This article shares insights from Elie Ofek, Malcolm P. McNair Professor of Marketing, 性视界 Business School and Julian De Freitas, Assistant Professor of Business Administration, 性视界 Business School who are pursuing research on the topics of artificial intelligence and organizations.

1. What drew you to this area of research and how did you first become involved in this work?

Our involvement began through conversations with marketing managers at the 性视界 Art Museums, who were exploring innovative ways to bring their collections to life and were open to using new technologies. This collaboration sparked the idea of using AI avatars to animate historical portraits and measure their impact in real-world settings.

We were naturally drawn to this idea by our interest in how emerging AI technologies鈥攅specially generative and interactive AI鈥攃an transform the way organizations engage with people. While personalization and automation have been studied, there is little empirical work on whether and how embodied, humanlike AI avatars鈥攁 new technology on the cutting edge of AI development鈥攃an foster more engaging interactions between consumers brands. 

2. What are some common misconceptions or barriers around the problem you鈥檙e working to solve?

A limited way to think about AI is that it is primarily a tool for efficiency or information delivery鈥攅ssentially a faster, cheaper way to push content. Yet this perspective does not consider how people naturally perceive these systems. And while use of generative AI for companionship is beginning to be appreciated, many managers under appreciate the technology鈥檚 potential for two-way, emotionally rich interaction that can be relevant for their business/organization. Our research directly explores this potential source of business value in a field setting. 

A barrier is skepticism: organizations worry about handing over control to a system that responds autonomously on their behalf, in case the AI鈥檚 behavior feels too artificial, inauthentic to the brand, or misaligned with its values. Thus, in implementing our research project, we have also needed to take care to develop safeguards鈥攕uch as content moderation, tone calibration, and strict alignment with institutional values鈥攖o ensure that we are protecting the brands of our field partners. Another source of skepticism is whether other modalities鈥攕uch as just an image or a video, which are one-way information provision, is sufficient or even superior to allowing consumers to interact with a brand, particularly when the agent is an AI avatar. Our research aims to examine this very issue and tease out the effect of different modalities.

3. What research is being done on this topic and how is your approach or perspective unique?

There is emerging research on personalization in marketing, human-computer interaction, and parasocial relationships, but most studies are conceptual or run in highly controlled lab settings, and do not focus on the potential of embodied, interactive AI avatars per se. Our approach is unique in three ways:

鈥�&苍产蝉辫;Real-world field experiments: we are testing AI avatars in live campaigns with the 性视界 Art Museums, measuring actual engagement, conversion, and visit behavior.

鈥�&苍产蝉辫;Systematic variation in interactivity: we directly compare static images, one-way scripted avatars, and two-way conversational avatars to isolate the effects of interactivity.

鈥�&苍产蝉辫;Psychological mechanisms: we examine the role of psychological processes in our effects鈥攕uch as perceived intimacy, relational intent, and status dynamics鈥攂ringing a social-psychological lens to AI design in marketing and cultural engagement.

4. What excites you most about this work and its potential impact?

We are most excited about the possibility of moving digital engagement from transactional clicks to meaningful, human-centered interactions. If successful, this work could provide a blueprint for cultural institutions, brands, and nonprofits to create experiences that deepen connection, inspire action, and expand access. For instance, the idea that a historical portrait in a museum could 鈥渢alk鈥� with a visitor鈥攁nd that this could spark curiosity, learning, and even a museum visit鈥攊s both academically fascinating and socially impactful, suggesting a fundamentally new way to engage new generations while sustaining important cultural institutions. 

5. How do you hope working with D^3 will amplify the impact of your work?

The essential infusion of funds from D^3 is allowing us to turbocharge this challenging research program, reaching insights that otherwise would not have been possible or may have come too late to meaningfully inform managerial practice. Through its network, resources, and convening power, D^3 then offers a unique platform to translate our insights into practice, as well as spark new collaborations that could scale these insights well beyond our initial museum partnership. 

6. What changes do you hope to see in your field as a result of the work being done in this area?

We hope to see a shift from viewing AI as a cost-saving novelty toward seeing it as a relational tool鈥攐ne that can extend human-brand connections beyond what has previously been possible. We hope the findings will suggest a new vision of brand engagement, in which AI-powered brands listen, adapt, and co-create meaning with customers; rather than just focus on one-way modalities. Furthermore, we believe these findings can inform managerial models of how AI integration efforts should merge with customer relationship and brand management efforts鈥攁n area that is sorely in need of empirically-informed conceptual frameworks. For some initial ideas, see: 

Finally, we hope to see more industry-academic partnerships conducting field-based, ecologically valid studies that address these questions, while still being mindful of how to do so in a sustainable manner that protects long-term customer and brand assets, since many company鈥檚 AI efforts have been failing and there is some apprehension among brands. For a number of analyzed failure examples, see 

7. What鈥檚 an essential area in which AI and digital technologies will reshape the way businesses or society operate in the long run that we may not be considering?

People treat today鈥檚 highly capable chatbots much more like they do other human beings, than they do other (non-living) technologies. We believe this has profound implications for how brands are engaging customers that are still underappreciated. 

