top of page
  • Writer's picturepevans

How platform strategies will shape the rise of AI-as-a-Service

The world’s largest companies, such as McDonald’s, General Motors, Merck, Vodafone, and Ping An Insurance, are deploying automated intelligent systems more systematically in their organizations.

These investments include robotic process automation (RPA), machine learning (ML), natural language processing (NLP) and deep learning systems that increasingly can be lumped together under the catch-all banner of Artificial Intelligence (AI).  While these technologies can be deployed separately, companies increasingly realize they can be even more effective if deployed in concert.

As in any market, there are two sides.  On the other side of the enterprise AI market, we have suppliers, both big tech and a growing number of startups.   To activate the market, the supply side is working not only to rolling out new enterprise-focused tools but also to develop delivery models that will make AI more readily consumable by large companies.

In particular, the supply-side is turning to platform strategies. The success of platform-driven approaches will depend on the investment path taken by the demand side of the market. Enterprise consumers will need to make significant transformations to become comfortable with and take full advantage of AI-as-a-service platforms.

Large cap companies will drive the market

While midcap and, perhaps, even some small nimble startups will play a role in enterprise AI deployment, it is safe to say that the world’s largest companies will shape the AI enterprise market.  Given their size, we can further posit that just 500 companies will have outsized influence.

The companies that make up the Fortune Global 500 generated $32.7 trillion in revenues and $2.15 trillion in profits in 2018.[1] On average, these companies have revenues of $60 billion and employ over 150,000 people.

Given their size and scale, these companies have the opportunity and resources to deploy AI at vast scale.  They can, therefore, capture gains from AI in ways that smaller companies cannot. Take selling, general and administrative (SG&A) expenses.[2]

These costs run at about 14% for the largest companies so tally to approximately $4.5 trillion for the Fortune Global 500.  If we assume that AI can make a sustainable 10 per cent dent in SG&G, this holds the potential to shave hundreds of billions in costs and improve profitability for these large cap companies.

At the same time, the world’s largest companies have complex organizational structures, typically operate across many regulatory jurisdictions (national, state and local), and have significant institutional inertia, which often weds them to existing systems and processes.  Achieving an organization-wide transformation that embraces the full potential of AI is a major multi-year undertaking.

More supply than demand

At present, there is a mismatch in the market for enterprise AI.  There is more supply than there is demand.[3] The supply-side of the market served by companies like IBM, Google, Amazon, Intel, Microsoft has ramped compute power, cloud services and a range of tools and capabilities.

We are seeing a growing number of companies indicate THAT they are incorporating AI into their growth, productivity, and innovation but overall investment levels remain modest compared to other areas of information technology investment.  The enterprise market for AI is about $12-15 billion today. This a small fraction of their total enterprise IT spending, estimated by the market research firm Garner to total $3.8 trillion worldwide in 2019.[4]

Causes for the Mismatch

There are several reasons for market mismatch.  AI has been hard for enterprises to deploy at scale.[5]  Companies must build significant in-house capabilities. They must adapt their tech stack, refigure their data stack, and even more challenging, create new organizational capabilities—or “org stack.”

These include answering questions such as: What is the optimal organizational design to drive AI initiatives and more fundamentally in which functional and process areas should AI be a near term priority?   Should authority to manage the full range of AI-related activities be centralized, handed over to an AI Center of Excellence or pushout out to the business and functions? Companies must also chart an AI strategy.

Should resources be devoted to growth via customer-facing front office use cases?  Or, should they be devoted to driving productivity by improving back-office function?  AI applications also have the potential to improve compliance, fraud detection, and other measures that help firms reduce regulatory costs and retain the value of their intellectual property.   A challenge of a general-purpose technology like AI is that it can serve many business objectives.

Adjustments are underway  

Developments are taking place that are changing the market balance.  Most large companies indicate that they will double spending on AI over the next three years.  Many are also undertaking changes to their tech stack, data stack and org stack that will improve their ability to effectively deploy AI more broadly throughout their enterprises.

On the other side of the market, suppliers recognize they must make AI easier to adopt and more effective in addressing specific use cases.

Let’s review these adjustments in more detail beginning with the demand-side of the market.

Demand-side adjustments

Several things will need to happen on the demand side to enable enterprise consumers to become more significant AI consumers.

Budget allocations: Building AI capabilities requires investment. Funds are required for technology, supporting infrastructure, data prep, and the acquisition or internal development of the required expertise.

Management authority: Not much will happen unless a company clearly articulates who decides, approves or disapproves AI activities.  Authority and responsibility must be assigned to drive deployment.

Talent development:  AI requires technical talent in the form of RPA, ML, NLP, and other technical capabilities. It also requires senior managerial talent.  To drive AND scale deployment at least some in the most senior positions in the company must understand the technology beyond a superficial level and know how it can support the firm’s competitive position.

Linkages: It is insufficient to establish an AI center of excellence.  The center must interact and coordinate with functions and business lines.  Deliberate measures must be taken to ensure that linkages are established and work to the benefit of the company’s AI deployment strategy.

Governance: Finally, internal policies and procedures related to AI need to be put to paper and publicized within the company.  Articulating the swim lanes helps employees know what they can and cannot do. Clear policies support transformation by clarifying appropriate and inappropriate actions.

It would be a mistake to think that the largest 500 companies are all progressing at the same pace across each of these five key areas.  There remains wide variation across the competencies required to become AI-enabled.

Some large enterprises are spending hundreds of millions of dollars and have made significant progress in addressing managerial authority, talent development, linkages with functions and business units and overall governance.  Others are just beginning.

