Product in the AI Era: AI Resource Guide
AI is changing how products are built, the shape of product teams, and how successful companies create value.
This AI Resource Guide attempts to bring together everything SVPG has published on AI – including articles, podcast episodes, and other recommended resources. It will continue to grow as we publish new insights.
AI has already impacted so much of how we create products, starting with how we build our products, but touching nearly every aspect of product including product strategy and product discovery. Yet the principles underlying strong product organizations endure.
As always, what matters in product is achieving business results – outcomes. Yet it’s no secret that while for most organizations the output has most definitely increased, the outcomes have not. This is referred to as the AI Productivity Paradox.
We are all about leveraging AI to accelerate and improve your chances of achieving the necessary outcomes.
The sections that follow cover the major ways AI is changing product teams – from strategy and product management to discovery, coaching, organizational design, and go-to-market. Each resource includes a brief summary and links to the original article or podcast.
AI Strategy: The Future of Product Teams
These articles address the strategic questions many companies are now facing: how much is actually changing, where the changes matter most, and what they should do about it. They cover team topology, business model risk, and why some companies are adapting faster than others.
Article: Preparing For The Future
Marty Cagan’s perspective on how generative AI is likely to reshape product teams and product work, drawing on four decades of experience through previous technology disruptions.
Key takeaways:
- AI dramatically reduces the cost and cognitive effort required to build software, enabling smaller product teams to own broader business problems.
- As AI automates more delivery work, the value of product managers, designers, and engineers increasingly comes from judgment, creativity, and discovery rather than artifact creation.
- The companies that thrive will be those that rethink how they build products, not simply those that adopt new AI tools.
Article: A Vision For Product Teams
A 3–10 year vision for how generative AI is likely to reshape product team size, roles, and organizational topology as the cost of software delivery continues to fall.
Key takeaways:
- As AI reduces the cost of software delivery, product teams become smaller while owning broader business outcomes.
- The core product trio of product manager, designer, and engineer remains necessary because discovery still requires distinct, deep forms of judgment.
- Lower barriers to building software increase competitive pressure while creating new opportunities for startups and product innovation.
Article: Team Autonomy and AI
Clarifies the difference between team empowerment and team autonomy, and explains how AI can help teams work across large, complex codebases with fewer dependencies.
Key takeaways:
- Empowerment is the ability to discover the best solution. Autonomy is the ability to build, test, and deploy that solution without depending on other teams.
- Legacy systems, specialized technologies, and the limits of cognitive load have long constrained team autonomy.
- AI tools that help engineers understand and safely modify unfamiliar code can reduce dependencies while improving product quality and maintainability.
Article: Tests of the Product Model
Argues that generative AI is a real-time test of which companies have the Product Model skills required to evaluate disruptive technology and use it to solve meaningful customer and business problems.
Key takeaways:
- The Product Model enables companies to respond quickly to major opportunities and threats rather than settling for superficial features, marketing claims, or outsourced solutions.
- Generative AI shows that established companies can respond successfully to disruption when they retain strong product and engineering capabilities.
- Companies that have not begun developing Product Model capabilities face growing risk, regardless of their size, industry, or current position.
Article: Disruption and Denial
Draws a direct parallel between the objections Marty Cagan heard about the internet in the 1990s and the objections he hears today about AI.
Key takeaways:
- The most common objections to AI mirror the resistance to the internet: that nothing fundamental will change, and that a company’s particular constraints make adoption impractical.
- AI is likely to change product team topology, product roles, and how teams discover and deliver solutions, not simply add another feature to existing products.
- Companies that use uncertainty and risk as reasons to stand still may be displaced by competitors that learn how to create substantially better intelligent products.
Article: Creating Intelligent Products
Argues that the most valuable intelligent products will combine deterministic and probabilistic approaches, and that product teams must learn to embrace probability as a source of value rather than an edge case to avoid.
Key takeaways:
- Probabilistic behavior is central to intelligence, not an alternative to a serious or reliable product.
