Virtual panel: How software engineers and team leaders can excel with artificial intelligence

Introduction

Artificial intelligence is now generally available and is being used by many software developers in their daily work. It is not only impacting individual work, but also the way that professionals work together in teams and how software teams are being managed.

In this panel, we’ll discuss how artificial intelligence is reshaping the way that software is being developed, and what mindset and skills are required for software developers and engineering leaders to become adaptable and resilient in the age of AI.

The panelists:

• Courtney Nash – Internet Incident Librarian & Research Analyst at The VOID
• Mandy Gu – Senior Software Development Manager @Wealthsimple
• Hien Luu – Sr. Engineering Manager @Zoox | Author of MLOps with Ray

InfoQ: How has the rise of artificial intelligence impacted the way that software is being developed?

Courtney Nash: From what we hear in the media and product pitches, AI is making development seemingly quicker and more productive (though the jury is still out on this objectively), but in doing so it is adding unforeseen complexity and the likelihood of unexpected surprises later on. This addition of complexity is in part due to our inability to peel off the top of the AI black box and see how or why it’s doing what it’s doing. We can’t inspect how an AI arrived at the code or solutions that it did, and AI tools can’t model the broader complexity of systems, with which they may interact without awareness.

This knowledge is most critical when things don’t go as planned. When AI-generated software fails, how will we know where to look, or what to investigate when trying to stop the bleeding and get things back up and running and learn from what happened and feed that back into the system?

When it comes to AI and automation in software systems, my research focuses mainly on our own mental models of these tools. This research tends to view AI as a way to replace human work, rather than supporting and augmenting it. These mental models create unrealistic dichotomies (“Machines are better at these tasks/Humans are better at those tasks”) that don’t reflect the realities of software development for today’s modern complex systems. Research from other domains has shown that automation (and now, AI) is built on a “substitution myth”, which stems from the belief that people and computers have fixed strengths and weaknesses, and therefore all we need to do is give separate tasks to each agent (computer/person) according to their strengths.

As long as software development and AI designers continue to fall prey to the substitution myth, we’ll continue to develop systems and tools that, instead of supposedly making humans lives easier/better, will require unexpected new skills and interventions from humans that weren’t factored into the system/tool design (Wrong, Strong, and Silent: What Happens when Automated Systems With High Autonomy and High Authority Misbehave?, Dekker & Woods, 2024).

Mandy Gu: A lot more code is being written by AI (or with AI assistance). Anthropic’s CEO predicted that AI generated code will account for ninety percent of the code being written within the next six months.

On one hand, this change could be a huge productivity boost for developers, potentially accelerating timelines for software delivery and reducing development cost. With the time they get back, developers can focus on high-level design, architecture, and more complex problem-solving. On the other hand, companies and organizations will need to make sure they have the right checks and guardrails so that code quality standards are still being met. There will also be a shift towards better documentation and stronger contextual awareness to take advantage of these tools.

Hien Luu: Software development covers a lot of ground, from understanding requirements, architecting, designing, coding, writing tests, code review, debugging, building new skills and knowledge, and more. AI has now reached a point where it can automate or speed up almost every part of the process.

This is an exciting time to be a builder. A lot of the routine, repetitive, and frankly boring parts of the job, the “cognitive grunt work”, can now be handled by AI. Developers especially appreciate the help in areas like generating test cases, reviewing code, and writing documentation. When those tasks are off our plate, we can spend more time on the things that really add value: solving complex problems, designing great systems, thinking strategically, and growing our skills.

The recent advancements of AI coding agents have pushed them far beyond simple autocomplete. Tools like Cursor, Claude Code, and others are becoming standard components of the modern developer’s toolkit. However, developers still need to provide oversight, making sure the generated code meets quality standards, does not introduce new bugs, and is secure. Careful review, solid testing, and good test coverage are still non-negotiable requirements.

InfoQ: What skills do software developers need to cope with the challenges and exploit the opportunities of artificial intelligence?

Courtney Nash: First and foremost, they need to be empowered to trust their hard-earned knowledge and expertise in the face of AI tools that often evade direct introspection or a clear explanation of how the model works. In order to cope with the glut of AI tools and models, they’ll need knowledge of the domain question so they know when the model is not working as intended (or worse, is hallucinating). In particular they’ll need a clear understanding of how the model operates, and what the model is trained on. They’ll need time to build experience with the model and how to work effectively with it, e.g., whether to handhold it in small steps or make it respond to everything at once.

