The Baruch Group

09/06/2026

Beyond the Hype: Top 5 Cybersecurity AI Use Cases With Real Impact
AI is reshaping cybersecurity — here are the five use cases that deliver results.
August 28, 2026

Top 5 Real-World Cybersecurity Use Cases for AI
Focus on proven AI use cases in cybersecurity
As AI-driven tools change how organizations detect, respond to and prevent cyberthreats, CISOs face pressure to distinguish proven AI use cases from overhyped promises. Gartner finds that 55% CISOs report they have faced costly workarounds from vendor overpromises in AI capabilities and technologies, and 43% feel that AI-assisted tools have failed to deliver promised outcomes. As Charlie Winckless, Gartner Vice President Analyst notes, “Focusing efforts on the most effective uses of modern AI technologies is critical amid the incessant hype around them.” CISOs see the strongest results when they apply AI in targeted ways.

Optimize existing tool usage via embedded AI assistants.
Embedded AI assistants can provide configuration guidance and improve event visibility and understanding. By integrating AI into existing security tools, organizations can maximize the value of their current investments and ensure more effective threat monitoring and response. Use copilots to boost team effectiveness with existing tools and simplify onboarding for new tools. Promote their use to support decision making and uncover new capabilities.

Enhance security operations with AI SOC agents.
AI can automate workflows, prioritize incidents and enrich data within the security operations center (SOC). This enables SOC teams to respond more quickly and efficiently to potential threats, reducing manual effort and improving overall incident management. Larger teams benefit most from augmentation, so match your scale and use cases — focusing on enrichment and proactive alert processing — rather than relying on a single vendor or approach.

Develop more secure code with AI code security assistants.
AI code security assistants (ACSAs) help find security issues during the development pipeline for custom code and provide guidance on mitigating them.

By integrating AI-powered tools into software development, organizations can proactively address vulnerabilities and strengthen application security from the outset. Use ACSAs with regular application security testing (AST) and LLM-based vulnerability tools — AST and LLMs find issues, while ACSAs guide remediation.

Manage third-party questionnaires.
AI can assist in filling and processing third-party cybersecurity assessment questionnaires. This streamlines the evaluation process for third-party vendors, reducing administrative burden and helping organizations maintain compliance with security standards. While AI tools can draft and process security questionnaires, humans must verify results and control models due to sensitive information.

Improve access to policies and standards with policybots.
AI improves accessibility to cybersecurity policy libraries through natural language queries. This allows users to easily search and understand relevant policies, supporting better compliance and awareness across the organization. Integrate LLMs with messaging systems and policy libraries, starting with a limited set and expanding gradually.

09/06/2026

AI Is Revolutionizing Strategic Decision-Making
New tools can improve human judgment by tirelessly generating, evaluating, and synthesizing insights. by Felipe A. Csaszar
https://bit.ly/3US0KMG
From the Magazine (September–October 2026)

Dimitris Ladopoulos
Summary. For decades, the limits of time and brain capacity meant that teams could consider only so many strategic options before making decisions. Strategy tools—SWOT analyses, portfolio matrices—were simple because that’s what planning meetings needed. AI changes those dynamics. It can generate and evaluate thousands of strategic options and build rich, continually updated views of markets and competitors. And it can stress-test potential plans through structured debate that isn’t influenced by internal politics. Since the same AI tools are available to all companies, lasting advantage will go to those that pair them with proprietary data, integrated workflows, and faster ex*****on. In practice, that means casting a wider net for strategic options before narrowing the field, replacing static models with real-time ones, and making AI-assisted challenges a routine part of major decisions.

Think about your last strategy offsite. Your team spent weeks preparing. You flew people in, booked the conference room, hired a facilitator. And after two days of debate, how many truly different strategic options did you walk out with? Three? Four? Now ask yourself: Was that because only three or four good options existed, or because that’s all your team had the time and mental bandwidth to develop and evaluate?

For most companies, the honest answer is the latter. The bottleneck in strategic decision-making has never been a shortage of possible directions. It has been the limited capacity of the human minds doing the work. We can hold only so much information in our heads, evaluate only so many alternatives in a strategic-planning cycle, and process only so many perspectives in a meeting before fatigue, politics, or the clock forces a decision. Scholars call this dynamic bounded rationality—the idea that human decision-makers, however capable, are constrained by finite attention, memory, and processing power.

