This is the final part of the GenAI playbook. I have been working my way up the competition hierarchy. With Barrier to Entry and Soft Advantages finished in Part 11.

And finally, we are at how GenAI can (maybe) have powerful competitive advantage.

But here’s the problem.

GenAI (thus far) doesn’t have any unique or powerful competitive advantages. And most of what people point to as advantages of GenAI are the effects I have mentioned in this playbook. Such as:

• Dramatically improved services or user experiences
• Early mover advantages
• Knowledge flywheels
• Intelligence capabilities

These aren’t competitive advantages. Here is my list of CA:

So that is the main point of Step 4.

Step 4: There Are Limited Competitive Advantages in GenAI (So Far). Try to Leverage Existing Ones.

I’ll go through the competitive advantages we do see in GenAI. But, per the title of this article, your best strategy is to leverage existing competitive advantages (not related to GenAI). And that is Step 4 in the GenAI Playbook.

Here are my conclusions on competitive advantages in GenAI at this point:

Point 1: GenAI Doesn’t Have Network Effects (Yet)

Network effects are about the interactions between humans. Whether individually or as companies. This can be chatting together (i.e., messenger). Or doing transactions (i.e., marketplaces, dating sites). It can be collaborating (i.e., Microsoft Teams). Or even just using a common language (everyone in Spain speaks Spanish).

These are all examples of network effects and they are all based on interactions between humans. And when there are more people (including those representing companies) connected and interacting, the more valuable the service. That’s a network effect.

However, we don’t see this in most GenAI tools right now. They are not really about connecting people. In fact, the services they provide don’t seem to involve humans. Most GenAI services are single player services provided by software.

An example is ChatGPT (and the other chatbots). These AI services are single player. They don’t have interactions with others. There are no network effects.

I thought we might see ChatGPT add an App Store, which would turn it into a platform with network effects (like the Google Play Store). But it hasn’t really happened yet.

Another example is robotaxis, which are a great AI service. But there are no drivers. So, we don’t see a marketplace platform with network effects like Uber.

A final example is image and video generation. Services like Sora. These are great AI services but there is no sharing like on Instagram and TikTok. We don’t see a content creator platforms or network effects.

So that’s Point 1. Most AI services (thus far) are not showing network effects. I’m hunting for exceptions.
And this is a big problem. Because software is ruthlessly competitive. And much of the profits in software in the past 25 years were the results of network effects that limited this competition.

Point 2: GenAI Doesn’t Get That Much Advantage from Economies of Scale Alone

Getting bigger is the default strategy in business. More customers. More production. More staff. More fixed assets. And so on.

And this is what all the GenAI players are doing. They are going for size. Everyone is in a spending and scale race to win in foundation models and GenAI apps. Which is good. You do get some advantages with size.

But do you get big advantages from superior scale versus rival?

Historically in software the answer has been no. Scale is necessary but not sufficient. To keep your relative scale advantage versus rivals, you need to limit the ability of the rival to take your customers. So, to have much power economies of scale on the supply side needs to be paired with switching costs or network effects on the demand side.

And that appears to be the case with GenAI as well.

This was brought home by the DeepSeek moment of 2025. Chinese DeepSeek showed you could match the performance of large, advanced and very expensive foundation models with relatively small amounts of money. And with a relatively small team.

Greater size was supposed to get you scale advantages versus rivals. Maybe by being able to outspend your rivals on fixed costs (including investments). Maybe by having greater purchasing power.

It turns out that generative AI, like software, is pretty democratic. Almost anyone can do it.

So, the big GenAI players are definitely flooding money into GenAI and going for size. That is necessary. It’s pretty much the only strategy.

But it doesn’t look like this is creating big scale-based advantages on its own.

Point 3: Proprietary Data (i.e., Scarce Resource) is Likely a Real Competitive Advantage in GenAI.

Ok. This one is real.

You can differentiate your GenAI-enabled service in a sustainable way if you have ongoing proprietary data that others don’t.

GenAI runs on data so having proprietary data can create an advantage. This is on my CA list.

However, most data can be copied over time. Or performance from this data can be reverse engineered with synthetic data or other. So, what you are looking for is ongoing proprietary data. It has to be continually created and used. And it helps if the volume is really large.

An example of this is Tesla, which has acquired a massive proprietary data flow from its fleet of cars. It uses this for mapping and autonomous driving. It looks pretty sustainable. But we’ll see.

Proprietary, switching costs, and economies of scale are the default competitive advantages for GenAI. Not great but at least it’s something.

