Lots of organizations can green-light AI. Far fewer can absorb it. That gap, between excitement and real, embedded use, keeps showing up even when ROI is compelling and leadership is visibly supportive. New research from 性视界 Business School AI Institute Frontier Firm affiliate Shunyuan Zhang and Das Narayandas reveals an uncomfortable idea contributing to this gap. In 鈥�,鈥� they highlight that the very people who must approve and champion these technologies are the same ones whose jobs could be fundamentally threatened by them.

Key Insight: The Three Threats of Self-Disruptive Technologies (SDTs)

Traditional adoption theories typically focus on whether organizations are ready, whether the technology is useful, and whether there鈥檚 institutional pressure to adopt. But these frameworks miss something critical: they assume decision-makers are neutral agents acting on behalf of the firm. Now, add to the mix AI systems with the potential to automate managerial judgment, analytics platforms that centralize decision rights, or algorithmic tools that replace experiential expertise with codified models. When the manager in charge of approving these technologies anticipates that they will shrink their own role or reduce their influence, the approval decision becomes identity-laden. These Self-Disruptive Technologies, as Narayandas and Zhang call them, trigger three forms of role-level identity threat. Role compression occurs when automation shifts core work from 鈥渄eciding鈥� to 鈥渕onitoring,鈥� compressing the judgment and expertise that defines a role鈥檚 distinctive contribution. Control shift happens when discretion moves away from the approving role (e.g. centralized to analytics teams or delegated to algorithms), removing the decision authority that makes roles defensible within organizations. Span erosion reflects the contraction of influence over people, budgets, or processes, undermining status and future opportunity even when the formal position remains intact.

What makes these threats particularly powerful is that they can dominate the approval calculus even when firm-level incentives favor adoption and economic cases are strong. A manufacturing supervisor might support efficiency improvements in principle but resist when the technology eliminates the judgment calls that justify their expertise. A procurement manager might delay adopting an AI tool that demonstrably reduces costs because it centralizes decisions that previously sustained their organizational influence.

Key Insight: Engineering the Solution – Identity-Compatible Advantage (ICA)

Here鈥檚 where the research gets actionable. Narayandas and Zhang argue for an approach of Identity-Compatible Advantage to require bundling new technology with governance and role-design mechanisms that make adoption personally and politically defensible for managers. ICA includes five complementary elements: role rechartering that redefines the role around higher-order judgment rather than routine decisions; decision guardrails that preserve authority through override rights and and governance structures; analytical overlays that frame technology as augmentative rather than substitutive; redeployment pathways that provide credible commitments to role evolution rather than elimination; and executive sponsorship that legitimizes identity transition and reallocates accountability. 

The research emphasizes that these mechanisms work as a bundle, not in isolation. For example, implementing guardrails without rechartering leaves meaning unaddressed, as you give the manager the power to override the AI (restoring some control), but because the AI still does the core work, the manager feels their daily expertise is useless (leaving their loss of purpose and contribution unaddressed). The framework shows that successful SDT adoption requires designing offerings where endorsement becomes personally and politically defensible. 

Why This Matters

Most AI automation discourse has fixated on individual contributors like programmers, graphic designers, and copywriters because their work products are visible and the substitution story is easy to tell. This research adds a missing piece: the managers and decision-makers, who control whether AI technologies get adopted in the first place, are themselves facing automation of their core judgment and authority. For executives and business leaders, the implications are profound. If you treat AI adoption as a purely rational calculation, you are likely to be met with 鈥渟ymbolic adoption,鈥� where your team pays lip service to innovation while quietly ensuring that the status quo remains undisturbed. By utilizing Identity-Compatible Advantage, leaders can implement the complex undertaking of AI adoption as an evolution of their teams, not a replacement of them. The future of work belongs to the firms that can successfully re-anchor identities around high-level strategy, risk ownership, and the human-centric decisions that no machine can replicate.

Bonus

The path to real AI adoption runs through design choices: how you frame AI, where you keep humans in the loop, and how you protect legitimacy. For another look at the dynamics of AI in the workplace, check out Drawing the Line on AI Usage in the Workplace.

References

[1] Narayandas, Das and Shunyuan Zhang, 鈥淪elling Self-Disruptive Technologies: Identity-Compatible Advantage and the Role-Level Microfoundations of Automation Adoption.鈥� 性视界 Business School Working Paper, No. 26-050 (February 9, 2026): 5.  

[2] Narayandas and Zhang, 鈥淪elling Self-Disruptive Technologies,鈥� 9.

Meet the Authors

is Edsel Bryant Ford Professor of Business Administration at 性视界 Business School.

is Associate Professor of Business Administration at 性视界 Business School. She and other HBS faculty contribute to the HBS AI Institute Frontier Firm Initiative.

Did you find this insight to be useful?

• Yes
• No

Would you recommend this insight to others?

• Yes
• No

Submit

The post The Manager鈥檚 AI Dilemma appeared first on 性视界 Business School AI Institute.

Think about the last time you contacted customer support. Did you start with a chatbot? If it failed to resolve your problem, how did you feel when transferred to a human agent? This dynamic defines our expectations of the modern customer service experience: the struggle to balance the cold speed of automation with the warm necessity of human empathy. However, in 鈥�,鈥� 性视界 Business School AI Institute Frontier Firm affiliate Shunyuan Zhang and Das Narayandas explain how the results from a year-long experiment involving 138 customer service agents and over 250,000 conversations are far more complex than the typical assumption.聽

Key Insight: AI Assistance Isn鈥檛 Just Faster, It鈥檚 More Human

The prevailing fear in customer service is that introducing AI will turn human interactions into robotic, assembly-line exchanges. Yet, when agents received real-time AI-generated reply suggestions, they didn鈥檛 just respond 22% faster to customer messages, they actually sent more messages and saw a measurable boost in the 鈥渉uman鈥� quality of the chats. AI freed agents from the cognitive burden of composing responses, allowing them to engage customers more deeply. The conversations became richer, not shallower: using large language models to categorize agent messages, the researchers found that AI-assisted responses scored higher in the key aspects of empathy, information, and solution, with the largest jump in empathy.