To provide just one example, one of us has found that users of so-called 鈥淎I companion鈥� apps often say farewell before logging off the app rather than simply quitting the app. These apps leverage this moment by employing 鈥渆motionally manipulative tactics鈥�, like making the user feel guilty for leaving, in order to prevent users from logging off at this point. These tactics work, increasing the number of messages users send and how long they stay on the app beyond the point they said farewell (relative to when the app simply says goodbye in turn, without using emotional manipulation). Notice that such tactics are not simply increasing engagement by how they recommend content to users, but by capitalizing on our social and emotional instincts. For more details, see: 

The D^3 Associates Program supports and accelerates faculty research into the ways AI and digital technologies are reshaping companies, organizations, society, and practice.

The post D^3 Associates Spotlight Series: Elie Ofek and Julian De Freitas appeared first on HBS AI Institute.

This article shares insights from Dr. Livia Alfonsi, Assistant Professor of Business Administration at 性视界 Business School, whose research studies labor markets and the transition from school to work, with a focus on how to help young workers find strong job matches and build early-career trajectories.

1. What drew you to this area of research and how did you first become involved in this work?

I spend most of my time studying labor markets in the Global South, particularly in Sub-Saharan Africa and South Asia, where historically large cohorts of young adults are entering the labor market with higher education and higher expectations than previous generations. Over the next decade, more than a billion young people will reach working age in developing countries. Yet in many settings, job creation has not kept pace, and competition for stable formal work is intense.

In these environments, the early years of a career can be fragile. When early job search efforts lead to low pay, unstable work, or repeated rejection, young workers can become discouraged and pull back from active search. Some drift into casual work, subsistence activities, or inactivity. That discouragement is not just a personal experience. It can translate into underutilized human capital at scale, even when education and training investments have risen dramatically.  

In prior research, I studied mentorship interventions that can help young jobseekers persist through uncertainty, recalibrate expectations, and navigate early setbacks. Those programs can be powerful complements to education and training. At the same time, job search is increasingly mediated through digital platforms, especially for younger cohorts. This led me to a new question: can conversational AI be designed to provide some of the continuity, encouragement, and practical guidance that supports persistence, complementing human sources of support when they are available, and extending access to that kind of guidance when they are not? Partnering with Rozee.pk, Pakistan鈥檚 largest job platform, and its AI career buddy, Rozeena, we are testing how conversational tone, message content, and memory callback features shape engagement and job search behavior at scale, in a setting where an AI mentor can support users across the process, from identifying opportunities and understanding the labor market, to practical steps like CV building and interview preparation, and to encouragement and follow-through during setbacks.

2. What are some common misconceptions or barriers around the problem you鈥檙e working to solve?

One common misconception is that job search is mainly an information problem. If young people just had better data about wages, labor demand, or application strategies, outcomes would improve. Another is that networks are the whole story. Networks do matter a lot, but their value goes beyond access to job leads. What’s often missing is sustained support鈥攕omeone to normalize setbacks, provide encouragement, and model what persistence looks like when early efforts don’t pay off immediately. This matters because the challenge is rarely about a single decision. It’s about maintaining momentum over weeks or months when feedback is scarce, rejections are common, and progress is hard to see. In reality, young workers face a twofold gap. First, they lack practical guidance about which steps are most effective at different stages of job search or career growth. Second, they struggle to sustain motivation and adapt strategy over time, especially when early efforts lead to silence or rejection. Both problems need to be addressed together. Information alone doesn’t help if you’re too discouraged to act on it. And encouragement rings hollow if it’s not paired with actionable guidance. That’s why studying how support is delivered鈥攊ncluding tone, timing, and continuity鈥攊s just as important as studying what information is shared.

3. What research is being done on this topic and how is your approach or perspective unique?

There’s growing research on AI in hiring and labor markets, often focused on screening, matching efficiency, or bias. There’s also a rich body of behavioral and psychological research on discouragement, belief formation, and how people respond to feedback during job search. But we still know very little about how to leverage AI systems that people actually interact with, including conversational tone, memory, and continuity, to foster persistence and improve decisions over time. 

Our approach is distinctive in three ways. First, we’re studying these questions at scale, in a real labor market. Working with Pakistan’s largest job platform, we evaluate randomized tests implemented within the product experience that vary how guidance is delivered and what the system remembers from past interactions, all embedded directly in the product experience. Second, we can link conversational interactions to extraordinarily rich administrative data: two decades of platform history, including job postings, applications, hires, and wage trajectories, alongside AI-era conversational logs. This lets us study not just immediate responses, like whether someone clicks on a job ad, but also longer-term shifts in search behavior, match quality, and labor market outcomes. Third, we’re testing communication strategies in addition to information provision, studying whether empathic framing or motivational language helps workers engage with advice, feel understood, and stay active through setbacks. Finally, we see this project as generating evidence that travels beyond AI. By treating conversational AI as a disciplined testbed, we can identify which communication strategies help young workers persist, whether the guidance comes from an AI agent, a mentor, or a career counselor. Those lessons can inform how support providers design interventions that are more credible, more motivating, and more effective.