- Intelligent products will commonly combine deterministic systems with probabilistic models.
- Regulated, mission-critical, and B2B products may be among the areas where probabilistic approaches create the most value, provided the risks are managed appropriately.
Article: Build vs Buy in the Age of AI
Explains why AI-powered user-programming tools will not eliminate enterprise software vendors, and how software built for both people and AI agents is changing the traditional build-versus-buy decision.
Key takeaways:
- Complex business rules and business logic, not technical difficulty, are why enterprise SaaS won’t be fully replaced by user-programmed tools.
- Standards such as MCP allow business services to be understood and used by software and AI agents, not only through interfaces designed for people.
- The likely future is both build and buy: companies will purchase core business services and create custom agents, applications, and workflows on top of them.
AI Product Management
These articles define what “AI product management” actually means, and describe the new challenges for the role as AI takes on more of the delivery work.
Article: AI Product Management
Explains why AI-powered products are especially challenging to build, and how probabilistic technology changes the work required to address value, usability, feasibility, and viability risks.
Key takeaways:
- AI product management is the work of creating AI-powered applications, and over time most product managers will be expected to develop these skills.
- Probabilistic products require teams to determine where AI is appropriate, what types and rates of error are acceptable, and how those errors can be mitigated.
- AI places greater responsibility on the product manager to understand the customer, the data, the technology, the business, and the legal and ethical implications of the product.
Article: AI Product Management 2 Years In
Reflects on the first two years of generative AI in product management, distinguishing the creation of AI-powered products from the use of AI to change how empowered product teams discover and build solutions.
Key takeaways:
- Despite rapid progress in the technology and a constant stream of announcements, the day-to-day work of empowered product management has not yet changed as much as many expected.
- The most important unanswered questions concern how AI will affect product discovery, product sense, creativity, innovation, and the judgment required of product managers.
- Generative AI is most promising when paired with strong product sense – and potentially dangerous when used without the necessary product foundation.
Article: Product, Design and AI
Clarifies the distinct contributions of strong product managers and product designers, and explains why product sense and design sense become more important as generative AI makes artifacts faster and easier to produce.
Key takeaways:
- On an empowered product team, the product manager is responsible for value and viability, while the product designer is responsible for creating the overall user experience.
- Product sense and design sense come from deep knowledge, practiced judgment, and distinct professional disciplines, not from the ability to generate a PRD, interface, or prototype.
- Generative AI can dramatically accelerate prototyping and product discovery, but the team must still determine whether a proposed solution is valuable, usable, feasible, and viable.
Article: The Era of the Product Creator
Explains how generative AI is enabling more designers, engineers, founders, and business leaders to take an active role in creating products, regardless of title.
Key takeaways:
- A product creator is anyone actively shaping a solution and addressing value, usability, feasibility, and viability.
- Generative AI is enabling more people with product sense and agency to contribute directly to product creation.
- Most serious products will still require a range of complementary skills, even as the boundaries between traditional product roles become more fluid.
AI-Powered Product Discovery
These articles cover the part of product work that AI is changing the fastest right now: prototyping. They explain what these new tools are for, what they aren’t for, and how to tell the difference.
Article: The Purpose of Prototypes
Explains why AI-based prototyping tools are best used to discover a successful solution rather than build the eventual production product, and clarifies what “just enough fidelity” actually means.
Key takeaways:
- The primary purpose of a prototype is to discover a solution worth building, not to produce a deliverable or an early version of the final product.
- The fidelity required depends on the specific product risk being tested: value, usability, feasibility, or viability.
- AI tools have made live-data prototypes fast and inexpensive enough to become a much more common part of product discovery.
Article: Prototypes vs Products
Explains the substantial gap between a convincing AI-generated prototype and a commercial-quality product, especially for product creators without an engineering background.
Key takeaways:
- A production product must address business complexity, operational demands, security, reliability, scale, and maintainability that a discovery prototype does not.