Key Insight: The Experience Equalizer

One of the most business-relevant findings in the study was that AI assistance didn鈥檛 benefit everyone equally. When the researchers examined how agent tenure moderated AI鈥檚 effects, they found that less-experienced agents gained far more from AI suggestions than their veteran counterparts. Essentially, the AI 鈥渄ownloaded鈥� institutional knowledge into the workflow of new employees: having access to these real-time suggestions was the functional equivalent of nearly five months of experience. This has profound implications for industries with high turnover, suggesting that AI can serve as a stabilized bridge, ensuring that a customer鈥檚 experience doesn鈥檛 suffer just because they happened to be connected to a trainee.

Key Insight: Not All Conversations Are Created Equal

The AI algorithm鈥檚 impact varied depending on why customers were reaching out in the first place. For example, when customers wanted to cancel subscriptions鈥攖raditionally difficult conversations鈥擜I helped agents identify underlying reasons and recommend alternative options, leading to notable improvements in customer sentiment. But repeat complaints told a different story. Although AI helped agents respond quickly in these scenarios, customer sentiment barely improved. These complaints stemmed from systematic operational issues, like recurring delivery problems, that no amount of empathetic, information-rich messaging could solve. The AI could help agents communicate better about problems, but it couldn鈥檛 actually fix them.

Perhaps the most counterintuitive finding emerged from examining what happened in the handoff from a bot to a human agent. Many companies use a 鈥渃hatbot first鈥� approach, where a fully automated bot tries to solve the problem before transferring the customer to a human. As we鈥檝e seen, AI-assisted agents are able to respond more quickly, and if the AI-assisted agent responded too quickly, customers suspected that they were still talking to a bot. The response speed that might normally delight customers became a liability, triggering what the researchers term a negative 鈥渟pillover鈥� from the initial bot failure. In these contexts, the study found that increasing the delay in human responses actually helped rebuild trust and improve sentiment.

Why This Matters

For executives deploying AI in customer-facing operations, this research delivers three strategic imperatives. First, resist the temptation to replace human agents entirely: augmentation delivers better outcomes than automation alone, particularly for handling nuanced, emotionally charged interactions. Second, deploy AI with precision: it鈥檚 most valuable in specific conversation types (like retention scenarios). Third, manage your AI ecosystem holistically. If you鈥檙e using multiple AI systems in sequence, recognize that they鈥檙e not independent. The companies that will win with AI aren鈥檛 those that deploy the most LLMs, they鈥檙e those that understand how these systems interact across the entire customer ecosystem and adapt their implementation accordingly.

Bonus

When emotions are involved, who people think is responding can shape outcomes as much as what is said. For another angle on AI and human emotion, check out It Feels Like AI Understands, But Do We Care? New Research on Empathy.

References

[1] Zhang, Shunyuan, and Das Narayandas, 鈥淓ngaging Customers with AI in Online Chats: Evidence from a Randomized Field Experiment.鈥� Management Science 72 (1) (2025): 84.  

[2] Zhang and Narayandas, 鈥淓ngaging Customers with AI in Online Chats,鈥� 84.

[3] Zhang and Narayandas, 鈥淓ngaging Customers with AI in Online Chats,鈥� 75-76.

Meet the Authors

is Associate Professor of Business Administration at 性视界 Business School. She and other HBS faculty contribute to the HBS AI Institute Frontier Firm Initiative.

is Edsel Bryant Ford Professor of Business Administration at 性视界 Business School.

Did you find this insight to be useful?

• Yes
• No

Would you recommend this insight to others?

• Yes
• No

Submit

The post The Fast-Talking AI Chat Agent appeared first on 性视界 Business School AI Institute.

Many of us have started using AI as an 鈥渁nswer machine鈥� to brainstorm ideas, analyze data, and pressure-test assumptions. You might even have a set of preferred prompts saved for just these purposes. But how often do you switch the order of the steps you give the AI, and what kind of influence does it have on the output? In 鈥�,鈥� a team including 性视界 Business School AI Institute鈥檚 shows that when organizations integrate AI into multi-stage innovation evaluation processes, the sequence creates a powerful but largely invisible tradeoff. By understanding how to structure this dynamic, we can gain a crucial advantage when identifying ideas that could bring value to our portfolios and organizations.聽

Key Insight: Navigating Between Novelty and Feasibility

Evaluating innovation is a high-stakes balancing act between two competing forces: novelty and feasibility. You want solutions that depart from established approaches (novelty) and that can realistically be built and implemented (feasibility). But as the authors note, evaluators can鈥檛 weigh everything simultaneously, so they prioritize one over the other at different stages of the process (criteria-sequencing): either novelty-then-feasibility or feasibility-then-novelty. These sequences lead to different results because the order acts as the initial filter. If a solution is eliminated in the first stage based on one criterion (e.g. feasibility), it is never evaluated on the second (e.g. novelty). Since evaluators apply these criteria in personalized ways, the order they use can lead to inconsistent decisions.

Key Insight: An AI Innovation Experiment

To see how AI could structure these heuristics, the researchers partnered with the crowdsourcing platform Hackster.io for a field experiment involving 353 evaluators and 132 open-source solutions. They utilized two distinct types of AI: Predictive AI and Generative AI. Predictive AI, which excels at identifying patterns from past data, was used to provide feasibility recommendations based on technical benchmarks. Generative AI, capable of recombining knowledge in unconventional ways, provided novelty-focused recommendations. Both systems provided 鈥淧ass鈥� or 鈥淔ail鈥� recommendations with explanatory content, with half of the evaluators receiving feasibility-then-novelty sequencing, and the other half receiving novelty-then-feasibility. The researchers predicted that the criteria-sequencing would create what they call a mean-variance innovation tradeoff: feasibility-then-novelty would allow evaluators to take greater risks with fewer options, resulting in higher mean innovation, while novelty-then-feasibility would cast a wider initial net, surfacing atypical solutions and producing higher variance.

Key Insight: Tradeoffs in the Pursuit of Breakthroughs

The results supported the researchers鈥� predictions, meaning the order of evaluation dictates the type of innovation an organization is likely to champion. The researchers also found in post hoc analysis that the AI鈥檚 format played a role: compared to a static summary, an interactive chatbot increased innovation variance, but led to a lower mean innovation rating. It appears that without a fixed, standardized summary to guide them, evaluators spent more time exploring diverse questions and ultimately relied more on their own judgment. As a result, evaluations became more complex, average quality declined, and the set of selected options became more diverse.This suggests that dynamic 鈥渢hought partners鈥� encourage more exploration and reliance on human judgment, while static AI recommendations act more as rigid guides. 

