Intelligent Software as a Service:

Will Generative AI Unlock the True Potential of Vertical SaaS?

Mike Heller
Navitas Capital
Published in
13 min readApr 12


Generative AI: A Primer

Chat-GPT, DALL-E, Stable Diffusion, Midjourney, Character.AI, Bard, Cohere. The tech world has been pulsing with buzz for the past several months related to this new wave of Generative AI solutions. They started as fun novelties serving up interesting content for Twitter threads, LinkedIn posts, and profile pictures. Even as novelties, leading thinkers quickly recognized the potential power of this technology to revolutionize our digital lives.

“It’s pretty stunning that what I’m seeing in AI just in the last 12 months is every bit as important as the PC, the PC with GUI [graphical user interface], or the internet. As the four most important milestones in digital technology, this ranks up there.” — Bill Gates

With Microsoft’s announcement and launch of the AI-powered Bing search engine and Edge browser (in partnership with OpenAI), Generative AI technology is officially going mainstream. Over the next several years, Generative AI technology will become increasingly embedded in the software and products we use every day. Before we dive too deep into exploring Generative AI’s potential impact, let’s take a step back and define this technology.

“Generative AI” is a pervasive but not broadly well-defined or understood concept. One could argue that, due to the rapid pace of innovation, the definition of Generative AI is still largely malleable and open to interpretation. For the purposes of this piece, we will rely on a definition proposed by the authors of All In on AI: How Smart Companies Win Big With Artificial Intelligence in a recent Fast Company article:

“Generative AI refers to artificial intelligence that can generate novel content, rather than simply analyzing or acting on existing data.”

The concept of creation is key to understanding Generative AI. Rather than crunching large data sets to deliver insights, Generative AI has the capability to create new pieces of content in a variety of different formats. Two of the most common types of content creation for Generative AI today are text and images.

  • Large Language Models (LLMs): According to Nvidia, an LLM is a “deep learning algorithm that can recognize, summarize, translate, predict, and generate text and other content based on knowledge gained from massive datasets¹.” Chat-GPT is the most notable example of a Large Language Model. However, several prominent LLMs exist — each with their own strengths and weaknesses (for a deeper dive into relative performance across LLMs, see Stanford’s HELM project). Due to the extraordinary scale of required compute power, well-funded companies are typically best-positioned to build high-quality LLMs.
  • Text-to-Image: These models interpret natural language descriptions of an image and create an entirely unique image that matches the user’s description. DALL-E, Stable Diffusion, and Midjourney are some of the best-known products in this segment of the market. These products use LLMs to encode text for image synthesis and then apply diffusion models to generate and enhance images. At a high level, diffusion models are largely constructed from publicly available internet data and systematically add noise and remove noise from images to enhance their photorealism and resolution (for more on diffusion models, Scale AI offers a helpful guide).

Equipped with a standard definition and basic understanding of Generative AI, we can now explore the potential of this emerging technology to create a new category of software — Intelligent Software as a Service (aka iSaaS) — that has the power to capitalize on the promise of vertical SaaS. In highlighting the opportunity for Intelligent Software as a Service, we will first review how software has evolved over the past 25 years. We will consider how the advent of enterprise software spurred the emergence of vertical SaaS, which has in turn laid the groundwork for Intelligent Software as a Service. Then we will examine a few examples within the Navitas Capital portfolio of companies applying Generative AI to create Intelligent SaaS products that maximize outcomes in vertical applications.

The Advent of Enterprise Software

We can trace the origins of enterprise software back to the material requirements planning (MRP) systems manufacturers used in the 1960s to better track inventory and production. As the technology continued to evolve through the 1990s, businesses outside of manufacturing began to adopt similar systems, and Gartner coined the term enterprise resource planning (ERP). The Oracles and SAPs of the world built databases for enterprise customers that united business information across functions (e.g., accounting, sales, HR, engineering, etc.). For the first time, these ERPs digitized data into a single source of truth for an organization.²

The 2000s marked a new generation of enterprise software companies, as data migrated from private company servers to the cloud. Salesforce (founded in 1999) was an early pioneer in delivering software via the cloud. Salesforce’s CRM product was not only easier to use than traditional ERPs, but it also benefitted from streamlined distribution. Customers could easily access Salesforce via the web, always working off of the most current software and real-time data. This novel method of distribution increased software penetration beyond large enterprises by lowering barriers to adoption.

