Video Summary
Only use AI to solve a real pain point — avoid the 'shiny object' trap of doing AI for its own sake.
AI fluency will be essential for PMs; basic coding and familiarity with tools like ChatGPT and AutoML help you prototype and lead.
Data quality and diversity often matter more than sheer volume; collect your own signals where possible.
AI product work is research-like: expect uncertainty, iterate with MVPs, and partner closely with research scientists and data teams.
To get buy-in, show comparable case studies, present clear rollbacks, and build trust incrementally.
Marily warns against the 'shiny object' trap: AI should only be used when it addresses a clear user pain or creates measurable value. Start by defining the problem and a high-level solution before investing in AI.
Start small: use existing tools (e.g., ChatGPT, AutoML) to prototype, hire a data-science intern or partner with researchers, and learn basic coding through courses or bootcamps to gain the mindset needed to lead AI projects.
It depends on the task. Simple problems can need little data, while complex models need large, diverse datasets. Prioritize collecting your own high-quality, diversified data rather than relying solely on generic datasets.
Use analogous case studies, present a clear plan with rollback strategies, and build trust through incremental experiments. Framing proposals with past successes helps stakeholders accept calculated risk.
"Don't do AI for the sake of doing AI; make sure there is a problem that needs to be solved in a smart way."
"We talk about a bunch of resources that will help you get started in the world of AI."
"I literally go to ChatGPT and I say, 'rewrite this mission statement for me.'"
"In the future, every product manager will be an AI product manager."
"You need to get comfortable with having a partner that's a research scientist who can produce a smart model."
In product management, it's essential to collaborate with research scientists who can leverage their expertise to create intelligent AI models. Many product managers (PMs) struggle with the uncertainties that research entails, preferring clear timelines and outcomes. However, embracing the unpredictability of research can lead to more innovative products.
A successful intersection in product development combines user desirability, business viability, and technical feasibility. Understanding this can make a significant impact on product launches.
"You shouldn’t be overwhelmed by these technologies if you don’t have a technical background, because you can learn these things."
PMs without a technical background should not feel daunted by AI. They can start learning by considering how to incorporate smarter features into existing products, such as personalization, fraud detection, or enhancing security.
Gathering and analyzing data is crucial; PMs should focus on leveraging any available data to improve user experiences with AI. Hiring a data science intern can be a practical first step to integrate AI into products.
"Don’t do AI just for the sake of doing AI; make sure there's a problem that needs to be solved."
It's vital for PMs to identify real problems that AI can address, rather than just implementing AI for its own sake. A clear problem definition helps guide product development and ensures that resources are effectively utilized.
The first step in considering AI should be to assess the actual need for it within the product, ensuring there is an audience and a problem worth solving.
"If it doesn't solve a real problem, then it’s probably not worth the investment."
PMs should be cautious about investing time and resources into AI projects that do not have clear ROI. Presenting a compelling vision with an MVP is often more effective than trying to train complex models from the start.
For early-stage projects, using existing datasets or simpler methods can yield better results than attempting to build something intricate when the market still requires validation.
"The amount of data you need for AI really depends on what you are trying to do."
The necessary volume of data for AI and machine learning varies greatly depending on the complexity of the task. Simple tasks might require minimal data, while more sophisticated applications demand large datasets.
Early-stage startups frequently lack sufficient data to train effective models. PMs should consider when it’s appropriate to build their models versus utilizing existing AI technologies available in the market.
"You do want to diversify; you do want to collect your own data."
PMs must focus on diversifying the data they use for training AI models rather than solely relying on generic datasets. Original data can improve product quality and create a competitive edge.
Determining when to launch a product with AI is a critical decision for PMs, who must balance accuracy and user expectations against development timelines.
"A model is like a kid's brain; it has the ability to take an input, recognize it, and output the probability of certainty."
The process of training a model involves providing extensive datasets, such as thousands of images of various animal species, to help the model learn to identify and categorize them.
Once the model has processed this data, it no longer learns through simple if-then logic. Instead, it identifies complex patterns that might not even be apparent to humans.
The output of the training process results in a model that can predict or classify images or text, including probability scores of certainty regarding recognitions made.
"This technology is real; it's just a matter of connecting the pieces to see them come to life."
The speaker highlights an impressive function of Google Glass developed within the AR/VR team, which translates spoken language in real-time, enabling communication between people who speak different languages.
Such advancements demonstrate that the future of technology can effectively bridge communication barriers, making the extraordinary capabilities of AI a practical reality.
"It will free up time for me to focus on the more strategic side of things."
Artificial intelligence tools like ChatGPT could enhance productivity within product management, rather than replace the role entirely.
These tools can automate routine tasks such as drafting documents and allow product managers to concentrate on strategic decision-making and innovative thinking.
"I encourage people to start learning how to code; it gives you a different mindset and confidence."
The speaker emphasizes the importance of coding skills for product managers amid the rise of AI technologies, suggesting that a basic understanding of coding will empower them to engage more deeply with AI applications.