Why This Matters

For business leaders and executives encouraging their employees to use AI-augmented workflows, this research fundamentally reframes the integration question. It鈥檚 not just about whether to use AI, or even which tasks to automate versus augment, it鈥檚 about recognizing that AI recommendations create structure that shapes human judgment in path-dependent ways. The sequence you choose could determine whether your organization builds a portfolio optimized for steady performance or one that swings for breakthrough innovation. The question isn鈥檛 whether AI will influence your decisions, it鈥檚 whether you鈥檒l deliberately design that influence, or let it emerge accidentally from your initial prompt. 

Bonus

For another look at how AI can shape outcomes by steering what kind of ideas people generate and select, check out 鈥�The Creative Edge: How Human-AI Collaboration is Reshaping Problem-Solving.鈥�&苍产蝉辫;

References

[1] Grumbach, Cyrille, Jacqueline N. Lane, and Georg von Krogh, 鈥淭he Mean-Variance Innovation Tradeoff in AI-Augmented Evaluations,鈥� 性视界 Business School Technology & Operations Mgt. Unit Working Paper No. 26-038 (2025): 1.  

[2] Grumbach et al., 鈥淭he Mean-Variance Innovation Tradeoff in AI-Augmented Evaluations,鈥� 2. 

[3] Grumbach et al., 鈥淭he Mean-Variance Innovation Tradeoff in AI-Augmented Evaluations,鈥� 33.

Meet the Authors

is a PhD Candidate and Research Associate at the Chair of Strategic Management and Innovation at ETH Zurich.

is Assistant Professor of Business Administration at HBS and co-Principal Investigator of the Laboratory for Innovation Science at 性视界 (LISH) at the HBS AI Institute.

is a Professor at ETH Zurich and holds the Chair of Strategic Management and Innovation.

Did you find this insight to be useful?

• Yes
• No

Would you recommend this insight to others?

• Yes
• No

Submit

The post How AI Can Spot Your Next Billion-Dollar Idea appeared first on 性视界 Business School AI Institute.

As we move into the era of agentic AI, what kind of influence will this emerging technology have on democracy and misinformation? In the new Science paper 鈥�,鈥� , Assistant Professor of Business Administration at 性视界 Business School and Faculty PI of the at the 性视界 Business School AI Institute, and an international, multi-disciplinary group of co-authors argue that we鈥檙e entering a phase where 鈥渕alicious AI swarms鈥� could use multi-agent systems to infiltrate communities, mimic human social behavior, and iteratively refine persuasion tactics in real time. By expanding misinformation into persistent manipulation, these systems threaten the information ecosystem that democratic societies depend on, but Goldenberg and his co-authors also outline technical, economic, and institutional measures that could meaningfully defend against this new danger.

Key Insight: AI Swarms Operate Like Digital Societies

Unlike earlier botnets, which relied on centralized control, rigid scripts, and human labor, AI swarms combine LLM reasoning with multi-agent architectures to function more like adaptive digital societies. The authors define malicious AI swarms as systems of persistent agents that coordinate toward shared objectives, adapt in real time to engagement and platform cues, and operate with minimal human oversight across platforms. Five capabilities make these systems especially potent. (1) Swarms replace centralized command with fluid coordination, allowing thousands of AI personas to locally adapt while periodically synchronizing narratives. (2) They can map social networks to identify and infiltrate vulnerable communities with tailored appeals. (3) Human-level linguistic mimicry and irregular behavior patterns help them evade detection. (4) Continuous, automated A/B testing enables rapid optimization of persuasive content. (5) Finally, their always-on persistence allows influence to accumulate gradually, embedding itself within communities over time and subtly reshaping norms, language, and identity. As the article notes, recent elections in Taiwan and India already saw a proliferation of AI-generated propaganda and synthetic media outlets, meaning that this threat is already here and poised to expand in the future.

Key Insight: The Harm Cascade

Goldenberg and his team argue that AI swarms could trigger a 鈥榗ascade鈥� of harms by systematically distorting the information ecosystem. By engineering 鈥榮ynthetic consensus鈥� and targeting different misinformation to different communities, these agents would have the power to undermine the independent thought essential for collective intelligence while simultaneously fragmenting the public sphere. This manipulation, together with coordinated synthetic harassment campaigns, could create a hostile environment that drives journalists and citizens into silence. The damage would compound as swarms 鈥榩oison鈥� the web with fabricated content that contaminates future AI training data. Ultimately, this sustained erosion of trust could corrode institutional legitimacy, rendering democratic safeguards vulnerable to collapse.

Key Insight: A Layered Defense Strategy

Rather than a single fix, the authors argue for a layered defense strategy designed to raise the cost, complexity, and visibility of swarm-based manipulation. The first layer is always-on detection: continuous monitoring systems that identify statistically anomalous coordination patterns in real time, paired with public audits and transparency to reduce misuse. Because attackers will adapt, detection alone is insufficient. A second layer involves simulation and stress-testing. Agent-based simulations can replicate platform dynamics and recommender systems, allowing researchers and platforms to probe how swarms might evolve to recalibrate defenses before major elections or crises. Third, the authors emphasize empowering users through optional 鈥淎I shields,鈥� tools that flag likely swarm activity, allowing individuals to recognize suspicious content. Finally, the paper highlights governance and economic levers as essential. Proposals include standardized persuasion-risk evaluations for frontier models, mandatory disclosure of automated identities, stronger provenance infrastructure, and a distributed AI Influence Observatory to coordinate evidence across platforms, researchers, and civil society. Crucially, the authors argue that disrupting the commercial market for manipulation may be among the most effective ways to reduce large-scale abuse.