The enterprise software solutions of the 1990s and 2000s were often horizontal in nature. That is, they could facilitate similar workflows across industries. Think Salesforce, Dropbox, Quickbooks, etc. These products are intentionally flexible and customizable. The broad adoption of horizontal software is largely based on adapting to a range of user needs. However, significant expertise and/or investment is often required to harness the true power of the software.

For example, most businesses today hire dedicated specialists to configure and manage their CRM systems. Although Salesforce’s basic functions are intuitive and accessible, only experienced users maximize its capabilities. Without expertise and/or investment, enterprise software often only delivers a portion of its promise and potential.

The Emergence of Vertical SaaS

Enter vertical SaaS — purpose-built software designed to optimally serve a set of workflows within a specific industry. The 2010s saw an “unbundling” of larger horizontal enterprise software into a series of industry-specific verticals. What are the benefits of vertical SaaS? It all comes down to adoption and retention.

  • Adoption: Let’s say you attend a conference and hear about a new industry-specific software product from a few well-respected colleagues. Mired in a year-long horizontal ERP customization that is sucking up time and resources, you are intrigued by the ease of deployment that your colleagues highlight. Not only is the software built for your organization’s needs, but it also includes out-of-the-box integrations and connections to other products you use today. These key characteristics of vertical SaaS remove potential barriers to adoption.
  • Retention: Once you adopt the new vertical software, the product workflows align with existing business processes. Training employees and customizing elements of the product take a fraction of the time that a horizontal enterprise software customization would. The nomenclature is familiar, and the user interface is intuitive. Put simply, it just works.

As vertical SaaS companies demonstrated advantaged metrics, venture investors began to flock to the category. In-market network effects help contribute to “winner-take-most” dynamics, increasing potential market share for vertical SaaS companies. Further, as Bessemer Venture Partners highlights in a 2017 paper, vertical software companies often operate with greater capital efficiency due to their targeted focus. Over the past decade, we have seen the emergence of category winners, such as Procore, Toast, and ServiceTitan. According to another Bessemer Venture Partners’ analysis, public vertical software companies saw a ~10x increase in market cap between 2010 and 2020.

Vertical SaaS companies earned this economic success by offering customers a better, faster, and cheaper deployment with a product suited to their specific needs that does not require advanced expertise or unique customization. ServiceTitan offers field technicians a tailor-made mobile solution to improve job cycles, Procore provides unique job costing capabilities, Toast extends its functionality to the consumer with mobile order and payment. Despite these advances, vertical software is only as good as its users. Users must still apply their own intelligence to manually execute workflows within the product. No matter how intuitive or seamless these tasks and actions may be, the manual nature of their execution introduces some element of risk.

The Potential of Intelligent SaaS

What if software products could limit reliance on manual workflows and still produce high-quality outcomes? With recent advances in Generative AI, traditional user tasks can be automated and optimized to an extent that was previously impossible. Generative AI’s creative ability can entirely remove big chunks of workflow inherent in software today.

We are already starting to see this trend play out in Microsoft’s collaboration with OpenAI. A traditional Bing search workflow resembles something like the following…

  1. Type in a narrow, well-defined query
  2. Review results in the form of website links
  3. Click on various links
  4. Read and synthesize information

Now let’s examine how the workflow might look for an AI-powered Bing search…

  1. Type in a broad, amorphous query
  2. Read already synthesized information

The inclusion of Generative AI technology completely eliminates steps 2 and 3 of the traditional workflow and vastly improves steps 1 and 4. With an AI-powered search, users can access information with even less friction and more efficiency.