4. What excites you most about this work and its potential impact?

What excites me most is the possibility of designing digital systems that reinforce agency rather than undermine it. Job search already feels opaque and discouraging to many young workers. There’s a real risk that technology makes this worse: more automated rejections, less human feedback, even less sense of progress. But conversational AI, designed thoughtfully, offers something different: personalization at scale. In many low- and middle-income settings, access to career guidance is uneven and formal support systems are limited. A tool that reaches people through WhatsApp, in local languages, with low friction, can meet workers where they already are. WhatsApp is part of daily life for billions of people, which means this model can, in principle, deliver guidance at a scale that traditional programs never could. The question is whether the design choices we make (empathic language versus neutral facts, recalling past conversations versus treating each interaction as new, proactive follow-up versus waiting for users to return) actually matter for outcomes. If they do, it means platform designers have real leverage to shape not just match efficiency, but worker persistence, confidence, and long-term trajectories. 

5. How do you hope working with D3 will amplify the impact of your work?

I’m grateful for D^3’s support, and I’m especially excited about joining a community that’s thinking rigorously about how AI and data are reshaping organizations and markets. What I value most about this collaboration is the chance to have a structured space to share early findings, stress-test interpretations, and learn from others tackling similar challenges across different domains. It will help ensure this project generates insights that are rigorous, actionable, and useful beyond this single context. That kind of feedback is especially helpful for a project like mine, which sits at the intersection of labor economics, behavioral science, and AI product design, and it depends on close, iterative collaboration with an industry partner, Rozee.pk. D^3’s ecosystem is invaluable because I can learn from parallel efforts across domains. The structured feedback loops, workshops, and cross-disciplinary conversations D^3 enables are especially helpful for a project that’s fundamentally about translating research insights into better platform design.

6. What changes do you hope to see in your field as a result of the work being done in this area?

I hope we move beyond thinking of digital labor platforms as static information boards and instead treat them as systems that shape how people persist, learn, and decide over time. In practice, that means taking seriously that the delivery of guidance, including tone, timing, continuity, and what the system recalls from prior interactions, can influence whether workers stay engaged in the labor market or withdraw after setbacks.

More concretely, I hope the field develops evidence-based principles for how to communicate difficult but important messages in a way that keeps workers moving forward. Many labor markets are changing quickly. Some career paths are becoming flatter, some skills are depreciating faster, and many workers will need to reskill or adjust expectations. A central challenge is not only identifying these shifts, but communicating them in a way that preserves motivation and agency. For example, how do we provide realistic feedback about prospects while still helping people take the next constructive step, whether that step is adjusting search strategy, pursuing training, or pivoting to a nearby occupation?

Finally, I hope research in this area broadens the set of outcomes and design questions we study. In addition to questions about matching, efficiency, and fairness, we should also ask how AI systems shape motivation, expectations, and follow-through, especially for young adults navigating uncertainty and groups that face barriers to opportunity. If AI is going to play a role in career guidance, we cannot forget the humanity of the user. Last, the most effective support may not look the same for everyone. It may need to adapt to different personalities, circumstances, and moments, offering encouragement when confidence is low, structure when someone feels stuck, and practical feedback when someone is ready to take action.

7. What鈥檚 an essential area in which AI and digital technologies will reshape the way businesses or society operate in the long run that we may not be considering?

One underappreciated shift is that AI is becoming part of the advice and feedback ecosystem that shapes how people make high-stakes decisions, especially career decisions. Over the years, we’ve moved from seeking guidance primarily from other people, to relying on search engines or online communities. Conversational AI is becoming the next default: the place people turn after a setback, when they feel stuck, or when they’re deciding whether to persist, pivot, or invest in new skills. As AI becomes embedded in job platforms, workplace tools, and general-purpose assistants, it will fundamentally influence how people interpret feedback, and decide what to do next. Career guidance (once delivered sporadically by mentors or counselors) will increasingly be mediated by systems that respond instantly, repeatedly, and at very low cost.

That’s fundamentally different from human advice, and we’re only beginning to understand the implications. This raises important questions such as; What kinds of advice build agency rather than dependency? How do we design systems that can communicate difficult truths without discouraging users? Designing AI that is not only capable, but responsible, trustworthy, and genuinely supportive in these high-stakes contexts is an essential frontier. It’s about whether we can design systems that help people navigate uncertainty with more confidence, make better-informed decisions, and build stronger long-term trajectories, particularly for workers who lack access to traditional support networks. That’s a design challenge with enormous implications for equity and inclusion in labor markets worldwide.

The D^3 Associates Program supports and accelerates faculty research into the ways AI and digital technologies are reshaping companies, organizations, society, and practice.

The post D^3 Associates Spotlight Series: Dr. Livia Alfonsi appeared first on HBS AI Institute.

This article shares insights from Alex Chan, Assistant Professor of Business Administration in the Negotiation, Organizations & Markets Unit at 性视界 Business School who is pursuing research on the topics of artificial intelligence and organizations.

1. What drew you to this area of research and how did you first become involved in this work?

My background spans both technology and healthcare, which naturally pulled me toward the 鈥渆ngineering鈥� side of economics鈥攎arket design. I became fascinated by how small changes in market rules, incentives, or even information presentation can meaningfully shape human behavior鈥攕ometimes with life-or-death consequences in settings like healthcare and organ allocation.