Why This Matters

For business leaders and professionals, this study reveals a threat that extends beyond electoral politics into the fundamental information ecosystem that underpins market confidence, consumer behavior, and corporate reputation. The same AI swarm technologies that  manipulate political discourse could target brand perception, financial markets, or industry narratives just as easily. The defense strategy outlined by the authors can similarly provide a roadmap for corporate action: implementing detection systems for monitoring threats to brand reputation, advocating for industry standards around AI transparency, and supporting governance initiatives that protect the broader information ecosystem. Executives who treat information integrity as core infrastructure will be better positioned to protect stakeholder trust, decision quality, and long-term resilience in an era of AI-enabled influence operations.

Bonus

For a look at how efforts to align AI systems with human preferences can unintentionally undermine trustworthiness itself, check out 鈥�AI Alignment: The Hidden Costs of Trustworthiness.鈥�&苍产蝉辫;

References

[1] Daniel Thilo Schroeder et al., 鈥淗ow Malicious AI Swarms Can Threaten Democracy,鈥� Science (391) (2026): 354.  

[2] Schroeder et al., 鈥淗ow Malicious AI Swarms Can Threaten Democracy,鈥� 355.

[3] Schroeder et al., 鈥淗ow Malicious AI Swarms Can Threaten Democracy,鈥� 357.

Meet the Authors

is an assistant professor in the Negotiation Organization & Markets unit at 性视界 Business School, an affiliate with 性视界鈥檚 Department of Psychology, and a faculty principal investigator in the HBS AI Institute’s Digital Emotions Lab.

Additional Authors: Daniel Thilo Schroeder, Meeyoung Cha, Andrea Baronchelli, Nick Bostrom, Nicholas A. Christakis, David Garcia, Yara Kyrychenko, Kevin Leyton-Brown, Nina Lutz, Gary Marcus, Filippo Menczer, Gordon Pennycook, David G. Rand, Maria Ressa, Frank Schweitzer, Dawn Song, Christopher Summerfield, Audrey Tang, Jay Van Bavel, Sander van der Linden, and Jonas R. Kunst

Did you find this insight to be useful?

• Yes
• No

Would you recommend this insight to others?

• Yes
• No

Submit

The post The New Influence War: How AI Could Hack Democracy appeared first on 性视界 Business School AI Institute.

Conventional sales wisdom treats persistence as virtue: stay in the conversation, overcome objections, keep the line alive. But recent research into the dynamics of sales conversations suggests that our bias toward persistence leads to a massive misallocation of resources. In 鈥�,鈥� a team including , Professor of Business Administration at HBS and co-founder of the Customer Intelligence Lab at the 性视界 Business School AI Institute, explain how they built a generative AI 鈥渟topping agent鈥� that watches sales transcripts in real time and chooses to quit or wait to maximize cumulative payoff. The result? The ability to lift expected sales by over 30%.聽

Key Insight: Quitting is an Optimization Problem

The authors argue that 鈥渄ynamic qualification鈥濃�斺�漷he decision of whether and when to quit a sales call that is unlikely to succeed鈥� [2]鈥攊s an under-studied decision problem with enormous operational stakes. In a dataset of 11,627 sales calls analyzed by the researchers, over 94% fail, with an average call length of 169 seconds. Humans tend not to be responsive enough to subtle, early linguistic cues that indicate a lack of progress, while simultaneously becoming too responsive to negative signals only after they have already invested significant time in the call. The authors show that early call language is informative enough that a fine-tuned LLM can predict eventual failure quite well after just 60 seconds. The critical challenge is not only in predicting failure, but determining the optimal moment to intervene, which the authors call an ‘optimal stopping’ problem.

Key Insight: Teaching Machines the Wisdom of Hindsight

The researchers鈥� innovation lies in recognizing that you can retrospectively identify the optimal moment to quit any historical conversation by comparing what actually happened with what would have happened at different quitting times. For each call in the training data, the algorithm calculates when quitting would have maximized the expected cumulative reward (balancing time costs against sales revenue). This creates a dataset of 鈥渆xpert鈥� decisions鈥攖he optimal quit-or-wait choice at every moment in every conversation. The researchers then fine-tuned GPT-4.1 to generate these optimal decisions given the evolving transcript. What makes this particularly powerful is how the stopping agent learns dynamic decision rules that evolve as the conversations progress. At 30 seconds, it focuses on whether the salesperson reached the correct person. By 60 seconds, it shifts to gauging interest levels. At 90 seconds, it keys into whether the prospect already has a similar product or service. 

Key Insight: Reclaiming Revenue Hidden in 鈥淒ead Air鈥�

The managerial implications of this research are profound, moving beyond mere cost-cutting to actual revenue generation. By implementing the 鈥渟topping agent鈥� at different levels of 鈥渁ggressiveness,鈥� the study demonstrated that a firm could reduce the time spent on failed calls by 54%. [5] Crucially, this isn鈥檛 just about ending calls faster, it鈥檚 about what happens next. When the time saved by quitting doomed calls was reallocated, expected sales increased by up to 37%. This represents a massive gain in productivity. In sales, the focus has often been on 鈥渆ffort motivation鈥濃�攗sing commissions and quotas to make people work harder, but this research argues that 鈥渆ffort optimization鈥濃�攈elping people work smarter by knowing when to stop鈥攎ight be the more powerful lever. The stopping agent also doesn鈥檛 replace the human element of persuasion. Instead, it serves as a silent assistant that protects the salesperson鈥檚 time. 

Why This Matters

For today鈥檚 business leaders, the takeaway is clear: efficiency is not just about doing things faster, it鈥檚 about choosing not to do the things that don鈥檛 work. In an era where AI is increasingly viewed through the lens of total automation, this research also offers a more sophisticated model. It demonstrates that the most effective use of generative AI isn鈥檛 to replace the human salesperson, but to provide them with 鈥渄ecision support鈥� that corrects for natural psychological biases. This methodology scales beyond sales to any domain with sequential decisions and observable outcomes. The question for leaders isn鈥檛 whether their teams face similar cognitive constraints, they almost certainly do, it鈥檚 whether they鈥檙e ready to systematically identify and correct them.

Bonus

For another use case where AI doesn鈥檛 replace humans, but offers the opportunity to improve judgment, break silos, and accelerate execution, check out 鈥�The Cybernetic Teammate: How AI is Reshaping Collaboration and Expertise in the Workplace.鈥�

References

[1] Manzoor, Emaad, Eva Ascarza, and Oded Netzer, 鈥淟earning When to Quit in Sales Conversations,鈥� arXiv preprint arXiv:2511.01181 (2025): 23. . See also . 