- Tools designed for prototyping and tools designed for production software solve different problems and are used differently by skilled practitioners.
- Building to learn and building to earn remain distinct activities, even when a prototype looks and behaves much like a finished product.
Article: Build to Learn vs Build to Earn
Uses Jeff Patton’s framing of “build to learn” and “build to earn” to explain the difference between product discovery and product delivery, and why that distinction matters even more as AI lowers the cost of building software.
Key takeaways:
- As the cost of delivery falls, discovering a solution worth building becomes the primary bottleneck and source of competitive advantage.
- Product discovery is building to learn; product delivery is building to earn. They require different tools, techniques, risks, and forms of testing.
- AI makes it possible to create and test more realistic alternatives earlier, but the product manager’s essential contribution remains product sense and responsibility for value and viability.
Article: Build To Learn FAQ
Answers the most common questions raised by Build to Learn vs Build to Earn, including how empowered teams frame discovery, what product managers contribute, how AI accelerates learning, and the proper role of the PRD.
Key takeaways:
- Product leaders identify worthy problems through product strategy; empowered teams are responsible for discovering solutions that work for both the customer and the business.
- The product manager is an individual contributor responsible for value and viability, not the team’s boss, sole decision-maker, or gatekeeper.
- AI supports delivery primarily through automation and code generation, while in discovery it supports rapid prototyping, decision-making, and the development of product sense.
AI as a Coach
These articles and podcasts cover a genuine shift in SVPG’s own position: after two decades of advocating human coaching as the primary way product people develop, we now believe foundation models can serve as a scalable, always-on product coach for most product creators.
Article: Product Coaching and AI
Explains why SVPG now recommends foundation models as personal, always-available product coaches for most product creators, while reserving human coaching for the leadership and organizational challenges where it has the greatest value.
Key takeaways:
- Foundation models given the right operating model, project instructions, and strategic context can now provide useful product coaching for most product creators.
- Always-available AI coaching can make the craft of product more accessible and accelerate the development of product sense, especially for people entering the field.
- Human coaches remain most valuable for product leaders navigating organizational politics, transformation, relationships, and strategic context.
Podcast: Coaching in the Age of AI
SVPG Partners Christian Idiodi and Marty Cagan revisit SVPG’s long-held view that great product people are developed primarily through human coaching, and explain why foundation models can now serve as scalable, always-available product coaches when given the right operating model and strategic context.
Key takeaways:
- The most important instruction for an AI product coach is which operating model it should coach toward: the Product Model or the project model.
- AI coaching can make high-quality product guidance available to people whose managers lack the time or product craft to coach them effectively.
- The purpose of an AI coach is to strengthen product sense and judgment, not replace thinking, automate discovery, or generate artifacts without understanding the problem.
Podcast: Coaching the Future of Product & Engineering with AI
SVPG Partner Christian Idiodi talks with Mike Fisher, CEO of MyFitnessPal and former CTO of Etsy, about why making product and engineering work cheaper may increase demand rather than reduce it, and what this means for team structure, product judgment, leadership, and the new costs created by AI.
Key takeaways:
- When technology makes creation cheaper and easier, customer expectations and the demand for innovation tend to rise rather than fall.
- As AI takes on more of the building, product judgment, leadership, and the ability to determine what is worth creating become more valuable.
- AI will change team size, roles, and operating costs, requiring leaders to rethink team structure, coaching, and resource allocation, not simply introduce new tools.
AI and Go-to-Market
As AI lowers the cost of building and increases the pace of product change, go-to-market becomes a more important source of differentiation.
Podcast: Product Marketing in the AI Era
SVPG Partners Martina Lauchengco and Christian Idiodi discuss why AI may be accelerating mediocre product marketing more than improving it, and why judgment, differentiation, and restraint matter more as the bottleneck shifts from building products to bringing them to market effectively.