Overcoming the intimidation associated with learning to code can result in enhanced problem-solving and creativity in product management.
"There are amazing resources like Coursera and online coding schools where you can learn at your own pace."
For those interested in learning to code, a variety of resources are available, including online courses and coding boot camps.
Suggestions include platforms like Coursera, Career Foundry, and General Assembly, which cater to different learning styles and enable aspiring product managers to acquire relevant skills effectively.
"I have a course coming up for current and aspiring product managers that want to build AI products."
The speaker is launching a new course aimed at product managers who are eager to integrate AI into their products, suggesting that they should understand what it takes to manage AI products successfully.
Engaging with resources that provide both theoretical knowledge and practical experience is essential for those who wish to thrive in the evolving landscape of AI in product management.
"AI product development is different; sometimes you're managing the problem and the product at the same time."
AI product development involves unique complexities compared to standard product management. It requires an understanding of whether a problem can be effectively addressed with a smart solution.
It's crucial for product managers to grasp how AI product management diverges from traditional methods, and to engage with AI researchers to gain context and insights into current possibilities.
"You may have incredible research ideas, but the results when you train the model may not be optimal."
Uncertainty is a significant challenge in AI projects; results may not align with initial hypotheses or predictions.
As the leader of the team, a product manager must encourage and guide the team throughout this process, helping maintain momentum despite setbacks.
Obtaining high-quality data for AI models is challenging, often requiring creative approaches to data collection, which may include direct outreach to gather user responses.
Career advancement paths may differ in research-heavy roles, with fewer launches, making early clarification with hiring managers about progress metrics essential.
"I always use examples and say, 'Here's how my proposal is similar to this previous successful project.'"
To secure investment for AI initiatives, it helps to leverage successful case studies from adjacent products, drawing parallels to justify new proposals.
Proposing a clear plan that includes rollback options can make stakeholders more amenable to taking risks. Establishing a culture that embraces failure is beneficial for innovation.
The more trust a product manager builds within the organization, the easier it becomes to advocate for potentially risky projects.
"We shouldn’t underestimate academia and research blogs; they provide valuable insights."
Staying updated on advancements in AI requires diligence, with resources such as research papers and blogs playing a crucial role in informing product managers about new technologies and methodologies.
The bridge between academic research and practical application is vital, necessitating collaboration between PMs and researchers to monetize innovative ideas and products effectively.
"My course is three weeks long, designed for current and aspiring PMs interested in AI solutions."
The course is structured to educate participants on the product development life cycle, particularly how it compares to traditional product management.
Key topics include idea generation for AI applications, understanding users' needs before they know them, and collaborating with research scientists to convert ideas into viable products.
The course consists of nine workshops, with a focus on practical skills that enhance performance in product management roles.
"It's just crazy to think that you can do that within three weeks."
Students in the course engage in hands-on exercises that require them to create and develop their AI products from start to finish. This experiential learning leads to impressive outcomes, including one team that successfully raised funding for their app.
The culmination of the course features presentations where students showcase their work, receive feedback, and express excitement about their creations. The learning experience emphasizes the importance of practical application over mere theoretical knowledge.
An example from the course is a student who developed a model capable of analyzing online medical images, such as X-rays, to identify potential issues. This model exemplifies the rapid pace of innovation and feasibility within a short three-week timeframe.
"One of the tools I would like to recommend is AutoML."
For those interested in building AI projects without coding skills, tools like AutoML can significantly lower the barriers to entry. AutoML allows users to train custom machine learning models using photographs, making the process accessible to non-programmers.
A case study provided involves a company that used drones to photograph wind turbines, replacing labor-intensive inspections. They processed these images with AutoML to determine which turbines needed maintenance, reducing their inspection time from several weeks to just a few hours.
"Treat creating my course like a product."
The course creator approached the development of the course systematically, identifying the target audience and their learning needs. This included engaging with potential students to gather insights on what they wanted to learn and answering specific questions.
Over time, the course evolved from targeting software engineers to focusing on product managers interested in AI. Key elements include finding the right audience, ensuring the appropriate duration for learning, and maintaining a personal connection with students.
Iterative feedback and a supportive environment proved essential in refining the course, resulting in a program that resonates well with participants.
"Teaching and crystallizing thoughts is one of the best ways to learn."
Creating the course not only involved consolidating knowledge but also forced the instructor to tackle challenging questions posed by students, enhancing their understanding of AI topics.
The iterative process of answering difficult inquiries and incorporating feedback made teaching an invaluable learning experience. This cycle of teaching and learning can lead to deep insights in the field of AI, showcasing the dual benefits of course creation for both educators and students.
"It's so important to be able to explain things in a simple way and have the storytelling to convince a kid."
"Try the male version—it's so much cooler than the king version."
"You can find me on Instagram; I also have a product channel on YouTube and I'm starting a newsletter."
