Service as Software Changes Everything
Over the last decade, Software as a Service (SaaS) has revolutionized the business landscape. It introduced low-cost, flexible applications as the standard, paving the way for more agile and scalable IT frameworks. Today, organizations, both large and small, can use powerful software solutions that were once beyond their reach.
Now, with the advent of artificial intelligence, this concept is evolving further. Service as Software is rapidly emerging, promising to bring even more capabilities to the table. “Service as software combines the core principles of both the SaaS and Business Process Outsourcing (BPO) delivery models, blending them into a new, AI-powered framework,” states Fred Giron, Senior Research Director at Forrester Research.
Service as Software, also known as SaaS 2.0, goes beyond merely adding AI layers to existing applications. It fundamentally focuses on automating business processes through intelligent APIs and autonomous services. This framework seeks to minimize or eliminate human participation through the use of AI agents that respond and adapt based on events, behavioral shifts, and feedback.
The result is autonomous software. “Traditional SaaS provides cloud-based tools that require human interaction. Service as software flips that concept; instead of having staff perform tasks, you’re utilizing APIs or software that accomplishes these tasks for you,” explains Mark Strefford, founder of UK-based consulting firm TimelapseAI.
This approach is exceptionally promising for managing niche, well-defined processes such as financial reviews, legal analysis, IT reporting, marketing and public relations evaluations, and general research. Although still in its nascent stages, service as software is poised to usher in further change within the enterprise. Giron suggests that it “might surpass the SaaS revolution.”
Boosting productivity is at the cornerstone of any successful enterprise. Despite advances in software automation and sophisticated AI tools, manual processes persist in most organizations. Service as software aims to bridge these crucial gaps by expanding the conception of cloud-based platform delivery.
An increasing number of vendors are stepping into the service as software realm, including organizations such as Klarna, Moonhub, Thoughtful Automation, Crescendo AI, Converzai, Adept, and Inflection AI. These companies typically offer pre-engineered agents to handle specific tasks, many of which feature voice-enabled interfaces and interactions.
Early adopters are already leveraging these tools for niche tasks, predominantly involving document processing, medical transcription, and automated invoice processing, Strefford notes. These cases often utilize unstructured data found in documents, messages, images, and various forms, converting it into structured, actionable information.
Simply put, service as software conducts the work itself rather than providing tools for humans to complete the tasks. “It’s more than just scanning data for patterns or matches. It determines actions based on the information,” Strefford elaborates.
For instance, AI-driven accounting software can automatically categorize transactions, file taxes, and monitor compliance. AI-powered marketing and sales solutions can identify leads, draft personalized messages, and even autonomously schedule calls or demos with interested prospects. Similarly, AI-enabled content creation tools can generate market research reports, legal summaries, or product descriptions from raw data.
SaaS 2.0 has become feasible due to significant advancements in AI systems over recent years. Beyond the much-talked-about generative AI and large language models (LLMs), machine learning and deep learning have also progressed significantly. “LLMs have facilitated service as software,” says Strefford, “but traditional machine learning algorithms remain extremely valuable, particularly in predictive analytics and workflow optimization.”
Combining various AI components results in outcomes greater than the sum of their parts. Giron explains, “AI continuously analyzes interactions, learns from both successes and failures, and enhances its performance over time. This continuous learning loop ensures that service delivery becomes increasingly intelligent, personalized, and effective.”
A major advantage of the service as software model is that it massively simplifies AI adoption while automating 50 to 70% or more of interactions, according to Giron. Instead of constructing complex AI models in-house, organizations can opt for pre-packaged solutions that deliver pre-designed AI-driven workflows. Like conventional SaaS, updates and patches occur seamlessly.
This results in access to new features and capabilities as the service provider introduces them. The approach fosters a continuous learning and optimization loop, promoting a more intelligent, personalized, and efficient work model, says Giron. SaaS 2.0 also supports a strategic framework that prioritizes measurable business outcomes and performance metrics.
However, human oversight remains crucial, at least for now. Strefford advocates a three-tiered model, especially as organizations familiarize themselves with the domain and begin pilot projects. He recommends full automation for low-risk tasks, human-AI collaboration for medium-risk activities, and maintaining human-led processes for high-value or high-risk operations.
“Trust is the key,” Strefford asserts. “It’s essential to comprehend the potential costs and repercussions of a system making an incorrect prediction or executing an incorrect action.” Naturally, these considerations differ by organization and industry, and business and IT leaders must factor in regulatory requirements, board confidence, geopolitical events, and the overall risk tolerance.
CIOs and IT leaders are advised to start modestly and scale iteratively. As organizations build confidence and trust, they can extend autonomy within a SaaS 2.0 component. “More AI initiatives have failed from overly ambitious beginnings rather than cautious starts,” Strefford emphasizes. Consequently, understanding the entire workflow, implementing oversight and protections, establishing measurement and validation tools, and maintaining a focus on outcomes is imperative.
Giron emphasizes that several factors can make or break an initiative. The quality of data and the ability to integrate across systems are vital. A framework for standardization is critical, including the tasks of cleaning, standardizing, and preparing legacy data. “Data labeling and annotation can be a time-consuming and resource-intensive task, requiring specialized expertise and tools,” he notes. Simultaneously, identifying and addressing potential biases within the data, alongside focusing on security and regulatory risks, remains a priority.
