Enterprise Software: From Clunky Forms to a GenAI Revolution

Enterprise software is BORING. For decades, we've been stuck in a world of clunky forms and rigid workflows that stifle innovation and productivity. Sure, there has been automation, but someone is still filling out what are essentially digital online forms for every quote, order, or invoice – it’s a drag. The best thing we offered users was more fields, application modules, and form-fill interfaces. But a revolution is coming, driven by Generative AI (GenAI). Get ready to ditch the form entry drudgery and say hello to a future where software anticipates your needs, understands your natural language, and adapts to your workflow.

 The Early Days (1990s): Rules and Relational Databases

The 1990s were dominated by business applications like ERP, CRM, and HRMS. These applications were built on a core principle: translating operational processes into business rules and structured data stored in relational databases (think Oracle, Sybase, DB2). Imagine applications for opportunity management, order management, and fulfillment – all governed by a set of rigid rules like you can have a maximum of 20 lines on a quote, or a zip code is mandatory before ship confirmation or an employee needs to have a manager ID before doing an employee transfer in the system. This translated to code, business logic, and a fairly simple user interface based on forms and data entry.

A significant shift occurred from mainframes to client-server architectures and relational databases, multiplying applications and modules. Waterfall and specific packaged business application implementation methodology were the dominant development approaches, but agile methodologies began to emerge in the latter part of the decade. Developers will remember the challenges of “out-of-the-box” versus customizations. Interestingly, unstructured data wasn't a major concern; it wasn't something anyone cared to store or analyze extensively. However, early content management systems (CMS) emerged, offering rudimentary ways to handle unstructured content like text documents, with heavy manual content tagging and administration support.

The Rise of the Web (2000s): Java Takes Center Stage

The 2000s ushered in the era of internet/browser-based applications (we also had applets), with Java taking the lead. While the core business logic and relational databases remained, the applications themselves moved to the browser, eliminating the need for clunky desktop installations. User experience saw advancements with better layouts and non-textual elements. However, the fundamental character of the business application remained essentially unchanged.

The explosion of data collected by these transactional systems led to the formalization of Business Intelligence (BI) and Data Warehousing disciplines, although data warehousing practices existed before the 2000s. They became more widespread with the rise of the internet and the need to manage larger and more complex datasets. Unstructured data remained trapped in "comments" or "blob" fields, primarily used for capturing data and basic text mining with regular expressions.

The Cloud Revolution (2010s): Mobility and Multi-tenancy

The 2010s witnessed a paradigm shift with the rise of cloud-based applications and multi-tenancy. Businesses could now build and deploy everything – user experience, logic, and data – in the cloud, significantly reducing the need for expensive private data centers. Mobile access became more widespread, but B2B adoption lagged behind B2C due to the text-heavy nature and complex validation requirements of B2B applications.

Unstructured data often remained confined to form fields. While Big Data was a major area of investment in the enterprise software space, the focus was on use cases such as understanding user behavior, monitoring machine performance, gaining customer insights based on engagement patterns, and achieving cost savings through moving to lower-cost Hadoop infrastructure. Data scientists began using machine learning for tasks like sentiment analysis and forecasting, but most modeling relied heavily on feature engineering based on structured data like purchases, industry, revenue, and employee count, often overlooking the valuable insights hidden within customer text inputs.

The GenAI Era (2020s): A New Frontier

The formation of the Generative AI (GenAI) field through the introduction of Large Language Models (LLMs) from 2020 work on transformer models and the broader field of AI/ML has shaken the foundation of the entire enterprise software landscape. Traditional rule-based, workflow-driven applications with good UI/UX and cloud-based relational databases are no longer a significant differentiator. The message is clear – "anyone can do that."

GenAI presents a groundbreaking opportunity, but there are challenges to overcome:

• Data for Training and Inference: What data do we feed the models to ensure effective training and Retrieval-Augmented Generation (RAG) inference?
• Probabilistic vs. Deterministic Nature: GenAI models are probabilistic, not deterministic. How do we ensure consistency in all scenarios?
• Response Times: Infusing GenAI into applications does not yield the near-instantaneous responses we are used to seeing on B2C/standard B2B applications.
• Re-architecting the Core: Simply applying GenAI as a "wrapper" or for ancillary use cases won't yield a significant impact. True transformation requires a ground-up re-architecture of the application.
• Root-Cause Analysis While Connecting to the Black Box: LLMs are built on complex neural networks with millions of parameters. They generate outputs based on learned patterns rather than predictable steps, making the input-output path opaque and challenging to trace for understanding predictions.
• Security of Data and Access: How to ensure that data being accessed is secure for AI modeling and inference without any leaks along the way.

The Future: Reimagining Enterprise Software with GenAI

This is a time of immense potential for enterprise software companies. Let's explore how we can leverage GenAI's capabilities to create entirely new software experiences that go beyond the limitations of the past:

• Beyond Forms: Imagine registration forms that disappear! GenAI could capture information in a natural, unstructured way, mimicking how we introduce ourselves in real life.
• Multimodal Input:  Move beyond form fields and accept user input through images, references, and even voice. e.g. submit an opportunity status update by sharing the notes on an image of a handwritten note.
• Structured Data from Unstructured Inputs: Convert unstructured data into structured data that feeds into core systems with established rules and workflows. For example, you simply forward your itinerary and accommodation details, and the expense is submitted.
• Self-Guiding Workflows: Workflows and rules adapt to handle exceptions within tolerances, eliminating the need for developers to code every possible variation. At the end of the quarter, the system allows for Orders to be booked with secondary input, such as partner PO #
• Real-time Integration: Instead of capturing information after the fact, capture it as it happens – for example, a discussion on invoice dispute, captures resolution and updates and resends the invoice.
• Dynamic Content Delivery: User interfaces adapt to individual preferences – color, size, location, shape, form, language

Conclusion

The transition from clunky, form-based enterprise software to intelligent, adaptive applications driven by GenAI represents a significant evolution. This transformation can enhance productivity, streamline workflows, and create more intuitive user experiences. The future of enterprise software is not just about improving existing systems but reimagining them to harness the power of GenAI fully.

Why do you say?