[2] Manzoor et al., 鈥淟earning When to Quit in Sales Conversations,鈥� 1.

[3] Manzoor et al., 鈥淟earning When to Quit in Sales Conversations,鈥� 2.

[4] Manzoor et al., 鈥淟earning When to Quit in Sales Conversations,鈥� 21.

[5] Manzoor et al., 鈥淟earning When to Quit in Sales Conversations,鈥� 2.

Meet the Authors

is an Assistant Professor of Marketing and a graduate field member of Computer Science at Cornell University.

is Professor of Business Administration at 性视界 Business School and co-founder of the Customer Intelligence Lab at The HBS AI Institute.

is the Vice Dean of Research and the Arthur J. Samberg Professor of Business at Columbia Business School.

Did you find this insight to be useful?

• Yes
• No

Would you recommend this insight to others?

• Yes
• No

Submit

The post The Power of AI Stopping Agents appeared first on 性视界 Business School AI Institute.

For the last two decades, businesses have operated under the shadow of the Web 2.0 era, where a handful of giants like Google built unassailable tech fortresses. Now, as generative AI transforms every industry, leaders face an urgent question: are we watching the same scenario again, or is this time genuinely different? In the new article 鈥�,鈥� a team of researchers, including , co-Principal Investigator of the Platform Lab at the 性视界 Business School AI Institute and Glenn and Mary Jane Creamer Associate Professor of Business Administration at 性视界 Business School, explains how the GenAI market is more dynamic and 鈥渃ontested鈥� than the headlines suggest. By analyzing the economic fundamentals of AI platforms, tracking hundreds of acquisitions across the AI value chain, and surveying more than 300 business leaders about their actual AI adoption patterns, their research reveals a market that remains remarkably open, provided you know where the real leverage points lie.

Key Insight: GenAI isn鈥檛 Repeating Web 2.0鈥檚 Playbook

While many observers fear that GenAI could follow the same path of consolidation as Web 2.0, the researchers found a different story. Whereas platforms like Facebook relied on network effects (where an increase in users created greater value for all users), GenAI tools currently function as individual productivity and knowledge aids without forming connections between users. Data feedback loops also appear weaker. While models do learn from interaction, the authors note that the complexity of GenAI interactions makes it harder to engineer the kind of self-improving loop that helped search and social products pull away from rivals. They even flag a downside risk, 鈥渕odel collapse,鈥� where training on synthetic data can amplify errors and bias over time rather than improve quality. Finally, there鈥檚 pricing. Unlike many Web 2.0 services, GenAI consumes substantial compute and energy, which creates real costs and pushes providers toward tiered pricing. That matters competitively, because it reintroduces a familiar dynamic: entrants can attack on cost, quality, or both, rather than being boxed out by a dominant incumbent offering a free service.

Key Insight: Strategic Moves Across the AI Stack

The GenAI economy is best understood as a stack of five layers: chip manufacture, design, compute infrastructure, foundation models, and applications. While the application layer is exploding with roughly 1,600 active firms, the top of the stack remains highly concentrated. This has fueled a flurry of vertical integration. For instance, NVIDIA has expanded downstream from chip design into model orchestration through acquisitions like Run:AI, while cloud providers like Microsoft and Amazon are moving upstream into chip design and securing exclusive partnerships with model developers like OpenAI and Anthropic. However, this integration isn鈥檛 always a sign of impending monopoly. Cross-layer moves can actually increase competition by reducing dependencies. For example, NVIDIA partnering with emerging cloud providers like CoreWeave creates competition for AWS and Azure. 

Key Insight: A Still-Open Market

The researchers surveyed 323 business leaders across industries and geographies in May 2025. Nearly 90 percent report some GenAI use within their organizations, but what鈥檚 striking is how they鈥檙e using it. The vast majority are multihoming鈥攗sing multiple models such as ChatGPT, Claude, and Gemini, simultaneously. While this could be users taking advantage of greater capabilities for specific tasks within certain models, the authors also suggest a broader economic hypothesis: multihoming enables flexibility and thereby prevents potentially costly lock-in at an early stage of the GenAI transformation.

Bonus

Even if GenAI competition stays 鈥渃ontested,鈥� you can still end up locked in through messy, unmanaged adoption. For a look at why AI strategy is organization design strategy, check out 鈥�The People, Processes, and Politics of AI ROI.鈥�

Why This Matters

For business leaders and executives, this is a strategy and execution problem disguised as a technology trend. GenAI鈥檚 current economics suggest the market may stay contestable longer than Web 2.0 did, and the move is to treat the moment less like vendor selection and more like competition positioning. Design your organization to learn fast, build internal muscle, and avoid early lock-in while tools, pricing, and performance are still moving targets. 

References

[1] Andrea Asoni et al., 鈥淐ontested Ground: Early Competition and Market Dynamics in Generative AI,鈥� Management and Business Review, 5(4) (2025): 55, .

[2] Asoni et al., 鈥淐ontested Ground鈥�: 58.

[3] Asoni et al., 鈥淐ontested Ground鈥�: 61.

Meet the Authors

is an economist and Vice President of Charles River Associates.

is Glenn and Mary Jane Creamer Associate Professor of Business Administration at 性视界 Business School and co-Principal Investigatory of the Platform Lab at the HBS AI Institute.

is a Vice President in Charles River Associates鈥� European Competition Practice.

is a Vice President in Charles River Associates鈥� European Competition Practice.

Did you find this insight to be useful?

• Yes
• No

Would you recommend this insight to others?

• Yes
• No

Submit

The post Is GenAI Heading for a Tech Monopoly? appeared first on 性视界 Business School AI Institute.

In the new paper 鈥�,鈥� researchers from 性视界 University, including 性视界 Businesss School AI Institute Associate Collaborators and , explore how visualization and LLMs can work together to make stories easier to navigate without flattening their complexity. They introduce Story Ribbons, an interactive system that uses LLMs to extract structured narrative signals鈥攃haracters, locations, themes, sentiment, and more鈥攆rom novels and scripts, then turn those signals into customizable storyline visualizations designed to support exploration while helping users calibrate trust in AI-generated insights.