These constraints are so fundamental that we rarely notice them, but they have quietly shaped every tool in the standard strategy playbook. The reason a SWOT analysis has four quadrants, a growth share matrix is a 2×2, and Michael Porter’s most famous framework has exactly five forces is not that the competitive world is actually so simple. It’s that the frameworks had to be simple enough for a human team to map them out on a whiteboard in a few hours.

For decades, those frameworks were the best we had. That’s no longer the case.

The current generation of artificial intelligence tools—particularly large language models (LLMs) and the multiagent systems being built on top of them—are not just additions to the planning tool kit. They’re technologies that directly relax the cognitive constraints that have shaped how companies make their most important decisions. AI can generate and screen thousands of strategic alternatives where a human team might be able to consider a mere handful. It can build and continuously update models of markets, customers, and competitors that are far richer and more dynamic than any static framework. And it can test ideas through simulated deliberation—synthesizing diverse perspectives and challenging assumptions without the groupthink, hierarchy, and time pressure that distort real-world strategy meetings.

In short, AI offers a path to unbounding the strategy process. The limits are not removed, of course, but they’re pushed outward—and where they end up matters enormously, because the implications go well beyond efficiency. When you can explore more options, you find better paths forward. When you can model your environment in higher resolution, you see more opportunities and threats. When you can stress-test a plan by simulating competitors, skeptical customers, and a devil’s advocate who never gets tired, you make decisions that are more resilient. The companies that master this new way of working won’t just do strategy faster. They’ll do it better—and build the next generation of competitive advantage.

For decades, the limits of time and brain capacity meant that teams could consider only so many strategic options before making decisions. Strategy tools—SWOT analyses, portfolio matrices—were simple because that’s what planning meetings needed. AI changes those dynamics. It can generate and e...

09/05/2026

4 Steps to Transform the “Middle Office” with AI
by H. James Wilson, Chetna Sehgal, Michael Zimmerman, Ali Arsanjani, Blaise Abderholden and Jimmy Priestas

August 20, 2026
Yaroslav Kushta/Getty Images
Summary. Companies are spending heavily on AI, but many are still struggling to turn that investment into measurable business results. One reason is that they’re overlooking the middle office, where work such as contract review, compliance, risk management, and accounts payable still requires a lot of human judgment. Accenture and Google research suggests this is where some of the biggest near-term opportunities lie. Companies can capture them by finding processes where difficult cases still get handed to experts, building AI systems that learn from those experts as they work, and measuring whether AI actually reduces those escalations. The final step is making sure the time AI saves gets put to good use, freeing experts to focus on complex decisions, relationships, and higher-value work.
Recent evidence on AI adoption has revealed a sobering paradox. Economists who linked large-scale adoption surveys to administrative payroll records found that although most employers had rolled out AI tools in areas where AI should have the highest probability of impact, the effect on earnings and hours two years on was statistically indistinguishable from zero. Despite capable technology and workers reporting productivity gains, the promised benefits failed to materialize on the bottom line.

Why? The researchers found that the time the tools saved was eaten by the work of running them. Employees spent their “freed up” time responding to the model’s output and manually wiring it into existing systems and decisions. The gains never surfaced as earnings or reclaimed hours because the models were not truly integrated. The companies lacked the integration that turns them into results.

Based on our proprietary cross-industry analysis and structured experiments in live operations over the past year, we (Accenture and Google) have identified four steps leading companies take to close the gap between adoption and benefit and transform the middle office. This gap between AI adoption and benefit is widest and most costly in the middle office, the largely overlooked domain between so-called front-office customer-facing functions and back-office routine administration. This is where companies handle non-routine, exception-heavy work such as contract reviews, risk management, and compliance. Decisions often depend heavily on human judgment to resolve ambiguities as business conditions change. And because, according to our analysis, the middle office accounts for more than four-in-10 working-hour tasks across 18 industries, it represents the enterprise’s largest single layer of work and one of its biggest untapped sources of AI value.

Yet most AI roadmaps jump from the back office, where automation is mature and the gains are thinning, straight to ambitious front-office bets like dynamic pricing or AI-assisted dealmaking, whose payoff always seems a year away. The middle office, holding the plurality of the work and the nearest large returns, gets passed over.