Key takeaways:
- As AI makes it easier to build products and generate marketing assets, the competitive advantage shifts toward clear positioning, meaningful differentiation, and strong go-to-market execution.
- When anyone can produce passable messaging, judgment, taste, and the discipline to remove rather than add distinguish great product marketing from work that is merely adequate.
- The fundamentals of product marketing have not changed, but AI has raised the bar for every customer-facing asset and increased the consequences of weak product marketing.
Frequently Asked Questions
What does SVPG mean by “AI product management”?
At SVPG, we distinguish between using AI to build any type of product, including conventional (deterministic) products, versus building an AI powered (probabilistic) product.
For the former (using AI to build any product), we think it is only a matter of time before every product manager is skilled in using AI to do their jobs.
For the latter (when you are building an AI powered product), then there are some unique challenges to these types of products, including some essential techniques, and this knowledge is what we mean when we refer to an “AI Product Manager”.
Will AI eliminate the need for product managers, designers, and engineers?
Not on empowered product teams. AI can generate PRDs, interfaces, prototypes, and code, but it does not replace the judgment required to determine which problems are worth solving or which solutions are valuable, usable, feasible, and viable. Work centered primarily on coordination, administration, and artifact production is more exposed than work requiring deep product, design, and engineering judgment.
What’s the difference between “build to learn” and “build to earn”?
Build to learn is product discovery: creating and testing prototypes to determine whether a solution is worth building. Build to earn is product delivery: producing the reliable, secure, scalable product customers and the business can depend on. AI supports both, but it is used more for prototyping and decision support in discovery and more for automation and code generation in delivery.
Can AI serve as a product coach?
We believe foundation models can now provide meaningful, always-available product coaching when they are given the right operating model, project instructions, and strategic context. In many if not most organizations, the leaders don’t have the time or knowledge to provide this coaching. An AI product coach can fill this gap. If you also have access to a strong human product coach, even better.
What is an “intelligent product”?
An intelligent product combines deterministic and probabilistic approaches to solve customer and business problems. Recommendation systems, autonomous systems, and generative AI products all rely on probability rather than determinism for at least part of their behavior.
Should we build our own SaaS solutions, or buy from vendors?
In most cases, the answer will be both. Complex, well-established business logic, such as compliance, pricing, finance, and industry-specific rules, often remains a strong reason to buy. Companies can then build custom agents, applications, and workflows on top of those services as software becomes increasingly accessible to AI agents as well as people. See this article.
What is the biggest risk of using AI in product work?
The biggest risk is outsourcing judgment to the tool rather than using it to improve thinking. Generating a PRD does not replace discovery, and accepting an AI-generated answer does not remove the need to evaluate it. AI is most valuable when it strengthens product sense and decision-making rather than bypassing them.
How is AI changing product marketing?
AI has made it easy to generate acceptable messaging and marketing assets, which raises the competitive standard rather than lowering it. As production becomes easier, positioning, differentiation, judgment, taste, and restraint become more important. The fundamentals of product marketing have not changed, but the consequences of doing them poorly have increased.
Recommended External Resources
We reference these outside sources directly in our own AI articles. We recommend them here for readers who want to go deeper.
The End of Programming as We Know It — Tim O’Reilly
Referenced in A Vision for Product Teams and Product, Design and AI. The clearest outside analysis we’ve found of how AI is changing the engineer’s role, and a natural companion to our own writing on product and design roles.
AI Engineering — Chip Huyen
Referenced in A Vision for Product Teams. Source of the observation that AI doesn’t introduce a new kind of thinking, it reveals what actually requires thinking, a point that shapes much of our own view on judgment versus automation.
Model Context Protocol Announcement — Anthropic
Referenced in Build vs Buy in the Age of AI. The technical announcement behind the protocol we describe as solving a 30-year-old architectural problem for connecting AI agents to enterprise software.
Just Now Possible — Teresa Torres
This podcast explores the unique challenges of building AI products (versus simply using AI to build any product).