In their recent paper “Measures for explainable AI: Explanation goodness, user satisfaction, mental models, curiosity, trust, and human-AI performance”, researchers Robert R. Hoffman, Shane T. Mueller, Gary Klein, and Jordan Litman propose a trust scale for AI that might be helpful for developers and their teammates when it comes to assessing how confident they are with a given AI tool or model:

• I am confident in the [tool]. I feel that it works well.
• The outputs of the [tool] are very predictable.
• The tool is very reliable. I can count on it to be correct all the time.
• I feel safe that when I rely on the [tool] I will get the right answers.
• The [tool] is efficient in that it works very quickly.
• I am wary of the [tool]
• The [tool] can perform the task better than a novice human user
• I like using the system for decision making

Ultimately, a skill that will be essential for developers, and difficult to replace with AI, will be knowing how to detect and listen to the subtle signals that point at a system (or a team with AI involved) working suboptimally. For example, have you been using LLMs/AI to paste over areas of friction or tooling gaps which you don’t revisit and end up glossing over, and how will you know that’s happening?

Mandy Gu: The elephant in the room is “whether AI will take over my job one day?”. Until this year, I always thought no, but the recent technological advancements and new product offerings in this space are beginning to change my mind. The reality is that we should be prepared for AI to change the software development role as we know it.

To adapt to these changes, software developers should embrace AI as a tool. As AI becomes a more effective tool, so will the benefits of learning how to use it:

• A rudimentary understanding of prompt engineering, the dos and don’ts can go a long way.
• Every software developer should try an AI code assistant at least once.

However, we should also be aware of the pitfalls and risks of using AI. How do we review and guarantee that the code written by an AI is held to the same quality standards as for a human? What do we do if an AI assistant asks for a secret to assist with debugging?

Lastly, software developers should lean in to the critical thinking abilities that make us human. As AI makes coding more accessible, a developer’s impact will shift towards architecture design and translating business problems into technical requirements (as opposed to purely focusing on execution).

Hien Luu: The skills software developers need to thrive in the age of AI span three main areas: AI technical skills, systems thinking, and soft skills. To stay competitive, developers should evolve beyond a “T-shaped” profile, deep expertise in one area with broad general knowledge, toward a “Pi-shaped” profile, with depth in multiple areas and the ability to bridge them effectively.

A solid understanding of AI/ML concepts, especially how large language models (LLMs) are trained, how they behave, and where their limitations lie, is becoming essential. Knowing the strengths of LLMs and their weaknesses, such as bias and hallucinations helps developers use them effectively while guarding against errors. One particularly valuable skill is prompt engineering, the ability to clearly and precisely communicate intent to AI systems. Developers who master this communication will be more productive, more effective, and better equipped to build AI-powered applications. As Andrew Ng said at the Interrupt conference in May 2025, “The ability to tell a computer exactly what you want it to do will be a crucial skill for developers”.

While AI is excellent at repetitive software development tasks like coding, test writing, and code reviews, it struggles with higher-level systems thinking, system design, architecture decisions, and solving complex problems. These areas are now more valuable than ever, and developers who invest in strengthening their skills here will increase their career resilience and amplify their value in an AI-driven world.

Soft skills, particularly critical thinking and problem analysis, are also crucial. The ability to break down complex issues, apply logical reasoning, and weigh trade-offs allows developers to evaluate AI-generated output with a discerning eye. These soft skills are key to maintaining quality, preventing subtle bugs, and avoiding the accumulation of technical debt. In short, these skills act as a safeguard against the risks of being overly reliant on AI.

InfoQ: How can engineering leaders leverage software teams in using techniques and tools based on artificial intelligence in their daily work?

Courtney Nash: Leaders supporting teams that are using or adopting AI must acknowledge and invest in the hard-earned human expertise that their employees possess. Instead of viewing AI as a replacement for perceived human weaknesses, they can build what are called Joint Cognitive Systems (JCS), which provides a new view of how computers and people can not only co-exist, but support each other’s work in novel and advantageous ways.

In their 2002 paper “MABA-MABA or Abracadabra? Progress on Human–Automation Co-ordination”, Dekker and Woods identified three critical aspects of a JCS:

Mutual Predictability

In highly interdependent activities like software development and operations, planning our own actions (including coordination actions) becomes possible only when we can accurately predict what others (including automation/AI) will do. Skilled teams become mutually predictable through shared knowledge and coordination activities developed through extended experience in working together. How will AI fit into those activities?

Directability

For AI to be a good team member in a JCS, it must also be directable. Directability refers to the capacity for deliberately assessing and modifying actions as conditions and priorities change. Effective coordination requires AI and humans to provide adequate responsiveness to the others’ influence as the work unfolds, and especially when the unexpected happens.