That’s a mistake. The key to closing the adoption-to-benefit gap is improving intelligence integration: the work of turning a capable model into a profitable one by wiring it into how decisions actually get made. The results can be dramatic. For instance, in a live experiment we conducted, improving intelligence integration in a typical middle-office workflow lifted overall success from 42 percent to 80 percent. Resolution of the hardest cases rose from 36 percent to 99 percent, while freeing experts for higher-value work. To help your organization achieve stronger process performance and benefit from humans taking on higher-value work, we recommend the following four steps https://bit.ly/4wRqAhh

Companies are spending heavily on AI, but many are still struggling to turn that investment into measurable business results. One reason is that they’re overlooking the middle office, where work such as contract review, compliance, risk management, and accounts payable still requires a lot of huma...

09/05/2026

Reckitt’s transformation: Three moves, one strategic reset

Reckitt didn’t sequence change. It did three moves at once: transforming its core, separating noncore businesses, and resetting its operating model. That decisive reset cut business complexity, accelerated decisions, and repositioned the company for stronger growth, higher margins, and sustained performance.

Creating clarity amid complexity
In 2024, Reckitt leaders found the organization at a pivotal moment: It was operating in more than 70 markets while navigating industry disruption, fundamental decisions about which businesses to separate and where to focus, and the prospect of a simultaneous transformation and business carve-out. Geopolitical and supply chain disruptions had affected performance. Prior transformation efforts had not delivered the expected impact or pace of progress, while increased M&A activity added further pressure.

As Reckitt Chief Human Resources Officer Ranjay Radhakrishnan describes it: “The starting point was one of deep complexity and growing doubt. Performance was shaped by a series of one-offs, and that created a shadow over the underlying strength of the business.”

Ranjay Radhakrishnan
Performance was shaped by a series of one-offs, and that created a shadow over the underlying strength of the business.

Ranjay Radhakrishnan
Chief human resources officer at Reckitt
Organizationally, the operating model had also become unclear. “The organization was, quite literally, sitting on the fence, on governance, accountability, and where decisions really sat, and that made change unavoidable,” he explains.

Rather than layer another incremental program onto an already complex system, leadership faced a clear choice: Recalibrate slowly, or reset decisively. While many organizations may have sequenced their next chapter, Reckitt chose to act on three fronts simultaneously: separating noncore businesses, establishing a new operating model, and materially restructuring its cost base. Reckitt partnered with McKinsey from strategy through ex*****on to deliver the transformation.

The solution
Running transformation, separation, and operating model redesign in parallel
Against this backdrop, Reckitt’s leadership moved to provide what Radhakrishnan describes as “electrifying clarity on the direction, what was core, what was noncore, and what that meant for the organization and for individuals.”

The transformation required a set of hard business choices: where to play, what to exit, and how to fundamentally reset the cost and operating model to restore performance. This meant concentrating Reckitt on a set of high-growth power brands, separating noncore businesses, and redirecting capital, talent, and leadership attention toward fewer, higher-return growth engines—ultimately focusing the business on 11 core brands while preparing other parts of the portfolio for separation or alternative ownership.

This shift clarified where Reckitt would succeed, reduced competing priorities, and aligned capital and leadership focus behind a core set of growth engines.

Critically, Reckitt did not treat these portfolio choices as a prelude to transformation; it ran them as part of the same integrated reset. That parallel approach gave separation real momentum. It forced clarity on what the future core needed to look like and ensured that governance, processes, and productivity were redesigned for both the retained business and the assets being carved out, ultimately enabling the carve-out to be completed in record time.

Redesigning the operating model and cost structure
The operating model redesign focused on simplification and accountability to accelerate decision-making, reduce structural cost, and improve how the business delivers on its growth plans. Reckitt moved from a center-driven global business unit model to one where decision-making shifted toward geographic areas and their respective markets, bringing accountability closer to consumers and customers.

In practice, this translated into a fundamental shift in how the business operated, delayering the organization from five management layers to three. Reckitt established Global Business Services to serve as a platform for enterprise-wide transformation, driving end-to-end process excellence, accelerating automation and AI adoption, and unlocking sustainable value across the business. It also improved the cost of goods sold by redesigning its supply network across manufacturing, logistics, and procurement, embedding productivity as a structural feature of the operating model rather than a one-off cost program.