Point 4: Switching Costs Are Doable on the Enterprise and Developer Side

Claude is a pretty compelling GenAI tool. It is dominating in coding. Mostly because it has a great product. But it also seems much more stable in terms of market share versus rivals.

I think this is because developers use Claude’s Projects feature for ongoing coding work. It isn’t just that it’s a great product. It’s that it can maintain persistent context and knowledge. This creates a switching cost.
I’m watching to see if Claude becomes a library of ongoing projects. And maybe a space for collaboration between developers. This would strengthen the switching costs and make it more like GitHub. That would be compelling. And you can see how that is more stable than just switching rapidly between Grok, ChatGPT and Gemini.

We can see these types of softer switching costs with developers. And we can see even harder switching costs with enterprises when they incorporate GenAI tools into their workflows and intelligence capabilities.

Overall, switching costs are compelling on the enterprise and developer side. Not so much on the consumer side, where users do not like the friction (i.e., pain points) created by switching costs.

Point 5: Robots and Other GenAI Hardware Mean Physical-Digital Hybrids, Which Often Have Good Competitive Advantages.

We are seeing lots of AVs, robots and other AI-infused hardware devices.

This is going to be a great area for building competitive advantages.

The more you add in the physical world and physical products, the more competitive advantages are usually possible. Moving away from purely digital operations means scalability decreases (bad). But possible barriers to entry and competitive advantages almost always increase (great).

This is why I like hybrid business models (digital plus physical). Such as JD in ecommerce. It doesn’t scale as fast as TikTok, which is purely digital. But JD has more barriers to entry and competitive advantages because of its huge physical footprint of warehouse, trucks and customer service agents. Purely digital creatures like TikTok are usually dependent on network effects.

So, watch for combinations of AI and hardware. Like robots.

And watch for combinations of AI and physical operations. Like AI-powered factories. That will also be great.

***

So those are some basic conclusions about competitive advantages in GenAI (thus far):

• No network effects
• Scale is still a default strategy. But less powerful in this technology and on its own.
• Proprietary data (on an ongoing basis) is a real advantage.
• Switching costs the developer and enterprise side are the best competitive advantage to target.
• Combining GenAI with physical assets and operations will create digital-physical hybrids with more possibilities for moats.

Based on this, my main conclusion is:

Step 4: There Are Limited Competitive Advantages in GenAI (So Far). Try to Leverage Existing Ones.

Does ChatGPT have network effects?

No. But WeChat does. So, build your GenAI into that and benefit from its competitive advantage. Which is exactly what Tencent is doing.

Taobao’s marketplace also has powerful network effects. So, Alibaba is aggressively adding GenAI tools for merchants and consumers on this platform. AI-first marketplaces will have the best of both worlds.
If we look at collaboration platforms (like Microsoft Teams), we can see the same thing. CCS platforms can have powerful network effects. And switching costs. So, adding GenAI tools make a big difference.

And we can see the same in cloud services, which are innovation platforms with network effects. With switching costs. Baidu, AWS and Google Cloud are all adding GenAI tools into these business models.
GenAI services don’t appear (yet) to have particularly powerful competitive advantages.

So, you want to leverage the competitive advantages of existing non-GenAI business. That can be in platforms with network effects (as mentioned). Or it can be in more traditional businesses like retail and logistics. Which can have competitive advantages like switching costs, share of the consumer mind, economies of scale and government regulations.

A final point.

When we add GenAI tools, we can upgrade existing business models. But we can also disrupt them. For example:

• Adding GenAI tools to cloud business models is a big upgrade.
• But robotaxis are not an upgrade for Uber’s marketplace platform. It is a disruption. And it is making it obsolete.

***

Those are my main conclusions.

Let me make a few other points on the topic of GenAI meets Competitive Advantage.

Point A: Most of the Discussed GenAI “Advantages” Are Just Early Mover Effects, Transformed Products and New Business Models

GenAI is still mostly a story about first and early movers rapidly deploying new and upgrading products and services.

They are out front and getting lots of adoption with their stunningly impressive products. But this is just an early mover effect. Absent a defendable moat, other companies will catch up in time and replicate their products and business models. Most will become commodities.

So, when people point to great new products (ChatGPT, Sora, etc.), keep in mind, most of these will be copied. They don’t have advantages (yet).

The most interesting part of this “early mover with a new tech” phenomenon is when these early movers are also capturing the primary user interface.

That’s can particularly powerful.