My interest in AI grew from two directions. On the research side, I worked early on questions around deep learning鈥檚 ability to extract patient preferences and clinically relevant signals from unstructured data like clinical notes. On the applied side, my time in industry deploying AI-enabled healthcare products made the promise鈥攁nd the risk鈥攙ery concrete: technology can match expert performance and save enormous amounts of time, but it also changes how people make decisions and how accountability is assigned. That combination convinced me that one of the next major market design challenges is not just building better AI systems, but integrating AI into human decision-making environments in ways that are robust, incentive-compatible, and ultimately welfare-improving鈥攅specially as we think ahead to more advanced systems.

2. What are some common misconceptions or barriers around the problem you鈥檙e working to solve?

A major misconception is that 鈥渕ore information鈥� automatically leads to better decisions. In the context of Explainable AI (XAI), for instance, many people assume that if you provide an explanation, decision-makers will naturally use it to make fairer, better choices. But in practice, transparency can create strategic discomfort: explanations can reveal biases, conflicts of interest, or decision rules that stakeholders would rather not surface鈥攅specially when there are financial incentives, reputational concerns, or legal exposure at stake.

One barrier, then, is that people may strategically prefer 鈥渂lack-box鈥� systems鈥攏ot because they love opacity, but because opacity can protect them from scrutiny or responsibility. Another barrier is that we often forecast AI鈥檚 societal impact by linearly extrapolating from recent waves of automation. That framing can miss how AI will reshape how preferences are expressed, how trust is formed, and how institutions evolve when cognition, forecasting, and persuasion become more scalable and more delegated to machines.

3. What research is being done on this topic and how is your approach or perspective unique?

A lot of the current research rightly focuses on the technical 鈥渉ow-to鈥� of AI鈥攂uilding more accurate models, improving interpretability methods, and optimizing performance. My perspective is complementary: I treat AI as a participant in a market or organization rather than simply a tool. That means I focus on how AI systems interact with incentives, power, accountability, and human behavior鈥攐ften in ways that aren鈥檛 visible if we only measure technical accuracy.

For example, in my working paper 鈥淧reference for Explanations: The Case of XAI,鈥� I don鈥檛 just ask whether an AI can explain itself鈥擨 ask whether people actually want explanations when real incentives and tradeoffs are present. Using incentivized experiments with real financial stakes helps reveal when transparency is demanded, when it鈥檚 avoided, and why.

More broadly, by combining market design and behavioral economics, I can study how AI decision-support, monitoring, or recommendation systems interact with factors like gender, race, hierarchy, and institutional constraints鈥攄imensions that pure computer science approaches often treat as 鈥渄ownstream鈥� but that frequently determine real-world outcomes. Market design also pushes us to analyze markets that don鈥檛 fully exist yet, which is increasingly important as AI changes what it even means to 鈥減articipate鈥� in a market.

4. What excites you most about this work and its potential impact?

What excites me most is the possibility of moving beyond the idea that AI progress is mainly about better prediction鈥攁nd toward the idea that progress is about better systems. If we design incentives and institutions well, AI can reduce cognitive overload, improve access to expertise, and make high-stakes decisions more consistent and less arbitrary. In healthcare, that can translate into better triage, more equitable access, reduced clinician burnout, and ultimately better patient outcomes.

At the same time, I鈥檓 excited by the intellectual challenge: AI changes the 鈥渞ules of the game鈥� in markets and organizations. We now have decision-makers who can delegate judgment to models, organizations that can scale monitoring and evaluation, and environments where explanations can be demanded, ignored, weaponized, or strategically suppressed. Understanding those dynamics鈥攁nd designing mechanisms that make good outcomes more likely鈥攆eels both urgent and deeply consequential.

5. How do you hope working with D^3 will amplify the impact of your work?

D^3 is an ideal home for this kind of research because it brings together technologists, economists, organizational scholars, and practitioners who are grappling with the same reality from different angles. I see D^3 as a 鈥渢ranslation layer鈥� between theory and deployment: a place where questions about incentives, governance, and real-world adoption can be stress-tested against how organizations actually operate.

I also hope D^3 will amplify impact through its convening power and practitioner ecosystem鈥攈elping connect research insights to real institutional design decisions, from product development and auditing to policy, procurement, and organizational governance. When the goal is not just to understand AI, but to shape how it鈥檚 used responsibly and effectively, that cross-disciplinary and real-world engagement is invaluable.

6. What changes do you hope to see in your field as a result of the work being done in this area?

I hope to see market design become a central lens for thinking about AI, including advanced systems that may begin to act more like autonomous agents in the economy. Rather than relying primarily on after-the-fact regulation or patchwork compliance, I want to see organizations design digital ecosystems from the ground up with incentives that support transparency, productivity, and fairness simultaneously.

In practical terms, that means shifting from 鈥淐an we build this model?鈥� to 鈥淲hat behavior does this system produce once it鈥檚 embedded in an institution with real incentives?鈥� It also means building stronger evidence around what kinds of transparency and accountability mechanisms actually work鈥攏ot just in principle, but in practice.

7. What鈥檚 an essential area in which AI and digital technologies will reshape the way businesses or society operate in the long run that we may not be considering?