Key Insight: Mapping the 鈥淪hape鈥� of a Story

Story Ribbons starts with a classic visualization idea, each character as a continuous path over time, but treats that form as a flexible interface rather than a fixed diagram. Characters become 鈥渞ibbons,鈥� where absence creates gaps, thickness can encode narrative importance, and color can even shift to reflect sentiment or custom attributes defined by the user. Crucially, the system supports 鈥渆xplanations on demand,鈥� [2] a feature that allows users to click on any data point and receive a justification from the AI, grounded in the text. This interactivity transforms the visualization from a static chart into a set of dynamic, testable lenses. 

Key Insight: AI and Iterative Correction

The research team discovered that simply prompting an LLM to analyze a story produces unpredictable and often flawed results. Their solution: a carefully engineered four-step pipeline that decomposes narrative processing into manageable subtasks, inspired by crowdsourcing workflows. The system first splits novels into chapters, then into scenes based on location changes. For each scene, the AI extracts summaries, conflict and importance ratings, sentiment scores, character lists, and supporting quotes from the text. Critical to the pipeline鈥檚 reliability are multiple 鈥渃orrection loops鈥� that catch and fix common LLM errors. When the system discovered that LLMs frequently hallucinate or modify character quotes, it added an exact string matching check. Another correction loop involved deploying a second LLM to consolidate duplicate references, such as when an initial pass fails to recognize that 鈥淛ane,鈥� 鈥淛ane Bennet,鈥� and 鈥淢iss Bennet,鈥� refer to the same person in Pride and Prejudice. Despite these pipelines, Story Ribbons users still identified limitations in the AI鈥檚 analytical capabilities, such as its ability to provide holistic analyses that synthesized across chapters. At the same time, users treated the AI as a 鈥減artner to bounce ideas off of,鈥� [4] using the tool to challenge their own interpretations, with clear applications in education, such as sparking classroom discussions or helping students find textual evidence for essays. 

Why This Matters

For business professionals and executives, this research is less about literature, and more about an AI playbook for turning messy, narrative-heavy domains into decision-grade insights. Strategy decks, customer interviews, incident reports, regulatory filings, even internal emails, all share the same obstacle: they鈥檙e rich in meaning but often hard to 鈥渟ee鈥� at scale. Story Ribbons offers a solution: use LLMs to extract structure, produce interactive interfaces and engaging visualizations, and design for trust with explanations and links to original sources. 

Bonus

As this article shows, reliability often doesn鈥檛 come from a single prompt, but engineered checks, layered methods, and system design. For another look at trust and what it takes to make large systems behave responsibly, check out Teaching Trust: How Small AI Models Can Make Larger Systems More Reliable.

References

[1] Yeh, Catherine et al., “Story Ribbons: Reimagining Storyline Visualizations with Large Language Models,” arXiv preprint arXiv:2508.06772 (2025): 1.  

[2] Yeh et al., 鈥淪tory Ribbons,鈥� 5. 

[3] Yeh et al., 鈥淪tory Ribbons,鈥� 2. 

[4] Yeh et al., 鈥淪tory Ribbons,鈥� 2. 

Meet the Authors

is a computer science PhD student at 性视界 University.

is Assistant Professor in the Department of English at 性视界 University.

is a PhD candidate in the Department of English at 性视界 University.

is a computer science and East Asian studies student at 性视界 University.

is Assistant Professor of Comparative Literature at 性视界 University.

is Gordon McKay Professor of Computer Science at the 性视界 John A. Paulson School of Engineering and Applied Sciences, and an Associate Collaborator at the HBS AI Institute.

is Gordon McKay Professor of Computer Science at the 性视界 John A. Paulson School of Engineering and Applied Sciences, and an Associate Collaborator at the HBS AI Institute.

Did you find this insight to be useful?

• Yes
• No

Would you recommend this insight to others?

• Yes
• No

Submit

The post AI Tools That Rewrite How We 鈥淪ee鈥� Stories appeared first on 性视界 Business School AI Institute.

Key Insight: Cyborgs – Fluid Collaborators

Cyborgs are the study鈥檚 most common type, 60% of participants, and they reflect Fused Co-Creation, where GenAI is woven throughout the workflow. The human still decides what they鈥檙e trying to accomplish, but GenAI often drives how the work gets executed: drafting, analyzing, generating options, and reworking outputs through back-and-forth interaction. This isn鈥檛 鈥榩rompt once, accept once.鈥� Cyborgs use interaction practices like modularizing (breaking tasks into steps), validating, adding data to rerun analysis, and pushing back when outputs conflict with their own view. 

Thanks to this human-AI integration, Cyborgs developed entirely new AI-related expertise, what the researchers call 鈥渘ewskilling,鈥� while maintaining their domain knowledge. But the study also flags a persistent risk: even with best practices like validation, GenAI can confidently reinforce the wrong path, meaning Cyborg fluency needs to include knowing when not to trust the conversation.

Key Insight: Centaurs – Strategic Selectors

14% of consultants worked as Centaurs, professionals who engaged in what the researchers call Directed Co-Creation. Like the mythical half-human, half-horse creature, these workers maintained a clear division of labor: humans decided both what needed to be done and how to do it, using AI selectively for specific support tasks. Centaurs used AI primarily through three practices: mapping the problem domain (asking AI for general information), gathering methods information (requesting specific techniques or formulas), and refining their own human-generated content. For example, one consultant asked GenAI, 鈥淗ow do I calculate the market size growth rate of some industry from 2013 to 2017?鈥� [3] After receiving the formula, they performed the calculation in Excel rather than delegating it to AI.

The payoff? Centaurs achieved the highest accuracy in business recommendations among all three groups. They developed and strengthened their domain expertise by treating AI as an intelligent search engine and writing assistant. However, Centaurs also faced a trade-off: while they upskilled in task-related capabilities, they didn鈥檛 develop new AI-related expertise. They remained cautious about changing their established workflows, with some expressing ethical concerns about taking credit for AI-generated work.

