The Enterprise Software Model Will Be Turned On Its Head

Every major technology transition has forced enterprise software to reinvent itself. The shift to client-server architectures, the web, SaaS, mobile etc., all initially met resistance. IT organizations always had legitimate concerns, and established software companies had equally understandable reasons to protect products and business models built for the previous generation.

Eventually, however, the economics and user benefits of each transition became impossible to resist. In this way, AI is profoundly transforming enterprise software and, as importantly, is challenging the assumptions on which much of the enterprise software industry has been built.

Enterprise Software Was Built Around Complexity

For decades, the enterprise software adoption model has followed a familiar pattern, a process that can take months or years, and cost hundreds of thousands or even millions of dollars, before an organization knows whether the software will genuinely improve the business. As a result, an enormous industry has grown around making this model work - systems integrators, consultants, implementation specialists, solution architects, training organizations, customer success teams.

Some of this complexity is of course unavoidable, nevertheless, an uncomfortable question is becoming increasingly difficult to ignore: how much of this complexity is genuinely required by the customer, and how much is required by the products and business models of the enterprise software industry?

AI is about to force us to find out.

(Andreessen Horowitz - Workday's Last Workday?)

The Software Is Beginning To Do The Work

One of the most important developments in AI is the rapid growth of Vertical AI companies that are embedding expertise directly into their products.

The fundamental proposition is changing. Traditional software companies sold tools that helped people perform work, but now the emerging generation of AI-native companies increasingly sells the work itself. If expertise, configuration, analysis, onboarding and increasing amounts of work can be embedded directly into a product, the economics of enterprise software begin to change.

The customer no longer needs a major implementation project before experiencing value and the vendor no longer needs a large professional services organization to deliver it.

In this way software companies can compete for budgets historically spent on labor and services rather than simply competing for existing IT budgets.

That is a fundamental change.

(Sequoia Capital - Services: The New Software)

(a16z - Vertical SaaS: Now with AI Inside)

But Much of Enterprise AI Still Looks Suspiciously Like Enterprise Software

We may already be seeing an early sign of just how disruptive this transition could become. In July, IBM warned that some enterprise customers had unexpectedly shifted quarterly spending toward servers, storage and memory to support AI infrastructure, contributing to delays in large software and technology deals. The immediate question for investors is when that spending returns to enterprise software. But perhaps the more interesting question is what kind of enterprise software customers will want to buy when it does. By then, will they still accept the same high-cost, high-friction products and adoption models, or will the AI investment itself accelerate demand for a fundamentally different generation of software?

There is a considerable irony in the current enterprise AI market. Almost every major software company is racing to use AI to transform its own business, with developers using AI to write software, marketing organizations using AI to create content, employees adopting lightweight self-service AI products, Customer Success teams deploying automation, and executives encouraging employees to find ways of using AI to become dramatically more productive.

In other words, many enterprise software companies are breaking almost every rule on which the traditional enterprise software model was built. They are adopting products quickly, starting small, experimenting before making major commitments, empowering individual users and teams, automating work previously performed by people and demanding rapid and measurable improvements in productivity.

Yet many of these same companies continue selling AI to their enterprise customers using essentially the same model they have used for decades: contact sales, schedule a demonstration, conduct discovery, define requirements, run a pilot, integrate the data, configure the platform, hire consultants, train the users and then, eventually, experience value.

Software companies are using AI-native tools and new adoption models to transform their own economics while continuing to sell their customers expensive, complex and increasingly AI-decorated versions of the previous generation of enterprise software.

Bottom line? They are not passing the economic and operational benefits of this new model on to the customer.

(IBM - Arvind Krishna's Letter to Investors, July 14, 2026)

Nowhere Is This More Apparent Than in Industrial Software

Industrial operations may be one of the clearest examples of this contradiction. Manufacturers, and other industrial companies, have spent decades investing in countless enterprise systems and the result is that most large industrial organizations do not suffer from a lack of software, they suffer from an inability to turn the enormous quantities of information generated by that software into better and faster operational decisions.

Because of this, the Industrial AI market is now growing rapidly with sophisticated platforms for contextualizing industrial data, advanced analytics products that optimize processes, AI copilots that help users interact with existing systems and new agentic platforms that automate specific operational tasks.

The challenge is that the market continues to operate on the assumption that significant value requires significant commitment: enterprise sales, extensive integration, data preparation, implementation programs, professional services and large contracts. Some of that may be necessary, but the emergence of AI-native products should at least force us to ask how much of it remains genuinely required and how much simply reflects the products and economics of the previous generation of enterprise software.

What If Value Came First?

Imagine a different enterprise software model, a user starts with a problem rather than a complex enterprise procurement process and immediately begins using a product that addresses their problem because that product is immediately accessible and already contains the relevant industry expertise, workflows, and specialized AI agents.

The user experiences rapid and meaningful value before the organization makes a major enterprise commitment. They invite their colleagues to experience the same speed of impact and then usage expands, additional workflows emerge, more legacy systems are connected, and more data is integrated so usage expands further across functions and then across sites.

In this way the journey becomes:

Bring a Problem → Experience Value → Invite Colleagues → Expand Usage → Experience More Value → Connect Systems → Enterprise Adoption → Experience More Value

Which is a very different model from:

Contact Sales → Discovery → Demonstration → Requirements → Proposal → Pilot → Integration → Implementation → Training → Value?

In this way, value and ROI move much earlier in the customer relationship, before major enterprise commitment, extensive integration and significant implementation costs. As adoption grows, more users, workflows, data and systems can be added, allowing the value of the product and the customer relationship to expand together.

(Bessemer Venture Partners - Building Vertical AI)

Product-Led Growth Is Not Just for Consumer Software

AI means that a product can increasingly explain itself, onboarding can become conversational, industry expertise can be embedded into AI agents, and systems can understand unstructured information without requiring every piece of data to be perfectly modeled. Customer Success can also become AI-enabled, allowing human experts to focus on relationships, organizational change and complex problems, while routine support, training and problems previously solved repeatedly by consultants can increasingly be handled once and embedded into the product.

All of this creates the opportunity for something that has historically been difficult to achieve, enterprise-grade capability with consumer-grade ease of adoption, meaning that low friction does not have to mean lightweight capability, low initial cost does not have to mean limited enterprise value and self-service does not have to mean the absence of security and governance.

The assumption that sophisticated enterprise capability must inevitably be accompanied by high cost, complexity and friction belongs to the previous generation of enterprise software, and AI gives us the opportunity to challenge it.

The Next Enterprise Software Model

CEOs are increasingly looking at today's enterprise software industry with the same disbelief with which we view previous generations of computing. Does it really need eighteen months to implement a product? Do we need to spend millions before users know whether it is valuable? Does every application require consultants, extensive configuration and training? Does enterprise-grade software have to be difficult to buy, implement and use?

Security, governance, integration and enterprise architecture will still matter. But CEOs will increasingly reject those requirements as justification for slow adoption, excessive implementation and inflated software economics.

The winners will embrace this change completely rather than simply adding AI to existing products and business models. They will allow customers to start easily with a specific problem, experience value before making a major commitment, expand through adoption rather than implementation and combine low friction and lower initial cost with the depth, security and governance the enterprise requires.

The software industry already knows this model works. It is using lightweight, self-service, AI-native tools to transform itself.

The question is how long it can continue selling its customers the opposite.