What is clear is that the large enterprise that is not beginning to develop and execute its AI strategy to drive growth, productivity and innovation is increasingly rare.  Indeed, executive teams increasingly recognize that AI investment will impact their competitive position. There is growing concern that if they do not make AI investments they risk falling behind.

Supply-side adjustments

To secure more demand, AI vendors are expanding their service offerings and innovating their delivery models.   A platform is an open architecture, together with a governance model, designed to facilitate interactions.

As articulated by Parker and Van Alstyne, a platform has three key attributes that have the purpose of facilitating efficient matching and interactions create value[6]:

  • Open architecture: allows 3rd parties, such as specialized machine learning software developers, to participate, add value, and innovate in a standardized way;

  • Interactions: provide the means by which value is created, which is finer grained than via an “exchange”;

  • Governance: gives the power to exclude bad actors, steer community behavior, and monetize.

There are several areas where we are seeing the AI-as-a-serve draw on platform strategies:

Big Tech platforms—Recognizing that demand for AI technology and tools will require making it easier to access and use, big tech companies are rolling out AI-as-a-service platforms.  Examples of these offerings to include Microsoft Cognitive Services, IBM’s PowerAI Enterprise, Amazon AI Services, and Google’s AI Platform.  These services offer developer-friendly tools that sit on top of open-source frameworks like TensorFlow the open-source machine learning library. They aim to make model training times faster and more flexible.[7]  Some also aim to promote interoperability between different AI frameworks given large enterprises are unlikely to use only one platform. Most are working to develop low code/ no code environments to make integrations for customers as easy as possible.  This approach is facilitated by creating web environments with drag-and-drop fields that users can manipulate to build applications.[8]

Bot stores/exchanges– Find turn-key solutions immediately accessible and functional delivered through downloadable apps. Users can find AI capabilities, in the form of open APIs, and apply to automated processes. Easily connect and integrate Digital Workers and existing systems to create a foundation of AI-enabled robotic process automation. Some of the companies that have set up Bot stores or exchanges include Automation Anywhere, Motion AI, UiPath, and Blue Prism.[9]

Development Platforms– Another noteworthy development is the rise of enterprise application platforms. The platform provides development tools and the development environment to build native apps to share code, libraries, and applications with no configuration required.  The idea is to reduce the friction of sharing and collaborating to speed development and deployment. These platforms cultivate large developer communities to support building applications.   For example, Outsystems now claims a community of over 210,000 members and 14,000 forum discussions. Other development platform includes Service Now, Appian and more recent entrants like Mat|r Project.  Development platforms have proven to be attractive acquisition targets. Xamarin, a “devapp” platform with 1.3 million developers, was acquired by Microsoft.   Mendix, another platform built on SAP, was purchased by Siemens.

Education platforms

The supply side also recognizes that education is crucial to activating demand.  As a result, they are turning to platform strategies to deliver AI training, often free.   A key target is enterprise professionals who will deploy AI solutions and software developers who will develop the applications.

Course offerings include areas like Bot Developer, Implementation Manager, Solution Architect, IQ Bot Developer, AI Computer Vision, and Machine Learning Building Blocks.   While the overall market is still small, it is growing quickly.

The market research firm Technavio estimates that the market for RPA training alone will exceed $200 million in 2021 up from $158 million in 2016.[10]  RPA vendors are setting ambitious targets. Automation Anywhere anticipates certifying more than one million individuals over the next five years.[11]

The big tech companies have also launched educational platforms.  Microsoft Learn is a platform offering free and paid programming combining short tutorials, browser-based interactive coding/scripting environments, and task-based achievements. These programs are evolving toward technical certifications.

Google’s education platform “Learn with Google AI” is similar, though it is focused on the company’s open-source AI framework TensorFlow and how to run models in Google Cloud.  IBM has the IBM AI Skills Academy. In some cases, suppliers have teamed up with large edtech platforms. For example, Amazon partnered with Codecademy to train developers on how to create AI-driven voice “skills” for Alexa.

In short, the supply side is developing a variety of platform strategies to accelerate large enterprise adoption.  There will continue to be the need for large companies to build in-house applications but those areas that can be standardized will be standardized and platform business models will rapidly drive this space.

The benefits of turning to external AI service ecosystems are significant.

  • they make it possible to deploy AI faster and cheaper than if they attempted to do it on their own, in-house.

  • they optimize the ability to tap and cultivate a community of scarce AI talent.

  • buying a managed service offering can have benefits from an ongoing governance and management standpoint.

At the same time, buyers must be prepared and build capabilities to effectively tap into these platform-driven ecosystems.   Indeed, enterprises developing their AI strategies today should consider how AI-as-a-service platforms will evolve over the next 3 years and invest accordingly.  Heavy investment in internal AI capabilities could become stranded assets.


While enterprise AI market may be imbalanced today, conditions are rapidly changing.  The demand side of the enterprise AI market- in terms of appetite for investment and clarification of where to apply AI – and the supply side – in terms of breadth and quality of offerings – will come into closer balance over the next three years.

The market is evolving rapidly.  Platform strategies will play a growing role.  The supply side of the market increasingly recognizes that platform strategies that link open-source frameworks, large developer communities, bot stores/exchanges, digital learning and technical certification will support faster enterprise AI adoption.

The big tech players are already working to assemble all-in-one-bundles that span the full ecosystem. There are also stand-alone ventures, which focus on one element, such as low code developer platforms. Savvy enterprise consumers will need to track these developments and build internal capacity to aligned to it.

The external platforms will not provide all that large enterprises need to meet their strategic objectives. Large enterprises will still need the capability to build applications internally. However, as the AI-as-a-service market quickly matures, the opportunities to shift resources to external ecosystem buy will grow significantly.


bottom of page