Key Insight: Self-Automators – Dangerous Delegates

The final group is the Self-Automator, comprising 27% of the participants. These professionals engaged in Abdicated Co-Creation, consolidating the entire problem-solving workflow into one or two interactions, copying all data into a single prompt, and accepting AI鈥檚 outputs with minimal engagement. 44% of this group accepted AI鈥檚 output without any modification, while the rest made only superficial edits. The researchers are careful here: abdication isn鈥檛 always bad. For routine tasks, tight deadlines, or problems that sit well within GenAI鈥檚 capabilities, full delegation may be efficient, and Self-Automators saw immediate productivity gains.

But when the work requires judgment鈥攆raming the real problem, deciding what鈥檚 important, evaluating competing narratives鈥攁bdication can hollow out the very capabilities that make professionals valuable. In the study鈥檚 framing, when you give up control over what you do, you rarely keep control over how you do it. These professionals developed neither domain expertise nor AI-related skills.

Why This Matters

For business leaders and executives, this research suggests AI implementation guidelines for strategic task alignment. When the cost of error is high, such as in financial forecasting or operational diagnostics, leaders should recommend Centaur behavior by requiring professionals to execute the core analysis themselves while using AI for targeted support. This will maximize accuracy while ensuring teams continue to deepen essential domain expertise. Conversely, for tasks requiring rhetorical flair, ideation, or stakeholder persuasion, the Cyborg mode should be encouraged. That style of iterative, integrated looping allows for greater creative extension and the development of cutting-edge AI fluency. Ultimately, the most valuable professional of the future might not be a fixed type, but an adapter thinker capable of toggle-switching between these modes. 

Bonus

This research builds directly on a core theme in HBS AI Institute research: understanding how AI reshapes not just productivity, but the nature of expertise and collaboration itself. To read more about how AI can functionally substitute for or augment human teammates, breaking down silos and enabling individuals to perform at team-level quality, check out The Cybernetic Teammate: How AI is Reshaping Collaboration and Expertise in the Workplace.

References

[1] Steven Randazzo et al., 鈥淐yborgs, Centaurs and Self-Automators: The Three Modes of Human-GenAI Knowledge Work and Their Implications for Skilling and the Future of Expertise,鈥� The Wharton School Research Paper, 性视界 Business School Working Paper No. 26-036 (December 08, 2025): 12.  

[2] Randazzo et al., 鈥淐yborgs, Centaurs and Self-Automators,鈥� 12.

[3] Randazzo et al., 鈥淐yborgs, Centaurs and Self-Automators,鈥� 26.

[4] Randazzo et al., 鈥淐yborgs, Centaurs and Self-Automators,鈥� 12.

Meet the Authors

is a doctoral candidate at Warwick Business School and collaborator at the Laboratory for Innovation Science at 性视界 (LISH).

is a Professor of Management at Warwick Business School (WBS) and a visiting faculty at 性视界 University, at the Laboratory for Innovation Science at 性视界 (LISH). She heads the Artificial Intelligence Innovation Network at WBS.

is the David J. McGrath Jr Professor of Management and Innovation, a Professor of Business Administration at the MIT Sloan School of Management. Her research focuses on helping knowledge workers and organizations develop and implement Predictive and Generative AI products, on-the-ground in everyday work, to improve decision making, collaboration, and learning.

is a postdoctoral researcher at 性视界 Business School. His research explores how human/AI collaboration reshapes knowledge work: the impact of AI on knowledge workers, its effects on team dynamics and performance, and its broader organizational implications.

is an Associate Professor at the Wharton School of the University of Pennsylvania, where he studies and teaches innovation and entrepreneurship, and examines the effects of artificial intelligence on work and education. Ethan is the Co-Director of the Generative AI Lab at Wharton, which builds prototypes and conducts research to discover how AI can help humans thrive while mitigating risks.

is Partner Value Creation & Portfolio Monitoring at Seven2.

is the Dorothy & Michael Hintze Professor of Business Administration at 性视界 Business School. He specializes in technology management, innovation, digital transformation, and artificial intelligence. He is also the Co-Founder and Faculty Chair of the HBS AI Institute and the Founder and Co-Director of the Laboratory for Innovation Science at 性视界 (LISH).

Did you find this insight to be useful?

• Yes
• No

Would you recommend this insight to others?

• Yes
• No

Submit

The post The Three Ways Professionals Work with AI – Which One Are You? appeared first on 性视界 Business School AI Institute.

Key Insight: Not Your Average Chatbot

Rather than simply exchanging text in a conversation as a chatbot would, agentic AI can plan, decide, and act across multiple steps at the user鈥檚 request. In the context of the Comet browser, this means the Comet assistant can navigate websites, click buttons, fill fields, and iterate towards a goal instead of simply responding with text. For example, when you ask an agent to 鈥渦nsubscribe me from all promotional emails I receive more than twice per month,鈥� [2] it doesn鈥檛 just tell you how, it actually searches your inbox, identifies the offending senders, and unsubscribes on your behalf. Given this emphasis on modifying external environments, they don鈥檛 classify all tool use as agentic, which helps focus attention on these new AI systems and capabilities as they move into use at work and in everyday life.

Key Insight: Agents Are Mostly Used for Utility and Knowledge Work

When the researchers introduced a hierarchical taxonomy spanning topics, subtopics, and tasks, clear patterns emerged about what people actually delegate to agents. Productivity and Workflow dominates at 36% of queries, with document editing, account management, and email management as the largest subtopics. Users also tend to stay within the same categories once they start delegating tasks in the short term, showing strong 鈥榮tickiness鈥� across personal, professional, and educational settings. When they do branch out, they are far more likely to shift toward productivity, learning, or media tasks. Over the longer term, a bigger query share gravitates toward productivity and learning-related tasks. As users repeatedly invoke agents for these categories of tasks, it suggests that agents do become part of cognitive workflows rather than one-off, simple tasks. 

Key Insight: A Personal Assistant for Personal Pain Points

Users deploy the agent to solve the specific friction points of their industry. Finance professionals are heavily focused on efficiency, dedicating 47% of queries to productivity tasks. Students are focused on utility, with 43% of tasks allocated to learning and research. In design and hospitality, it鈥檚 even easier to see how context-specific usage dominates, from media work for designers to travel planning for hospitality staff. Ultimately, the data shows that the agent is highly versatile and reflects the specific needs of its user. In an educational context, it is a specialized research engine while in a professional context, it becomes a multi-purpose assistant. Personal contexts account for over half of all query volume. The environments where agents operate reinforce this pattern: usage clusters tightly around a small set of platforms like Google Docs, email platforms, and LinkedIn.