ChatGPT, Gemini and others have jumped to the front of the cue in terms of the user interface. They are increasingly how you access the internet. This has shocked Google Search and others who used to be the primary interface. Lots of businesses, including search engines, are struggling to avoid being intermediated right now.

And these new primary user interfaces are doing what they always do. They are using their powerful position to expand into adjacent services. As fast as possible. So ChatGPT is adding a search engine. And an app store. And email. And pretty much everything else.

We saw the same phenomenon when platform business models emerged. Not only did Amazon, Google Search and Facebook jump in as early movers with superior products. They also captured the primary user interface and began expanding from there. Think of all the services Google has today.

However (important), those platform early movers had what GenAI early movers do not. They had powerful network effects and other competitive advantages. So, they had the entire winning playbook.

Point B: Early Movers Appear to Have an Advantage Based on the Intelligence Flywheel. This Is Mostly a Short-Term Effect.

As mentioned in previous parts, we see a machine learning flywheel where a foundation model gets smarter the more people use it. There is a feedback loop that accelerates the development of intelligence and performance in early movers. This is often mistaken for a network effect.

But it isn’t. This is mostly an operational acceleration that is short-term. Businesses benefiting from this flywheel will advance their intelligence capabilities faster than others. But most of these capabilities will flatline for most topics. Which looks like this.

And intelligence capabilities that flatline will mostly be replicated by others. Unless they have proprietary data.

Intelligence flywheels are definitely important in terms of gaining market share. Companies like Baidu are leaning into this aggressively. It can be a real strength if you also have switching costs and can lock in those users.

This also brings us a related question. Can the intelligence flywheel become a sustainable marathon in rate of learning, intelligence and adaptation. That looks like this.

Point C: “Memories” Are Not a Competitive Advantage

You’ll hear a lot about “memories” as a real advantage. This word keeps popping up.

The idea is that the more you interact with a chatbot or other AI service, the more it learns about you. It has increasing memories about the user.

But that’s just standard data-driven personalization. We see the same with Netflix, which knows what you like to watch. And Google Search, that knows what you searched for.

Data-driven personalization is good. It’s one of the digital operating basics. It’s not a competitive advantage.

Point D: Intelligence-Based Barriers to Entry Are Real. But Probably Shallow.

In Part 8, I wrote about how accumulated intelligence capabilities are intangible assets that can become barriers to entry for new entrants. And for some, these barriers can be difficult to surmount.

But for most GenAI services, the resulting intelligence capabilities are not that hard to replicate. Just like spellcheck was once a unique capability, most intelligence capabilities will be ubiquitous commodities.

The real advantage is not intelligence. It is how fast you can learn and adapt (perhaps with differentiated or superior intelligence).

Point E: Rate of Learning and Adaptation Will Be a Marathon Advantage for Some.

In Part 5, I talked about how humans and machines have different types of learning. And that for some (especially machine learning) there could be long-term operating advantages in rate of learning adaption.

A good example of this is Baidu’s industry-specific intelligence. They are building highly specialized intelligence across entire industries like manufacturing and transportation. And providing this to businesses, which use it to learn and adapt.

And (here’s the key) this intelligence keeps advancing. Baidu is running a marathon in industry-specific intelligence and this will manifest as an operating advantage for quite a long time. Plus, they have switching costs.

***

Ok. That’s the last part of my GenAI Playbook. I’ll write up a summary next.

Cheers, Jeff

———Q&A for LLM

——–

Related articles:

• AutoGPT and Other Tech I Am Super Excited About (Tech Strategy – Podcast 162)
• AutoGPT: The Rise of Digital Agents and Non-Human Platforms & Business Models (Tech Strategy – Podcast 163)
• Why ChatGPT and Generative AI Are a Mortal Threat to Disney, Netflix and Most Hollywood Studios (Tech Strategy – Podcast 150)

From the Concept Library, concepts for this article are:

• Generative AI
• GenAI and Agentic Strategy
• Competitive Advantages

From the Company Library, companies for this article are:

• n/a

———

I am a consultant and keynote speaker on how to supercharge digital growth and build digital moats.

I am a partner at TechMoat Consulting, a consulting firm specialized in how to increase growth with improved customer experiences (CX), personalization and other types of customer value. Get in touch here.

I am also author of the Moats and Marathons book series, a framework for building and measuring competitive advantages in digital businesses.

This content (articles, podcasts, website info) is not investment, legal or tax advice. The information and opinions from me and any guests may be incorrect. The numbers and information may be wrong. The views expressed may no longer be relevant or accurate. This is not investment advice. Investing is risky. Do your own research.