Clear roles and responsibilities further simplified governance and clarified decision rights, shifting ownership from a predominantly center-led model to one with greater market accountability. Global functions were realigned with a new geographic structure to support this shift.

09/04/2026

The Boundaries of Automation: A Theory of Persistent Human Participation
Fares Fourati, Hinrich Schütze, Eyke Hüllermeier, Iryna Gurevych
The rapid progress of AI has intensified the long-standing pursuit of automation: replacing human participation with algorithms wherever possible. Implicit in this pursuit is the assumption that humans remain in the loop only because current AI systems are not yet sufficiently capable. This paper challenges that assumption. Rather than asking how far automation can extend, we ask where its conceptual limits lie and argue that human participation may persist even with highly capable AI systems for three distinct reasons. Technical or complementarity grounds arise when humans contribute capabilities or perspectives unavailable to AI. Normative or developmental grounds arise when participation itself is valuable for human agency or learning. Most importantly, emergence grounds arise from target emergence: in some activities, the target is not fully specified in advance but instead emerges through the interaction itself. In these cases, human participation is not merely a means of improving ex*****on but is constitutive of the target being produced. Human--AI co-construction, understood as the joint production of outcomes by humans and AI systems, is therefore not simply a temporary response to imperfect AI, but a persistent feature of activities whose objectives emerge through participation. This perspective has important implications for the limits of automation and for the design, evaluation, and ethics of future AI systems.
Subjects: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Emerging Technologies (cs.ET); Machine Learning (cs.LG); Multiagent Systems (cs.MA)
Cite as: arXiv:2607.21547 [cs.AI]
(or arXiv:2607.21547v1 [cs.AI] for this version)

https://doi.org/10.48550/arXiv.2607.21547 https://arxiv.org/abs/2607.21547?mkt_tok=Mjk4LVJTRS02NTAAAAGj-2NPxstVsDLpbLcNRHPXFSNkQrnBdakDaYxl052mTD8GejP__6NHHkro01_u5U9HtzPYs4BajfVtiZhex3IHM1gta34TB5DmNCxPNDvu7k42UtQsSRps

Please feel free to contact me for cs.ai,or directly click here to make your best offer(查看报价直接点击这链接).

09/04/2026

What is human-in-the-loop?
Human-in-the-loop (HITL) refers to a system or process in which a human actively participates in the operation, supervision or decision-making of an automated system. In the context of AI, HITL means that humans are involved at some point in the AI workflow to ensure accuracy, safety, accountability or ethical decision-making.

Machine learning (ML) has made astonishing strides in recent years, but even the most advanced deep learning models can struggle with ambiguity, bias or edge cases that deviate from their training data. Human feedback can help both improve models and serve as a safeguard for when AI systems perform at insufficient levels. HITL inserts human insight into the “loop,” the continuous cycle of interaction and feedback between AI systems and humans.

The goal of HITL is to allow AI systems to achieve the efficiency of automation without sacrificing the precision, nuance and ethical reasoning of human oversight.

Benefits of HITL
Human-in-the-loop machine learning allows humans to provide oversight and input into AI workflows. Here are the primary benefits of human-in-the-loop:

Accuracy and reliability

Ethical decision-making and accountability

Transparency and explainability

Accuracy and reliability
The goal of automating workflows is to minimize the amount of time and effort humans have to spend managing them. However, automated workflows can go wrong in many ways. Sometimes models encounter edge cases that their training has not equipped them to handle. An HITL approach allows humans to fix incorrect inputs, giving the model the opportunity to improve over time. Humans may be able to identify anomalous behaviors using their subject matter expertise, which can then be incorporated into the model’s understanding.

In high-stakes applications, humans can impose alerts, human reviews and failsafes to help ensure that autonomous decisions are verified. They can catch biased or misleading outputs, preventing negative downstream outcomes. Continuous human feedback helps AI models to adapt to changing environments.

Bias is an ongoing concern in machine learning, and although human intelligence is known for being quite biased at times, an additional layer of human involvement can help identify and mitigate bias that is embedded into the data and algorithms themselves, which encourages fairness in AI outputs.

Ethical decision-making and accountability
When a human is involved in approving or overriding AI outputs, responsibility doesn’t rest solely on the model or its developers.