But AI-powered search is far from perfect. It doesn’t take much internet sleuthing to find examples of Bing providing wrong information or appearing just plain creepy. The Large Language Models behind Bing’s AI-powered search are incredibly broad, accessing sources of information across the internet. But there is a limit to the publicly available data that Generative AI can use to train advanced models. What happens when AI has read everything? Perhaps more importantly, how can we best apply Generative AI to commercial applications when we know it is prone to “hallucinations” and mistakes?

The power and quality of Generative AI is magnified when applied via vertical software products. Access to high-quality training data is essential in developing high-performing AI models. Vertical SaaS companies possess a deep and narrow data set on their company servers. With the proper technical expertise, this novel data can be incorporated into the latest research to deliver higher-quality results and outcomes than LLMs can produce alone. Whereas LLMs are often an unpredictable black box that can produce the likes of Sydney, by applying smaller models with a more defined data set, iSaaS companies can more effectively establish guardrails in the commercial context. Further, with a targeted focus and fewer model parameters, iSaaS businesses typically avoid spending millions of dollars on compute power, enabling more capital-efficient growth.

Intelligent products powered by Generative AI technology are the new differentiator for vertical software. The traditional benefits of vertical SaaS lay the groundwork for an impactful deployment of Generative AI technology.

  • Purpose Built: Vertical SaaS companies have a nuanced understanding of standard workflows in a specific industry and function. This understanding informs the thoughtful application of Generative AI to remove certain workflows while preserving others.
  • Unique Data: Industry-specific information improves the quality of Generative AI output. Models can be trained with private data sources to optimize for specific outcomes with greater accuracy and higher quality. This approach establishes clear guardrails and limits “hallucinations.”
  • Integrations: Often in the commercial context, the correct answer relies on data that can only be accessed via integrations with other software products. Vertical SaaS companies have already invested heavily in these types of integrations to best serve their customers and can easily access these answers.

Intelligent SaaS is more than just a future theoretical business model. Businesses in the market today are capitalizing on this approach to rapidly gain adoption and distinguish themselves from traditional vertical SaaS companies. We can look to the Navitas Capital portfolio to explore a couple of examples of iSaaS companies.


EliseAI is an advanced AI assistant powering business conversations that automatically answers emails, texts, and phone calls by applying Generative AI technology. EliseAI’s initial vertical focus is on residential real estate rentals, working with property managers and owners. For most consumers, renting an apartment is a stressful process, marred by slow, incomplete, and inconsistent responses from leasing teams and agents. With the power of Generative AI, Elise always responds in a timely manner with 100% informational accuracy and feels like a real person.

EliseAI uses LLMs, other models, and (most importantly) its database of tens of millions of industry-specific conversations. By combining this novel training data with significant technical expertise and domain knowledge, EliseAI generates industry-specific conversations with superior coverage and enhanced accuracy. Said differently, in responding to consumers, EliseAI doesn’t just act like another generic human, she acts like a best-in-class leasing agent and property manager. When shopping for an apartment, consumers often require information that is not publicly available, such as tour availability, unit specifics, and building policies. EliseAI has access to this information through existing integrations and can immediately and accurately answer consumers’ questions in a way an LLM never could.

Like most industries, multifamily businesses have their choice of deploying enterprise software, vertical SaaS, or Intelligent SaaS to manage their sales processes. Some leasing teams decide to use products like Salesforce or HubSpot and run into the typical challenges of applying horizontal SaaS to an industry-specific use case. They hire CRM experts to customize workflows and architect data structures. Others decide to adopt vertical SaaS solutions, such as CRMs tailored to the apartment leasing journey. These CRMs and lead management tools are better suited to existing workflows and offer pre-built connections to relevant data sources. But they still have a speed limit. The software is only as fast, accurate, and consistent as the user.

In contrast, iSaaS solutions, like EliseAI, drastically reduce the need for this type of software altogether. EliseAI automatically updates renter status, handles tasks and reminders, and manages correspondence — thus converting what was once the slowest part of the process into now the fastest. This automation frees up leasing agents to focus on higher-impact activities while eliminating mistakes and avoiding incomplete and/or poor data entry. They no longer spend tedious hours engaged in vertical SaaS workflows or submit to hours of onboarding and training.

