Product Coaching and AI

Why It Matters

It offers a scalable way to up‑skill millions of product creators, directly boosting organizations' ability to deliver outcomes. AI coaching accelerates product sense development while easing reliance on scarce human mentors.

Key Takeaways

• Human product coaching scarce; AI offers scalable alternative
• Foundation models now provide reasonable product guidance
• AI coach works best with clear strategic context
• Combination of AI and human coaches yields optimal outcomes
• Adoption depends on security concerns and cultural readiness

Pulse Analysis

Product organizations have long struggled with a bottleneck: effective coaching is often limited to a few senior leaders, leaving many product managers and designers without the guidance needed to translate ideas into outcomes. Generative AI, especially the latest foundation models, is reshaping this landscape. By ingesting a company’s strategic context and product operating model, these models can deliver real‑time feedback on roadmaps, user stories, and market analysis, effectively acting as a 24/7 mentor. This shift not only democratizes access to expertise but also reduces the time it takes for new product talent to develop a nuanced product sense.

Implementing an AI‑driven product coach requires disciplined prompt and context engineering. Users must explicitly define whether they are learning the product model or the project model, supply relevant metrics, and continuously validate the model’s recommendations against human judgment. While the advice is no longer flawless, error rates have fallen dramatically, making AI guidance comparable to many mid‑level managers. The most effective approach pairs AI’s scalability with seasoned human coaches who can navigate complex political dynamics and strategic nuances, ensuring that the technology amplifies rather than replaces human insight.

The broader industry impact is profound. As firms adopt AI coaching, they can accelerate transformation initiatives, close skill gaps, and maintain competitive advantage in markets where speed to market is critical. Early adopters face typical concerns—data security, privacy, and cultural resistance—but the pressure to stay ahead is intensifying. Over the next few years, we can expect AI product coaches to become a standard component of product development toolkits, complementing human mentorship and reshaping how organizations cultivate product leadership.

Product Coaching and AI

Martina Lauchengco · September 3, 2024

This article represents a significant change to what we have been advocating for the past two decades. But hopefully you will agree it is necessary, and that it also represents a substantial step forward.

We have been gradually but consistently raising the alarm about the need for product owners and feature team product managers to upskill to play the role their product team needs in order to deliver outcomes. If your CEO, (and/or your engineers and designers) decide that your role is simply product management theater, then your job may be in jeopardy.

Ironically, for many people, their use of AI is simply exposing the theater, rather than highlighting their potential contribution.

For example, if you’re just using AI in order to accelerate the project model by showing how trivial it is to aggregate requests and feedback, generate roadmaps, and create PRD’s and/or user stories, realize that an agent – or your engineer or designer – could just as easily do this themselves.1

The Limitations of Human Product Coaching

Thus far, we have argued that the best way to learn the product model, or more generally, to learn how to be strong at product, is through product coaching. And further, we argue that it is the primary responsibility of your manager to provide that coaching.

That was how I learned product, and it’s how most of the best product people I know learned product. We have shared widely and loudly how coaching is a top leadership principle at the most consistently innovative product companies.

Moreover, if a company wants to transform, we know that until and unless the product leaders and the product creators can earn the trust of the stakeholders, little will actually change. Normally it is effective product coaching that enables product people to earn that respect.

The problem with all this is that especially at companies that have not yet transformed, so few product people have a manager that is both willing and able to provide this coaching.

Either because the managers have never worked this way before; or because they simply don’t have the time (especially if their company is following the trend of increasing the number of direct reports per manager).

So the result is that more people than ever before are in desperate need of product coaching, at a time when many companies have more real opportunities, as well as more serious threats, than ever before.

We consider the lack of effective product coaching to be the primary obstacle to more people and organizations becoming strong at product, at the very time that so many companies are in serious need of these skills.

Certainly good training can help, as can strong external product coaches, but even if your company is willing to pay for this (most don’t), none of that is really a substitute for having someone that is both an expert in product craft, and also knows your company’s strategic context, to be there to guide you for at least your first year or two.

Our industry needs a scalable, affordable and accessible product coaching solution for the literally millions of product creators, and tens of thousands of product leaders, that are in urgent need of up‑skilling help.

The Model as Product Coach

Over the past year, we have been experimenting with using generative AI to help address this problem, first with custom GPT’s, and more recently, with the foundation models themselves.

This should not be a surprise. We’ve all seen as people in so many roles evolve their work to using the models as assistants, agents, thought partners, and teachers.

What is new, is that over the past several months, the models have consistently improved, and at the same time, we are learning how to better inform the models of our particular goals, constraints and context.

As prompt engineering has evolved into context engineering, we have been learning just what context is necessary for the type of collaboration and engagement needed for effective product coaching.

The result is that we are now advocating that product creators and product leaders use the foundation models as their personal product coach.

To be very clear, if your manager is willing and able to serve as a strong product coach, by all means you should take advantage of this, and consider yourself very fortunate.

But for the rest of us, we now believe that the foundation models, when configured with appropriate project instructions and your company’s strategic context, can provide product coaching at least as good as most managers.

Are the models as good as a strong human product coach?

Not yet, but we believe that’s the wrong question. The right question is: “Is an AI product coach able to help most product creators and product leaders develop their product sense, and contribute at the level that is necessary?”

For product creators, we believe the answer to that question is now “yes.”

For product leaders, especially of larger organizations, we believe the combination of the model‑as‑product‑coach, with a strong human product leadership coach, gives you the best chance of getting to a successful outcome.