Why This Matters

For business leaders and executives, this study serves as a critical signal amidst the noise of AI speculation. The data confirms that we are moving from an era of generative AI to agentic AI, and AI-powered browsers may provide the onramp. Operationally, start where tasks are frequent, where environments are concentrated, and where risk can be bounded through supervision. The shift in user behavior over time indicates that once employees hurdle the initial learning curve, these tools can become sticky, essential components of the digital workflow.

Bonus

To understand more about how agents fit into the evolution of AI from tool to teammate, check out When Software Becomes Staff.

References

[1] Jeremy Yang et al., 鈥淭he Adoption and Usage of AI Agents: Early Evidence from Perplexity,鈥� arXiv preprint arXiv:2512.07828 (2025): 2.  

[2] Yang et al., 鈥淭he Adoption and Usage of AI Agents,鈥� 6.

[3] Yang et al., 鈥淭he Adoption and Usage of AI Agents,鈥� 17.

[4] Yang et al., 鈥淭he Adoption and Usage of AI Agents,鈥� 22.

Meet the Authors

is an Assistant Professor of Business Administration at 性视界 Business School and affiliated with the HBS AI Institute.

is a Data Scientist at Perplexity.

is an UX Researcher at Perplexity.

is Co-Founder and CTO of Perplexity.

is Co-Founder and Chief Strategy Officer at Perplexity.

is VP Global Affairs & Deputy CTO of Perplexity.

Did you find this insight to be useful?

• Yes
• No

Would you recommend this insight to others?

• Yes
• No

Submit

The post The Agentic AI Reality Check appeared first on 性视界 Business School AI Institute.

Key Insight: Seeking Predictions While Avoiding Explanations

In the main experiment, participants acted as loan officers for a private U.S. lender deciding how to allocate a real, interest-free $10,000 loan between two unemployed borrowers. An AI classified one borrower as low risk and the other as high. Participants could see the AI鈥檚 prediction and, in many conditions, they could choose whether to see an explanation of how the model reached its risk assessment.

Roughly 80% of participants opted to see the risk scores, but only about 45% chose to see explanations when given the chance. When their bonus was aligned with the lender (they earned more if loans were repaid), participants were more likely than others to seek the prediction, but significantly more likely to avoid explanations, especially when they were told those explanations could involve race and gender. In one condition that made fairness auditing salient, lender-aligned participants were about 10 percentage points more likely to skip explanations than neutrally paid participants. 

Crucially, this avoidance wasn鈥檛 about disliking extra information in general. When demographic information was removed and replaced with arbitrary details, the gap in explanation-avoidance between incentive conditions almost vanished. People weren鈥檛 shunning explanations as such, they were avoiding what the explanation might force them to confront about discrimination and their own profit-maximizing behavior.

Key Insight: Systematic Underevaluation

To separate moral self-image from pure decision quality, a second experiment removed fairness trade-offs and focused on prediction accuracy. Participants evaluated a loan labeled “high risk” by an AI, potentially due to a two-year employment gap. They first stated their willingness to pay (WTP) for an explanation revealing whether the gap was the driver of the high risk label. Crucially, participants then received free private information explaining that the gap actually resulted from pursuing a full-time professional certificate (benign towards risk), and not a termination (increasing risk) as would commonly be assumed. This private information made the purchased explanation more valuable, a concept the paper calls 鈥渃omplementarity,鈥� because if participants knew that the high-risk AI label resulted from the employment gap, then the addition of the private information told them that the AI label was not to be trusted. In other words, the participants should integrate the private information with the explanation to form a more accurate assessment.

Yet, when WTP was elicited a second time, after participants received this related private information, valuations dropped 25.6%. Valuations only increased (by 23.7%) when participants were explicitly guided through the complementarity logic. This represents a novel behavioral bias: people systematically fail to recognize when explanations would help them integrate their own knowledge with algorithmic outputs. 

Why This Matters

For business professionals and executives, this research is a warning that deployment of AI is not purely a technical challenge, it鈥檚 also a behavioral one. In high-stakes decisions like credit, hiring, pricing, healthcare, and safety, your employees could eagerly consume AI predictions while quietly avoiding the explanations that would expose uncomfortable trade-offs or discriminatory patterns. That avoidance can skew outcomes, undermine fairness, and create hidden risk. At the same time, teams may systematically under-invest in explanations even when they would improve forecasting by helping experts combine their own domain knowledge with AI outputs. The bottom line: investing in transparent AI systems is insufficient. You must also architect the decision environment and incentive structures that ensure transparency gets used rather than ignored.

Bonus

If you鈥檙e interested in how explanation avoidance fits into a broader pattern of human and AI collaboration challenges, Persuasion Bombing: Why Validating AI Gets Harder the More You Question It shows that when professionals do try to validate model outputs, AI can respond by pushing back and working to persuade users to accept its answers. Or if you鈥檙e thinking about the governance implications of explainable AI, Evidence at the Core: How Policy Can Shape AI鈥檚 Future argues that regulators should insist on robust evidence and transparency, from pre-release evaluations to post-deployment monitoring, so that organizations can鈥檛 simply offer explainability features on paper while leaving them unused in practice.

References

[1] Chan, Alex, 鈥淧reference for Explanations: Case of Explainable AI,鈥� 性视界 Business School Working Paper No. 26-028 (December 5, 2025): 2.  

[2] Chan, 鈥淧reference for Explanations,鈥� 2. 

[3] Chan, 鈥淧reference for Explanations,鈥� 7.

Meet the Author

is Assistant Professor of Business Administration at 性视界 Business School and HBS AI Institute Associate. He is an economist interested in how market failures occur, how such failures lead to divergence in economic outcomes, and how to design incentives and engineer markets to remedy these market failures.

Did you find this insight to be useful?

• Yes
• No

Would you recommend this insight to others?

• Yes
• No

Submit

The post Explanations on Mute: Why We Turn Away From Explainable AI appeared first on 性视界 Business School AI Institute.