Some decisions require ethical reasoning that may be beyond the capabilities of a model. For example, an algorithmic hiring platform’s recommendations might disadvantage certain historically marginalized groups. While ML models have made major strides over the last few years in their ability to incorporate nuance in their reasoning, sometimes human oversight is still the best approach. HITL allows humans, who have better understanding of norms, cultural context and ethical gray areas, to pause or override automated outputs in the event of complex dilemmas.

A human-in-the-loop approach can provide a record of why a decision was overturned with an audit trail that supports transparency and external reviews. This documentation allows for more robust legal defense, compliance auditing and internal accountability reviews.

Some AI regulations mandate certain levels of HITL. For example, the EU AI Act’s Article 14 says that “High-risk AI systems shall be designed and developed in such a way, including with appropriate human-machine interface tools, that they can be effectively overseen by natural persons during the period in which they are in use.”

According to the regulation, this oversight should prevent or minimize risks to health, safety or fundamental rights, with methods including manual operation, intervention, overriding and real-time monitoring. The humans involved must be “competent” to do so, understanding the system’s capabilities and limitations, trained in its proper use and with authority to intervene when necessary. This oversight is intended to encourage the avoidance of harm and proper functioning.

Transparency and explainability
By catching errors before they cause harm, HITL acts as a safety net, especially in high-risk or regulated sectors like healthcare or finance. HITL approaches help to mitigate the “black box” effect where the reasoning behind AI outputs is unclear. Embedding human oversight and control into development and deployment processes helps practitioners identify and mitigate risk, whether that’s technical, ethical, legal or operational risk.
(IBM)

09/03/2026

Decision Making And Problem Solving
AI Is Undermining Leaders’ Judgment. Here’s What to Do About It.
by Leonid Sudakov and Nathan Furr

August 19, 2026

HBR Staff; AI; Guescri/Getty Images
Summary. As organizations gain more “intelligence” with AI tools, leaders are being trained out of the very capacity that creates competitive advantage—original judgment. Two interlocking capacities shape a leader’s ability to produce original thinking: breadth of perception (noticing weak signals and adjacent patterns beyond the obvious) and independence of interpretation (forming your own view rather than deferring to the model’s or the group’s). Organizations can protect judgement by adopting two practices: structured curiosity and intentional dissent.

A recent report from Harvard researchers studying innovation showcased a troubling pattern. To test how AI impacts human judgment, the researchers conducted a field experiment, asking 228 experienced evaluators to assess 48 submissions to an MIT global social impact innovation challenge. The researchers tested three conditions: 1) human-only evaluations; 2) LLM evaluations with a narrative explanation of the decision-making; and 3) LLM evaluations with no further context. They then compared the evaluators’ decisions against those made by four experts affiliated with the innovation challenge.

What they found was striking: When the AI tool encouraged human evaluators to reject submissions that the expert panel approved, the humans largely went along with those recommendations, filtering out high-potential innovations. What’s more, they less were likely to override an incorrect recommendation from the AI when offered a narrative explanation. Instead of improving the evaluators’ judgement, the narrative explanation made it worse.

This finding isn’t isolated. Recent large-scale research shows the intelligent tools meant to augment judgment are, in some cases, eroding the very capacity they support.

At a moment when judgement is becoming an ever more important competitive differentiator, that should cause leaders pause. Here’s what they can do to protect it.

The Original Judgment Paradox
For a decade, organizations have invested massively in decision-support infrastructure promising ever-better decisions. Yet a paradox is emerging: As organizations gain more “intelligence,” their leaders are being trained out of the very capacity that creates competitive advantage—original judgment. By original judgment we mean the uniquely human capacity to see beyond the dominant narrative and make choices that represent your values under uncertainty. It underlies nearly every story of competitive advantage—Sam Walton ignoring the wisdom that one cannot build a discount retailer in a town of fewer than 50,000 people, Steve Jobs integrating calligraphy into the personal computer, or Jensen Huang betting on unproven hardware emulation to radically accelerate prototyping at Nvidia.

Two interlocking capacities shape a leader’s ability to produce original thinking: breadth of perception (noticing weak signals and adjacent patterns beyond the obvious) and independence of interpretation (forming your own view rather than deferring to the model’s or the group’s). In our work on leadership judgment, we’ve come to see these not as personality traits or flashes of genius but as observable mechanisms explaining why some leaders spot inflection points and act with conviction while others, equally intelligent, default to compliance.

