Why AI Product Companies Are Moving Into Services And What Should Existing Service Companies Do

​AI companies spent the past few years building increasingly powerful models. Now they are facing a harder question: how do those models create value inside a real enterprise?

In the 17th century, England faced a massive problem that sounds familiar even today—its ambitions extended far beyond its reach.

They wanted to extend their reach to the Caribbean, thousands of miles across the Atlantic, and found a way through Privateers—privately owned ships authorized by the Crown to attack enemy ships and pursue its interests on its behalf.

What made them an effective strategy was not simply that they could sail that far. Their strategic value came from carrying England’s capabilities into places where England itself could not be.

AI is arriving at a similar strategic challenge.

The biggest AI models already have enormous capabilities. But it is very evident that capability alone does not create value. The real advantage comes from knowing how to apply that capability to a specific mission, adapt it to the environment and extract the maximum possible value from it.

This is where the Forward Deployed Engineer becomes important. Their role has very little to do with simply deploying a model, but rather with taking a powerful, general-purpose capability of models and making it work harder for a specific business case by connecting it with the right data, systems, workflows and decisions.

The industry is beginning to put serious efforts and resources behind this model. OpenAI has launched a new deployment company focused on helping organizations put its models into production. Similarly, Microsoft has launched Microsoft Frontier Company with a $2.5 billion commitment and 6,000 industry and engineering experts, focused on successful enterprise AI deployments. Beyond OpenAI and Microsoft, AWS has also committed $1 billion to an AI deployment initiative built around the Forward Deployed Engineering model.

In a way, it is the same strategic lesson Britain learned centuries ago: capability creates outcome only when you can take it where the mission is and put it to work.

Why AI’s Economics Are Moving Downstream

There is an economic idea from that same era that offers another useful lens: specialization.

Published on March 9, 1776, An Inquiry into the Nature and Causes of the Wealth of Nations by Scottish economist Adam Smith illustrated the idea of specialization with his famous pin-factory example. It showed how dividing work into specialized tasks could dramatically increase output.

The software industry, more specifically SaaS, carries that idea of specialization. SaaS products are typically built around a defined problem for a defined group of users. A CRM manages customer relationships. An HR platform manages people operations. The product is specialized. Their business model largely revolves around building a product once and distributing it repeatedly at scale.

AI is different.

The most powerful AI models are increasingly general purpose. The same model can write code for a developer, analyze documents for a bank, support a customer service team or help a manufacturer make sense of operational data.

But I firmly believe in the opportunity this generalization of AI models brings for the service industry. The more general the underlying intelligence becomes, the more specialized its application may need to be. And this is exactly where the AI economy starts moving downstream—from building and distributing intelligence to specializing how that intelligence is applied.

This creates an interesting economic intersection:

• Distribution plus Repetition equals Product economics
• Specialization plus Application equals Services economics

Companies like OpenAI and Microsoft are trying to combine both. Build and distribute a general-purpose capability at scale, then specialize its application for each business, workflow and outcome.

The Service Industry Already Has An Uncomfortable Advantage

One of the greatest advantages service companies have is difficult to build quickly: institutional knowledge of how enterprises actually work.

This institutional knowledge holds answers to some of the most business-critical questions, such as: Will AI work with our legacy systems? What can we safely automate? Where does human judgment stay? How does it fit existing workflows?

​This institutional knowledge matters even more as AI moves downstream, toward the systems, workflows and decisions where value is actually created.

What if this institutional knowledge could become a strategic advantage in the AI era?

What if the patterns service companies have learned across industries, workflows, architectures and transformations could become part of how they engineer, deploy and govern AI?

That is a much more interesting opportunity than simply adding AI to a services portfolio.

It means using what the service industry already knows about enterprises to make AI more contextual, more reliable, more useful and more specialized.

The New Rules For Scaling Services

I don’t think services need to become products. But I do think services need to learn from what products have taught us about creating and scaling value.

From People To Platforms

Move expertise from individual people into reusable engineering platforms, frameworks, tools and accelerators.

From Projects To Products

Seeing the same problem repeatedly? Stop solving it from scratch. Instead, turn what you have learned into reusable IP, components and products.

From Hours To Outcomes

Shift the unit of value from effort delivered to outcomes created.

From Delivery To Deployment

Move beyond delivering the solution to making it work in the real enterprise environment.

Conclusion

The Privateers offer a useful lesson. Capability only creates potential until it reaches the environment where it can create value.

AI has already entered a similar space. After building remarkable general-purpose intelligence, now the value is moving toward the harder part: making AI deliver the outcome.

That is why AI companies are now moving downstream into deployment and services. This creates a strategic opening for service companies—but only if they can turn their accumulated institutional knowledge into something more scalable than headcount.

The next service model will build on that knowledge differently: people into platforms, projects into products, hours into outcomes and delivery into deployment.

I don’t see this as competition between products and services. I see it as convergence of two kinds of value: the scale of software and the context of services.

Source: Forbes