One underappreciated shift is that AI won鈥檛 just replace tasks鈥攊t will reshape the institutional infrastructure through which preferences, negotiations, and decisions happen. As personal AI agents become more common鈥攁gents that summarize options, negotiate on our behalf, filter information, and even execute transactions鈥攎arkets may increasingly become 鈥渁gent-to-agent.鈥� That changes what it means to have a preference, how trust is built, and how persuasion and manipulation operate at scale.

This raises foundational design questions:

• How do we represent and protect human preferences when they鈥檙e expressed through intermediating AI systems?
• What new markets and norms emerge when AI can cheaply generate convincing arguments, tailored messaging, or strategic explanations?
• What does accountability look like when decisions are the output of human-AI teams鈥攐r of automated negotiations between agents?

In the long run, the big opportunity (and challenge) is designing the mechanisms鈥攊dentity, provenance, incentives, auditing, governance鈥攖hat make delegation to AI socially beneficial rather than destabilizing. That鈥檚 where market design and institutional thinking become essential.

The D^3 Associates Program supports and accelerates faculty research into the ways AI and digital technologies are reshaping companies, organizations, society, and practice.

The post D^3 Associates Spotlight Series: Alex Chan appeared first on HBS AI Institute.

BOSTON, November 18, 2025 鈥� Faculty at the Digital Data Design Institute at 性视界 today announced the launch of the Frontier Firm AI Initiative, a collaboration with Microsoft and its clients that aims to deepen understanding and accelerate the practice of building Frontier Firms. As defined by the Digital Data Design Institute at 性视界, Frontier Firms are human led, agent operated organizations that put AI at the core of their strategy to transform operations, accelerate innovation, and amplify human capacity. The research into the journey of Frontier Firms will be a catalyst for redefining long-held paradigms of work. Hosted by the Digital Data Design Institute at 性视界, this Initiative will develop applied research on human-AI collaboration, upskill global C-suite leadership, and deliver new insights and tools to disrupt conventional business thinking. 

The Chair of Digital Data Design Institute at 性视界 and Dorothy and Michael Hintze Professor of Business Administration at 性视界 Business School (HBS), Karim Lakhani, will be joined with fellow HBS faculty members Iavor Bojinov, Rafaella Sadun, Rem Koning, Shunyuan Zhang, and Kadeem Noray to drive forward the portfolio of experiments. Their efforts will focus in five key areas: future-state operating models for effective human-AI collaboration in core business functions, agent boss (an initial management theory for AI agents), agentic workflows, building a Frontier Firm radar based on AI-native startups, and the effect of new technologies on firm demands for skills and labor.

Jared Spatro, Chief Marketing Officer, AI at Work at Microsoft articulated, 鈥淚t鈥檚 no longer a question of 鈥榠f鈥� AI is right for business鈥攍eaders today are grappling with 鈥榟ow鈥� to become a Frontier Firm. This Frontier Firm AI Initiative is addressing a critical gap in the marketplace, giving leaders the education and practical tools they need to help their people and organizations navigate this transformation.鈥�

The inaugural cohort of organizations聽embarking on the path to become Frontier Firms聽include Barclays, BNY, Clifford Chance, DuPont, Eaton, Eli Lilly and Company, EY, GHD, Kantar, Levi Strauss & Co., Lumen Technologies,聽Mastercard, Nestl茅,聽and others. Organizations will聽participate in large-scale field-based experiments in AI that explore AI-first work patterns as well as participate in custom workshops that translate the results of the research into practical guidelines for organizations innovating their operating models with AI.听

About Digital Data Design Institute at 性视界

The Digital Data Design Institute at 性视界 (D^3) provides research-driven insights, accessible to anyone in the world, on using AI and digital technologies to advance business and society. Emerging from 性视界 Business School under the leadership of Dean Srikant Datar and founded on the premise that AI technology is only half of the answer and that businesses must also revamp their processes to harness AI’s potential, D^3 is made up of a global network of multidisciplinary faculty, researchers and scientists, business leaders, and entrepreneurs. For more information, please visit d3.harvard.edu.

The post D^3 and Microsoft Launch Accelerated AI Research Initiative appeared first on HBS AI Institute.

Read a write up from the  of the report here:

The post State of the Market: An Industry Analysis of Tech-Enabled DEI Products appeared first on HBS AI Institute.

This article shares insights from Boris Groysberg, Richard P. Chapman Professor of Business Administration who is pursuing research on the topics of artificial intelligence and organizations.

1. What drew you to this area of research and how did you first become involved in this work?

It came through my field work.   

2.  What are some common misconceptions or barriers around the problem you鈥檙e working to solve?

There is a misconception that large AI initiatives can be implemented without thinking about the organizational and talent management impact and without consideration as to how the organizational structure might be adjusted to optimize these initiatives.  While everyone is talking about AI, virtually no one is talking about the organizational side of implementing AI. Because of this, there is not a lot of information available. This dearth of information is a barrier.

3.  What research is being done on this topic and how is your approach or perspective unique?

While much research has been done on the potential uses of artificial intelligence, the amount of investment in this area, and the types of skills required to effectively implement AI strategies, there has been little research on the organizational implications of AI. Our approach considers the following questions (among others): Where does AI leadership sit in the org chart?  Are AI resources centralized or dispersed?  How do these functions report upward? How does this vary based on the uses of AI, the scope of a company鈥檚 AI initiatives, and the company鈥檚 size and industry? Does the current org chart make sense in an AI environment?

