureau of Economic Research. https://doi.org/10.3386/w33777

Jiang, W., Park, J., Xiao, R. J., & Zhang, S. (2025). AI and the extended workday: Productivity, contracting efficiency, and distribution of rents (NBER Working Paper No. 33536). National Bureau of Economic Research. https://doi.org/10.3386/w33536

Kossek, E. E., & Lautsch, B. A. (2012). Work–family boundary management styles in organizations: A cross-level model. Organizational Psychology Review, 2(2), 152–171. https://doi.org/10.1177/2041386611436264

Mennega, N., & de Villiers, C. (2021). A quarter century of gender and information systems research: The role of theory in investigating the gender imbalance. Gender, Technology and Development, 25(1), 112–130. https://doi.org/10.1080/09718524.2020.1859712

Pang, M.-S., Kankanhalli, A., Aanestad, M., Ram, S., & Maruping, L. M. (2024). Digital technologies and the advancement of social justice: A framework and agenda. MIS Quarterly, 48(4), 1591–1610. https://aisel.aisnet.org/misq/vol48/iss4/6

Tarafdar, M., Cooper, C. L., & Stich, J.-F. (2019). The technostress trifecta: Techno eustress, techno distress and design. Theoretical directions and an agenda for research. Information Systems Journal, 29(1), 6–42. https://doi.org/10.1111/isj.12169

Trauth, E. M. (2023). An overview of the individual differences theory of gender and IT. In E. M. Trauth & J. L. Quesenberry (Eds.), Handbook of Gender and Technology (pp. 22–54). Edward Elgar Publishing.

Zanhour, M., & Sumpter, D. M. (2024). The entrenchment of the ideal worker norm during the COVID-19 pandemic: Evidence from working mothers in the United States. Gender, Work & Organization, 31(2), 625–643. https://doi.org/10.1111/gwao.12885

 

For Further Discussion

This serves as an overview of the project and does not include the complete work. To further discuss this project, please email Emily McNeill.

Quinnipiac University Interdisciplinary Program for Research and Scholarship

Open to students of all majors, QUIP-RS provides up to $5,000 in funding for undergraduate students to conduct research or complete creative projects alongside faculty mentors. This intensive 8-week program enables students to develop scholarly skills while encouraging discussion about successes and shortcomings with fellows and mentors.

Learn more about QUIP-RS

Explore Our Areas of Interest

We've sorted each of our undergraduate, graduate and doctoral programs into unique Areas of Interest. Explore these categories to discover which programs and delivery methods best align with your educational and career goals.

Explore Business and Finance at Quinnipiac

Explore Social Sciences at Quinnipiac