At the corporate level, EliseAI removes previous constraints to testing various types of marketing spend. With a traditional CRM, each dollar spent acquiring a new lead is magnified by the time and resources required to nurture and convert that lead. Low-quality and low-intent leads are especially expensive. With EliseAI, the cost to nurture and convert leads remains consistent regardless of lead quality or intent to purchase. This empowers businesses to test different types of marketing spend to improve conversion.

In terms of ultimate business impact, the numbers speak for themselves. In a published case study, EliseAI generated a 310%+ increase in appointments booked, a 110%+ increase in appointment conversion, and 2+ hours saved per day for leasing agents. Equipped with Generative AI capabilities, EliseAI brings Intelligence to vertical SaaS and delivers a step function increase in value creation compared to traditional software products.

Oda Studio

Oda Studio is a virtual design software platform for the built world that is applying Generative AI technology to automatically enhance images, design spaces, and manage digital content. Anyone who has spent time in a poorly designed space recognizes the power of design choices to impact how a space feels and how you feel while in it. Design matters. Especially when deciding where to live, we like to understand how a space might look and feel. Achieving this understanding today relies on a combination of professional photographers, interior designers, and digital artists. Staging a home often amounts to managing an entire project across multiple stakeholders. Oda Studio condenses this project into a single Intelligent SaaS platform. Users upload a photo of their space, and Oda Studio’s algorithms automatically enhance the image according to user preferences, virtually stage various room layouts, and visualize a range of photorealistic furniture options.

Although the DALL-E’s of the world can generate standalone portraits of Darth Vader ice fishing, creating unique layouts within the context of an existing space presents a different type of challenge. Design elements must relate to each other and coalesce to create a comprehensive scene. Without this context-dependent Generative AI expertise, lamps can end up on top of beds and chairs can end up on ceilings. As users personalize their design preferences on the platform, Oda Studio’s models improve to deliver even more optimized design options. These Generative AI design options are then embedded within existing marketing and sales channels (e.g., property websites, listing providers, retailers, etc.). Consumers can seamlessly interact with Oda Studio’s iSaaS products as part of their standard shopping experience.

Virtual design capabilities do not represent a new category of software. Adobe’s Photoshop empowers designers across industries to “create gorgeous photos, rich graphics, and incredible art³.” Trimble’s SketchUp offers 3D modeling software purpose-built for real estate and construction. However, existing enterprise software and vertical SaaS solutions fall victim to typical shortcomings. They are limited by user expertise, lack of integrations, and efficiency.

As an Intelligent SaaS solution, Oda Studio automatically generates design options that adapt to users’ preferences. Using Oda Studio, consumers and designers efficiently create accessible personalized spaces vs. theoretical renderings. By removing elements of the traditional workflow via iSaaS, Oda Studio delivers higher-quality designs that feel like home and, in doing so, helps accelerate the purchasing process. Ultimately, Oda Studio streamlines and customizes the design process for living and workspaces, enabling built-world businesses to save thousands of dollars and weeks of work on their projects.

Software Ate the World; Can it Now Achieve its True Potential?

Enterprise software helped to bring business online. Vertical SaaS improved the user experience by offering easier and more intuitive purpose-built workflows. But any software workflow has a ceiling of efficiency and quality. The promise of iSaaS lies in the thoughtful application of Generative AI technology to remove manual workflows. Through an iSaaS product suite, businesses like EliseAI and Oda Studio consistently deliver higher quality business outcomes at a faster speed with more accuracy and personalization. Intelligent SaaS — it could be the next wave of innovation that forges the potential of Generative AI and vertical SaaS into a new reality.

Disclaimer: For the purposes of clarity, the author has distilled complex topics into key points. Many of the points contained herein could be debated and may include a variety of exceptions.






Mike Heller
Navitas Capital

LA-based investor in technologies creating value for real estate and construction