4.  What excites you most about this work and its potential impact?

The potential to provide practical advice to executives who are looking to implement AI strategies.

5.  How do you hope working with D^3 will amplify the impact of your work?

The opportunity to connect with others working in the fast-moving AI field will likely provide invaluable information on how organizations are approaching AI, what are their primary organizational challenges, which approaches are working and which approaches are not.

6.  What changes do you hope to see in your field as a result of the work being done in this area?

We hope to see a more thoughtful approach to organizational structure and talent management and how it should be adapted to  an AI environment.

7.  What鈥檚 an essential area in which AI and digital technologies will reshape the way businesses or society operate in the long run that we may not be considering?

AI and digital technologies may well upend the traditional corporate organizational structure that has been in place across large organizations for the past 50 + years.

The D^3 Associates Program supports and accelerates faculty research into the ways AI and digital technologies are reshaping companies, organizations, society, and practice.

The post D^3 Associates Spotlight Series: Boris Groysberg appeared first on HBS AI Institute.

Key Insight: The Power of Synthetic Customers

By combining LLMs with traditional research methods, companies have the opportunity to simulate consumer sentiments like willingness-to-pay (WTP) to make product innovation faster and cheaper. The authors鈥� research shows that for simulated tests of categories like toothpaste and tablets, LLM-created synthetic customers produced realistic and accurate preferences for many familiar attributes. What鈥檚 more, teams could explore dozens or even hundreds of ideas by using these synthetic consumers as an initial filter, overcoming traditional limitations in scope.

Key Insight: The Competitive Advantage of Proprietary Data

While usage of LLMs out of the box showed promising results, companies that incorporate their own historical customer data were able to achieve better results. For example, the authors noted that LLMs often rate novelty higher than actual humans, and as a result synthetic customers were initially positive about pancake-flavored toothpaste. Fine-tuning the LLM with data from an actual study helped to correct this enthusiasm and produce WTP results more in line with actual human sentiment. The researchers found similar results when testing hypothetical features, like built-in projectors for laptops.听

Key Insight: Strategic Integration, Not Replacement

The most successful application of this technology comes from understanding it as an augmentation tool rather than a replacement for traditional research. Given that LLMs are trained on static data, they may not reflect current market conditions without receiving frequent updates and new data. This allows companies to follow a new innovation roadmap: broaden the top of the innovation funnel by using AI, but keep the bottom narrow through sharper, more cautious analysis.

Why This Matters

Synthetic customers might not totally replace human research, but they can dramatically enhance it. For business leaders and executives, this represents a fundamental shift in the speed and scope of innovation strategy. The ability to rapidly test multiple prototypes or concepts at low cost could mean faster time-to-market, reduced development risk, and more efficient resource allocation. Organizations that build internal AI-powered customer simulation capabilities could gain a significant competitive advantage from fine-tuning models with their proprietary data, creating a virtuous cycle where better data leads to better insights. At the same time, decision makers and marketing professionals must be vigilant to recognize and respond to the shortcomings of these new technologies and tools.

Bonus

Learn more about the authors鈥� original research, and go a step further with the GenAI + Marketing Learning Module from D^3. You鈥檒l learn the basics of engaging an LLM, with broadly applicable and actionable techniques to create content, automate tasks, and revolutionize workflows. Then the program will take a deep-dive to discover how AI can redefine your early-stage marketing research.听

References

[1] James Brand et al., 鈥淯sing Gen AI for Early-Stage Market Research,鈥� 性视界 Business Review, July 18, 2025, .听

[2] Brand et al., 鈥淯sing Gen AI for Early-Stage Market Research.鈥�

[3] Brand et al, 鈥淯sing Gen AI for Early-Stage Market Research.鈥�

Meet the Authors

is Principal Researcher and Economist in the Office of the Chief Economist at Microsoft.

is the Marvin Bower Associate Professor of Business Administration at 性视界 Business School and co-founder of the Customer Intelligence Lab at the Digital Data Design (D^3) Institute at 性视界. She studies omni-channel and e-commerce markets, and her research focuses on data-driven marketing, with an emphasis on how businesses can leverage their own data, customer data, and market data to improve outcomes.听

is Senior Director of Economics in the Office of the Chief Economist at Microsoft.

Did you find this insight to be useful?

• Yes
• No

Would you recommend this insight to others?

• Yes
• No

Submit

The post Larger, Faster, Cheaper: The Future of Market Research with AI appeared first on HBS AI Institute.

This article shares insights from Ashley V. Whillans, Volpert Family Associate Professor of Business Administration at 性视界 Business School who is pursuing research on the topics of artificial intelligence and organizations.

1. What drew you to this area of research and how did you first become involved in this work?

I’ve spent years studying workplace stress and well-being, watching organizations pour money into wellness programs with . By 2026, global corporate spending on wellness will top $94.6 billion, yet stress levels continue to rise鈥�45% of employees report feeling stressed 鈥渇requently鈥� or 鈥渁ll the time.鈥� The disconnect is staggering鈥攃ompanies are spending more than ever but achieving less. What struck me was how poorly we understand the variability of employee responses to well-being initiatives. Some employees thrive with certain interventions while others see no benefit, or worse, feel more stressed. I have found that high-stress employees  an additional $5.3 million annually per 1,000 employees, yet most wellness programs completely miss these workers. This insight led my postdoctoral fellow  and I to explore how we might leverage AI to predict these differential responses before implementation. We realized that generative AI agents could simulate everyday employee behavior by using the intensive longitudinal data we have been collecting for years鈥攄ata that captures employees鈥� everyday patterns of thinking, feeling, and behaving. 