While the models are by no means perfect, we have found that the frequency and severity of what we consider “wrong” or “unhelpful” answers or advice has dropped to the point where most answers or advice range from reasonable to quite good, at least with the three major foundation models (Claude, Gemini and GPT).2

Importantly, in its role as a product coach, you will need to instruct your model to prioritize the operating model of product you are trying to learn (at a minimum, are you trying to learn the product model or the project model)?

This is because there are many different voices in the product world, and the advice and principles vary widely, so it should be no surprise that the foundation models can appear confused unless you clarify the approach you are working to learn.

Keep in mind that a foundation model is not a deterministic product. The coaching advice you get today can be different tomorrow, and not necessarily better. We’ve seen that in our testing and actual use. In fairness, human coaches can have that characteristic as well. We are also betting that the model‑as‑coach experience will continue to improve.3

Also keep in mind that you are not there to blindly accept the statements and advice from the model; you need to question and seek real understanding, look for critique and mistakes rather than affirmation.

If you want a good place to start using the model‑as‑product‑coach, work on developing your product sense. We are encouraging everyone to at least try this out.4

7×24 Personal Product Coach

But let’s pause for a minute to consider the implications of this, because we think this is very profound:

Any aspiring product creator with the necessary agency, whether that person is in San Francisco, or Sao Paulo, or Lagos, or anywhere else in the world with an internet connection and a connected device, now has 7×24 access to the advice and assistance of an experienced product coach, representing the aggregated learnings of some of the best minds in product.

To get a sense of what is possible in terms of rapidly developing your product knowledge, after configuring your product coach, now you can start using the model‑as‑coach to learn about your company, your industry, your competitive landscape, your domain, the sales and marketing considerations, the financial considerations (both costs and monetization), the compliance, legal and privacy constraints, the key metrics used to assess your company’s health, your different types of users and customers, your enabling technology, how your product team contributes to your overall product strategy, and how your team relates to other product teams.

This knowledge is table stakes if you hope to develop strong product sense and become a strong product creator or product leader.

Barriers to Adoption

Just as with the advent of the Internet, we can all expect that adoption of this approach of model‑as‑product‑coach, and AI technology in general, will follow the dynamics of the technology adoption curve.

We are already seeing this playing out.

Some companies are moving forward with generative AI as fast as they can – often with their leaders aggressively pushing their people to embrace – and others are more conservative and hesitant to provide open access to these tools to their employees until they are more comfortable with the implications, especially regarding security and privacy.

As a reminder, in the early years of the Internet, many companies refused to store their data in the cloud due to the same concerns. The same adoption curve played out with personal computers, and then with mobile devices.

That said, the impact of this technology is so dramatic compared to anything we’ve experienced, and the competitive disadvantage to abstaining is so painful, I am already seeing even conservative companies move forward more quickly than in the past.

Human Product Coaches

What about human product coaches?

We have been working for the past 5 years to build a global network of product coaches that we believe are strong and knowledgeable about the product model, and this network is larger and better than ever. However, realistically, they represent a drop in the bucket compared to what the industry needs.

We have been encouraging product coaches to focus on a company’s product leaders, especially those that are new to the product model, to help them navigate the company politics involved in transformation, and to create the strategic context (the product vision, the product strategy, the team topology, and the team objectives) that the product creators depend on to discover and deliver business results.

For those that have never had the opportunity to see strong product leadership, this is very hard to do well. The problems at this level are mostly people problems, involving relationships and power dynamics. There is a lot of nuance and judgement necessary as well as very strong knowledge of product craft. This is where a human product leadership coach can make all the difference for a company.

To be clear, I am still the number‑one fan of human product coaching, but I think those of us that serve as human product coaches need to focus on where we can have the greatest impact, which is with the product leaders, and embrace the model‑as‑product‑coach for the millions of product creators out there struggling to develop strong product sense and learn the craft of product.

The Zero‑to‑One Problem for New Product Creators

There is one other important point related to this.

One of our most popular articles discussing the impact of AI is A Vision For Product Teams.

It is still too early to say how well that article will predict the future, but at the end of the article, I shared a worry, which was that while it seemed very clear that experienced product creators would be very much in demand (and so far, they are), I was worried that people new to product would find that entry was blocked because the bar was too high for anyone that didn’t already possess the necessary experience.

Today I’m happy to say that I think I was wrong about that. I didn’t envision that the models would be able to get good enough, fast enough, that they could help to dramatically accelerate the learning curve for aspiring product creators and product leaders. I’m seeing this with all types of product creators including product managers, product designers, and especially engineers.

With the model as coach, and with coaching provided essentially continuously, you can progress dramatically faster than when you depended on your weekly 1:1 for your coaching.

We will share more in upcoming articles about specific techniques and best practices for using a model as your personal product coach, and we do expect this to continue to evolve and develop. But for now, we encourage everyone to start experimenting with the model‑as‑coach, and see if this can’t help you rapidly raise your level of expertise in product craft.

Special thanks to SVPG Partners Chris Jones and Christian Idiodi, and Product Coaches Thomas Fredell, Marcus Castenfors, and Elias Lieberich, for their thoughts on early versions of this article.

---

Footnotes

1. More about this serious anti‑pattern in an upcoming article.

2. For those wondering, yes, we considered training a custom model and making that available as a product. But experienced product people will know that doing a good job with this would be a significant undertaking, and with the foundation models rapidly improving, the lifespan of that effort would likely be limited. So we’ve been focused on the foundation models reaching the level we believe is necessary. Also, we view the model‑as‑coach more like our content – we want the knowledge freely available to everyone.

3. The recent announcement of Skills from Anthropic will, we are hoping, represent yet another step forward.

4. You can find an example set of project files and project instructions at www.svpg.com/examples.