Common Ground

Effective coordination requires establishing and maintaining common ground, including the pertinent knowledge, beliefs, and assumptions that the involved parties share (see my answer to the next question for more on this topic!). Common ground enables everyone to comprehend the messages and signals that help coordinate work. Team members must be alert for signs of possible erosion of common ground and take preemptive action to forestall a potentially disastrous breakdown of team functioning.

Leaders will have a real challenge, because most AI systems are not designed with these considerations in mind, so they will have to be creative and flexible to support their team working with an AI that is incapable of true joint cognitive work.

Mandy Gu: Engineering leaders should encourage adoption of AI tools and make it easy for software developers to try out these tools. Instead of waiting for the industry to align a winning tool, we should move quickly on reversible decisions.

Leaders should also make sure the secure way is enabled by default, with the right configurations, checks and balances being set from day one. In addition, engineering leaders need to invest in education and training to help their teams leverage this new technology effectively.

Lastly, leaders should continue to build a culture of writing clean code that is simple to understand, which will go a long way for humans and AI alike.

Hien Luu: Engineering leaders need to take an active role in guiding their teams through the AI transformation. This starts with clear communication of expectations, establishing a strong AI governance framework, and implementing comprehensive measurement systems.

Leaders should emphasize that AI is there to augment, not replace, human developers, and that the productivity gains it brings should be reinvested into higher-value, creative, and critical-thinking tasks, not simply used to increase workload.

A robust AI governance framework is essential to avoid security and compliance pitfalls. It should define clear guidelines for AI tool usage, outline adoption criteria, and prevent the uncontrolled spread of unvetted tools within the organization.

Measurement should also be holistic. In addition to traditional metrics like utilization, code acceptance rates, and developer satisfaction, leaders should track indicators of long-term health such as code quality, test coverage, maintainability, and feature delivery velocity. Good tracking ensures AI adoption is driving sustainable productivity and not introducing hidden technical debt.

InfoQ: What can companies do to cultivate a culture of resilience and enable their software developers to thrive in chaos and uncertainty?

Courtney Nash: This question can’t possibly be answered properly in a few paragraphs, but I’ll do my best and suggest some further reading and resources for people to dive further into the answer. I’ll start with a few definitions.

Resilience is the opposite of brittleness; it is the ability to bounce back, to adapt and respond when situations go awry and exceed known solutions, all without breaking down or experiencing catastrophic failure. Resilience is neither reliability, delivering the same outcome every time, nor redundancy, backups and similar methods for effectively supporting reliability.

As system safety and human factors researcher Dr. David Woods has wisely said, “Resilience is a verb”.

MIT Professor Edgar Schein defines organizational culture as “the pattern of basic assumptions that a given group has invented, discovered, or developed in learning to cope with its problems of external adaptation and internal integration, and that have worked well enough to be considered valid, and, therefore, to be taught to new members as the correct way to perceive, think, and feel in relation to those problems”. (Organizational Culture and Leadership, Edgar Schein, 2010.)

In this regard, culture is a collection of values (derived from assumptions and experience) upon which its constituent members act. Like resilience, it is active. Existing members of a culture must be explicit and transparent about what those values and associated behaviors are, and new members must be actively taught these values and behaviors by those existing members. Culture is alive. It is not a handbook or a one-off onboarding training video. This transparency is a critical piece of building a sustainable culture that includes AI: how will leaders factor AI into a living, changing, active culture?

Combining these two views, cultivating a culture of resilience hinges on investing in expertise. It succeeds when leaders trust their employees and give them autonomy to express their expertise and disseminate it throughout the organization without fear of blame or retribution. A culture of resilience celebrates learning from failure.

Some suggested further reading on this topic includes:

Mandy Gu: In response to the uncertainty introduced by AI, companies should make sure we embrace these technologies safely and securely. Make good data security practices easy with the right checks and balances, and leverage GenAI deployment options offered by their cloud providers (e.g., Bedrock for AWS) to ensure data from AI integrations stays within your cloud tenant.

Companies should also be transparent about their AI strategies to address potential anxieties and give their developers the space and resources to continuously learn and upskill in these changing technological environments.

Hien Luu: A culture of resilience starts with psychological safety and open communication, creating an environment where developers can share their learning journeys and mistakes, express ideas, ask questions, and voice concerns without fear of judgment or retaliation.

Leaders play a key role here. When they openly share their own challenges and learning curves with AI, it signals to the team that vulnerability is not only acceptable but valued. Hack days, regular developer meetups, and other forums for sharing lessons learned (including mistakes) help developers see they’re not alone in navigating change. These practices foster connection, mutual support, and the confidence to experiment, even in times of uncertainty.