2. What are some common misconceptions or barriers around the problem you’re working to solve?

One misconception is that employee well-being requires a one-size-fits-all solution. Leaders often believe that if they implement the “right” wellness program, it will work for everyone. But  suggests that high-stress employees鈥攚ho make up to 45% of the workforce in our recent analysis鈥攔espond very differently to interventions than their low-stress peers. 

Another misconception is that we need to choose between speed and accuracy when implementing workplace changes. Many companies run costly pilots that take months or make decisions based on gut instinct. Many leaders are understandably cautious about AI’s ability to accurately predict human behavior, especially as related to something as complex as workplace well-being because they have seen too many tech solutions overpromise and underdeliver. That is why we are focusing on rigorous validation against real-world data, to demonstrate that these simulations can genuinely help organizations make better people management decisions.

3. What research is being done on this topic and how is your approach or perspective unique?

While other researchers and practitioners in this space are using AI to analyze workplace data or provide chatbot therapy, we’re doing something different: We are creating dynamic digital twins of employees to predict their behavioral and psychological responses to organizational changes. Traditional agent-based modeling in management science has聽historically relied on simplified rules聽that fail to capture and simulate how employees’ patterns of thinking, feeling and behaving evolve over time. Our approach integrates intensive longitudinal psychological data (that we have collected through our research) with objective behavioral monitoring data (obtained through proprietary datasets that we obtained from employee monitoring tools) to create dynamic generative agents that exhibit realistic fluctuations in stress, engagement, and productivity. Unlike the vast majority of existing work on generative agents, we are not just tracking static demographics; we are modeling the temporal dynamics of workplace psychology. This means we can simulate not just whether a well-being intervention might work, but why it works for some employees (vs others), how its effectiveness varies over time, and any impact a psychological intervention might have on related workplace outcomes such as productivity.听

4. What excites you most about this work and its potential impact?

What gets me up in the morning is the possibility of ending a trial-and-error cycle that costs organizations millions and exhausts employees and managers alike. Imagine if leaders could test ten different versions of a well-being intervention virtually, see exactly which employees would benefit the most from it, and implement only the most promising approach, or even better, customize the design of the intervention for each segment separately. The possibility of this AI-led dynamic approach has the potential to reduce the $5.3 million annual cost per 1,000 employees that stress creates while improving worker productivity. Beyond the numbers, I’m excited about democratizing access to sophisticated predictive capabilities鈥攕maller organizations could use these predictive tools to make evidence-based decisions that previously required massive research budgets to design pilot programs, implement, and test at scale. 

5. How do you hope working with D^3 will amplify the impact of your work?

D^3’s support comes at a crucial moment when we need to move from proof-of-concept to real-world validation. The funding allows us to work with three industry partners, creating a feedback loop between academic rigor and practical application. Being part of D^3’s network also facilitates critical connections with researchers tackling AI challenges across completely different domains. These cross-pollination opportunities are invaluable. In our case, someone working on AI in finance might have insights about risk prediction that transform how we model the longitudinal employee stress response. D^3’s emphasis on translating research into practitioner tools aligns with our goal of making this research accessible beyond academia.

6. What changes do you hope to see in your field as a result of the work being done in this area?

I envision a future where no organization implements a major employee initiative without running simulations to understand differential impact first. This change would shift how we think about workplace interventions鈥攆rom hoping they work to knowing how they’ll work and for whom. We need to move beyond the current approach where 85% of companies offer wellness programs but stress levels continue to rise. Our field must embrace the complexity of human behavior at work rather than defaulting to simplistic solutions. If we succeed, the standard practice to implement any employee initiative that affects workers lives will be to test these interventions virtually first, just as engineers simulate building performance before construction.

7. What’s an essential area in which AI and digital technologies will reshape the way businesses or society operate in the long run that we may not be considering?

Stress, burnout, disengagement, distrust, and disillusionment can spread through organizations like wildfire, but we’ve never been able to model these dynamics accurately. With AI simulations, we can predict not only individual responses, but how one person’s improved well-being or trust might cascade through their team, or conversely, how a stressed manager might impact an entire department’s productivity. This network modeling could reshape how we think about organizational health鈥攎oving from treating individual “symptoms” to understanding and simultaneously managing the entire organizational ecosystem. The implications of this predictive approach extend beyond well-being to the design of teams, offices, and org charts. 

With AI simulations, we can intentionally and efficiently architect the future of work. The organizations that master this predictive capability won’t just manage workplace stress better鈥攖hey’ll build resilient cultures that more adeptly turn human potential into competitive advantage.

The D^3 Associates Program supports and accelerates faculty research into the ways AI and digital technologies are reshaping companies, organizations, society, and practice.

The post D^3 Associates Spotlight Series: Ashley V. Whillans appeared first on HBS AI Institute.