In short, original judgment is what separates strategic breakthroughs from operational efficiency—and as intelligent systems spread through boardrooms, that capacity is eroding precisely where it matters most. https://bit.ly/4gx6enl

As organizations gain more “intelligence” with AI tools, leaders are being trained out of the very capacity that creates competitive advantage—original judgment. Two interlocking capacities shape a leader’s ability to produce original thinking: breadth of perception (noticing weak signals an...

09/03/2026

What employers are looking for in the age of AI — and four ways to provide it
For applicants whose knowledge of AI stops at the name ChatGPT, Nature spoke to academics, employers and early-career researchers to find out what you need to expand your understanding.
By Ben Deighton

The world of artificial intelligence can be intimidating, but researchers have tips on how to get started. Credit: Jade Gao/AFP via Getty

Artificial intelligence is becoming a must-have skill for both employers and funders. Data compiled by the jobs site Indeed show that, in the United States, science jobs listing ‘AI’ as a required skill are rising sharply, whereas the overall number of science jobs has fallen (see ‘AI in demand’).
Source: Indeed

However, instead of worrying that AI systems will be taking science jobs, researchers can get ahead by using and interpreting AI tools to build their understanding. And they don’t need to know everything. Computer scientist Regina Barzilay runs a specialist course at the Massachusetts Institute of Technology (MIT) in Cambridge to teach scientists AI skills. She likens it to cooking: “You don’t need to learn every recipe on Earth to feel comfortable in the kitchen … but you need to have this very basic understanding,” she says, much like “you don’t need to be a computer scientist to use a computer.”

Nature spoke to nine recruiters and researchers to ask them what the most important skills will be for scientists of all kinds in the AI era.

In short
Dos and don’ts to build your AI knowledge:

• Do experiment with AI technologies in your working day.

• Do take a course in AI and machine learning.

• Do ask AI to write code for a project, then take the code apart and learn how it works.

• Don’t propose a project using AI tools without being able to show that you have the skills to back it up.

• Don’t believe everything you get from a chatbot.

• Don’t try to learn everything at once.

Be curious
Recruiters repeatedly told Nature that, rather than specific technical skills, they’re mostly looking for candidates who are curious to learn more about AI and are able to demonstrate that on their CVs.

“It’s the willingness to just roll up your sleeves and get into it, whether it’s informal training, whether it’s just spending some hours on open-source educational content,” explains Vijay Shah, dean of research at the Mayo Clinic in Rochester, Minnesota, which had almost 30 open research positions in May.

This is more important than specific skills, he says, because the field is moving so quickly. If you set a litmus test that requires people to be able to use AI tools such as OpenClaw or AlphaGenome, for instance, it will quickly be outdated, he adds.

Get to know how AI thinks
One pitfall when using AI tools is putting too much faith in the results that they generate without examining the data critically. “Often it’ll give you results that it thinks you want to see, based on something you’ve said,” says Walsh. “It’s a people pleaser.”

Something that needs careful checking, according to Barzilay, is any percentage likelihood given by AI, because it might not match the observed reality. For instance, if an AI model says that there is an 80% probability that a radiology image depicts cancerous cells, “What you need to do is to say, ‘can I trust this 80%?’”

This means recognizing whether the probability estimation has been calibrated against current test data to ensure its estimations match the real distributions.

It’s really important to get “a sense for the capabilities of the model, but also [to] develop a sense about what that means when it runs on a machine or when it interacts with data sets”, says Dominik Lukeš, a consultant at the AI Competency Centre at the University of Oxford, UK.

Because some people simply try a large language model and get “sort of wonky answers”, Lukeš thinks that it’s worth investing time in trying to understand how AI works. “There are people who’ve spent hundreds of hours learning how to take advantage of it, thinking about the process, rethinking their process and all of a sudden they will say [they] can’t live without it and ‘oh, this is an enormous game changer’.”

Know what you know
In an AI-enabled era, it feels like most research proposals are packed with ideas on how the technology can be used. However, proposal writers often don’t demonstrate that their team has the experience to implement its ideas, say hiring managers. Writing ‘we will use AI to achieve our goals’ isn’t especially credible without detail and experience to back it up. https://go.nature.com/4ccM8O0

For applicants whose knowledge of AI stops at the name ChatGPT, Nature spoke to academics, employers and early-career researchers to find out what you need to expand your understanding.