InfoQ: What’s needed to keep software development sustainable and ensure the mental well-being of software developers and team leaders?

Courtney Nash: System safety researcher Sidney Dekker notes a few key aspects of organizations that foster resilience and a culture that best supports the people doing that work. These organizations:

• Never take past success as a guarantee of future success.
• Keep risk discussions alive even when everything is fine and dandy.
• Continuously update their mental models of how their system(s) work.
• Actively consider diverse inputs. They seek minority opinions, and stay open minded about what they hear in those opinions.
• Unilaterally allow individuals to “stop the production line” without consequence if they feel something is going to go awry.

The other research I always point people towards, when the topic of sustainable work, mental well-being, and burnout comes up, is from Dr. Christine Maslach. She’s a leading expert on burnout in a variety of industries, and has spoken at a number of tech conferences over the past decade. Her research captures key areas in which leaders should invest to keep work sustainable.

Maslach identified six main strategic areas that leaders need to invest in to keep work sustainable and avoid burnout:

• A manageable workload
• Providing people with agency or control over their work
• Ensuring people feel appropriately rewarded for the work they do
• A sense of belonging to a community
• Fairness in the work environment
• Ensuring people feel their work is aligned with the shared values of the organization

A mismatch along any of those axes can lead to burnout down the road. The more mismatches there are, the more likely it’s going to happen. Having alignment along all of these is like having “money in the bank” for when people do need to stretch and work a bit harder when called for. Factoring AI into this model is a new frontier that is largely ignored in the gold rush to new products, models, and computing power.

AI poses a unique set of challenges to all six of these areas of investment for leaders and their teams. Leaders who work towards helping AI and software developers co-exist as “team players” will be more likely to have sustainable, higher performing teams than those who view AI as a substitution for human performance and expertise.

Mandy Gu: Engineering leaders need to address any AI anxieties that may be lingering. Leaders need to be transparent about AI strategies and their expectations, and make it easy for anyone in the company to share feedback about these strategies.

Companies will also need to give teams space to learn and adapt to these new technologies, and lay the foundations to leverage AI effectively. In some cases, companies may also need to reposition their productivity metrics to reflect the work being completed. While it’s tempting to conflate a ninety percent reduction in code being written as cutting delivery time by the same amount, there is so much more that goes under the tip of the iceberg than just writing the code.

Hien Luu: Clear communication about what AI can and cannot do is essential. When AI is positioned as a powerful tool to augment developers rather than replace them, it helps reduce anxiety about job security and fosters greater acceptance and adoption.

Providing accessible, ongoing training that fits into a developer’s busy schedule is equally important. Pairing training with regular “office hours” or open Q&A sessions creates a safe space for learning and troubleshooting. Together, these measures help ease feelings of being overwhelmed, support continuous growth, and keep software development sustainable in the long run.

Conclusions

Using artificial intelligence, software can be developed more quickly. Productivity can also be increased, as repetitive parts can be handled by AI. But AI can add unforeseen complexity and increase the likelihood of unexpected surprises later on. Developers still need to provide oversight, making sure the generated code meets quality standards. Checks and guardrails need to be in place to ensure that code quality standards are being met.

To cope with the challenges and exploit the opportunities of artificial intelligence, software developers need knowledge of the domain question so they know when the model is not working as intended, or hallucinating. To embrace AI as a tool, software developers need to understand AI/ML concepts and prompt engineering. However, they should also be aware of the pitfalls and risks of using AI. Critical thinking and problem analysis are crucial to evaluate AI-generated output.

Engineering leaders can leverage software teams by considering AI and teams as joint cognitive systems and supporting each other’s work in novel and advantageous ways. They should encourage adoption of AI tools and make it easy for software developers to try out tools, and continue to build a culture of writing clean code that is simple to understand. Leaders can establish an AI governance framework, and implement comprehensive measurement systems to support usage of AI.

To cultivate a culture of resilience, companies should invest in expertise. A culture of resilience starts with psychological safety and open communication. Leaders should trust their employees and give them autonomy to express their expertise and disseminate it throughout the organization without fear of blame or retribution. Companies should be transparent about their AI strategies to address potential anxieties and give developers space and resources to continuously learn and upskill in these changing technological environments.

To ensure the mental well-being of software developers, engineering leaders need to address any AI anxieties that may be lingering, and give teams space to learn and adapt to new technologies. When AI is positioned as a powerful tool to augment developers rather than replace them, it helps reduce anxiety about job security and fosters greater acceptance and adoption. Leaders who work towards helping AI and software developers co-exist as “team players” will be more likely to have sustainable, higher performing teams than those who view AI as a substitution for human performance and expertise.