Jen Stave Jen Stave

Key Insight: Agentic AI as Managed Teammates

Agentic tools are moving beyond chatbots and image generation. Unlike traditional automation that follows rigid scripts, AI agents function more like human employees: capable of independent decision-making after being given high-level goals and objectives.

Key Insight: An Uncertain Future

On one hand, the impact of AI is already here and being measured, as evidenced by how the use of AI agents at Salesforce led to a 17% customer service cost reduction over nine months. But the article also raises a range of undecided questions related to economic capture, quality and accountability, and the right balance between human and AI worker numbers.

Why This Matters

For forward-thinking executives, increasingly the question isn鈥檛 whether to adopt agentic AI, but how to operationalize it productively and responsibly. While the efficiency gains are compelling, success requires thoughtful integration by leaders who are ready to address challenges of workforce transition, quality control, and ROI measurement.

Bonus

To read more about Agentic AI and digital labor, read 鈥�,鈥� co-authored by Jen Stave, for the 性视界 Business Review.

Did you find this insight to be useful?

• Yes
• No

Would you recommend this insight to others?

• Yes
• No

Submit

The post When Software Becomes Staff appeared first on HBS AI Institute.

Key Insight: Accelerate or Anchor?

The study identifies a foundational behavioral pattern: when an agent鈥檚 current task closely resembles a stored memory, the outputs tend to closely match as well. This 鈥渆xperience-following鈥� correlation mirrors how humans often rely on familiar patterns, and it can accelerate learning when the stored example is correct. However, it鈥檚 also not without risks. If erroneous or low-quality experiences are stored in memory, they can be applied to future tasks, thereby decreasing the agent鈥檚 overall performance. This means that the quality of stored examples is paramount, as bad memories don鈥檛 just linger, they can create a propagating error feedback loop.

Key Insight: Selective Addition

If the experience-following property shows why quality matters in LLM agents, then addition shows how to control it, and a clear finding from the study is that indiscriminate memory growth actually hurts performance. In tests with three different agents, covering electronic health records (EHRs), the LLM-based autonomous driving agent AgentDriver, and a network security agent, storing every task and output (鈥渁dd-all鈥�) performed worse than using no memory addition at all. However, using strict evaluation criteria and filtering before storage led to an average 10% performance boost, so memory improvement is less about hoarding information than curating a high-quality knowledge base.

Key Insight: Improvement through Deletion

Even with careful addition, not all stored experiences are equally useful over time. Some look similar to new tasks (鈥渉igh input similarity鈥�), but consistently produce poor output (鈥渓ow output similarity鈥�). The authors term this 鈥渕isaligned experience replay,鈥� and show that pruning these entries improves long-term outcomes. Removing experiences with repeatedly low utility (鈥渉istory-based deletion鈥�) offered the best boost to performance while effectively and efficiently maintaining memory size. From a strategic perspective, this practice mirrors audits of playbooks, datasets, and best practices to ensure that institutional knowledge remains in top shape.

Why This Matters

The results from this research should give business leaders important context for thinking about how to choose and deploy AI agents: more data isn鈥檛 automatically better, and AI鈥檚 鈥渆xperience鈥� can actually be a liability, entrenching errors and bloating infrastructure. Disciplined curation, by selectively adding high-value experiences and strategically deleting low-value or misaligned ones, yields not only better accuracy but also more efficient, adaptable systems. In a world where executives may be involved in decision-making around LLM agents for their organizations, it鈥檚 important to have a blueprint for keeping AI agents sharp, reliable, and resilient, just like they plan for the training and advancement of their human employees. By understanding and investing in the processes that keep your AI鈥檚 memory in top-shape, your business will be equipped for tomorrow鈥檚 challenges.

References

[1] Zidi Xiong et al., 鈥淗ow Memory Management Impacts LLM Agents: An Empirical Study of Experience-Following Behavior,鈥� arXiv preprint arXiv:2505.16067v1 (May 21, 2025): 2.

[2] Xiong et al., 鈥淗ow Memory Management Impacts LLM Agents,鈥� 5.

[3] Xiong et al., 鈥淗ow Memory Management Impacts LLM Agents,鈥� 9.

Meet the Authors

is a PhD student in computer science at 性视界 University, advised by Himabindu Lakkaraju.

is a PhD student in computer science at Michigan State University.

is a PhD student in computer science engineering at University of Minnesota – Twin Cities.

is a PhD student in computer science and engineering at Michigan State University.

is a University Foundation Professor in the computer science and engineering department at Michigan State University.

is an Assistant Professor of Business Administration at 性视界 Business School and PI in D^3鈥檚 Trustworthy AI Lab. She is also a faculty affiliate in the Department of Computer Science at 性视界 University, the 性视界 Data Science Initiative, Center for Research on Computation and Society, and the Laboratory of Innovation Science at 性视界. Professor Lakkaraju’s research focuses on the algorithmic, practical, and ethical implications of deploying AI models in domains involving high-stakes decisions such as healthcare, business, and policy.

is Assistant Professor of Computer Science at University of Georgia.

Did you find this insight to be useful?

• Yes
• No

Would you recommend this insight to others?

• Yes
• No

Submit

The post Smarter Memories, Stronger Agents: How Selective Recall Boosts LLM Performance appeared first on HBS AI Institute.