Innovation and excitement are surging in applied AI. Recent displays of capabilities in areas such as generative AI have further boosted the technology’s profile.
Of course, the industry is still young, with plenty of opportunities for growth as organizations increase their adoption of AI and their spending on the technology. McKinsey analysis suggests that the value at stake from AI can reach $15 trillion.1 For now, however, market penetration is still very low: about 50 to 60 percent of companies have deployed AI but have not scaled it.2 We estimate a ceiling of about 30 percent market adoption in AI’s mature use cases for individual business functions. The technology’s commercial potential has drawn a flood of private investment, which hit a record $93.5 billion in 2021.3
This attention is partly based on the industry’s high growth. But at the same time, our analysis of software and applied-AI companies shows that AI companies are less efficient at generating revenue compared with their software-as-a-service (SaaS) counterparts. Still, the sector will likely continue to be commercially and technologically significant. We offer a short perspective on the possible direction of the growing industry.
How AI revenue looks different from SaaS revenue
Our analysis of 187 software and AI companies in our benchmark database shows that while applied-AI companies have promising growth profiles, their revenue is less repeatable and efficient compared with that of traditional SaaS companies (exhibit).
Specifically, the average applied-AI company has a growth efficiency (the ratio of realized annual recurring revenue to marketing and sales spending) of about 70 percent, compared with about 95 percent for their SaaS counterparts. According to our analysis, three key factors appear to contribute to this difference.
Higher costs of marketing and sales. Prospective buyers don’t necessarily consider applied-AI solutions and their propositions to be mission critical, particularly because companies may lack a robust track record of applied-AI use cases—although the buzz surrounding generative AI is starting to shift this mindset. At the same time, the sales path for applied-AI solutions is often unclear, including considerations such as which stakeholders have budgets for AI solutions. The lack of both clear targets and an obvious way forward extends the sales cycle. As a result, the lifetime value of applied-AI customers is lower than it is for SaaS customers, even with nearly identical average customer churns of about 15 percent per year. However, the push to adopt generative AI may change this in the near future.
Less efficient spending at scale. The applied-AI companies in the top quartile of our data set spend 40 percent of their revenue on general and administrative expenses and 50 percent of their revenue on R&D. Both numbers are more than ten percentage points higher than their equivalents for their SaaS peers.
Traditional software is developed once and can be shipped an infinite number of times. In contrast, applied-AI companies’ spending efficiency tends to be limited as they scale. Applied-AI companies take on manual activities such as data cleansing and model tuning more often, pay more for scarce AI talent, work more directly with data because of demand for specialization and responsiveness, and bear considerable ongoing costs for data storage and computing for product development.
Costly professional services. Because of the scarcity and high cost of AI talent, the strongest-performing applied-AI companies provide AI-specific professional services to their buyers.
However, these offerings may affect revenue growth: fees may increase to the total contract value, but the services are often not easily repeatable, cannot rapidly scale, and are costly to maintain.
Considerations when thinking about applied-AI companies
Six considerations can help stakeholders think about applied-AI companies as the industry develops.
A clear ROI story
Applied-AI investments would ideally have a clear market niche with adjacencies. In informal interactions, experts in the industry suggest that a company with a niche in the $15 trillion total addressable market might have a serviceable available market worth $1 billion to $3 billion and focus on a market segment that combines factors such as geography, industry, the end customer, and business function.
The ideal applied-AI company would also have a combination of captive users, private data sets, or other protective assets stemming from their technology, data, or machine-learning capabilities. These advantages can help them stay ahead of competitors that may use open-source tools or public data to develop their offerings.
Stakeholders could also look for signs of early customer evangelism. This authentic, unsolicited enthusiasm for their products could confirm that the AI use case is critical to their customers, rather than a nice to have. Beyond that, network effects—from data assets or expertise—may be a sign that a company’s offerings could be resistant to commoditization and margin compression.
Better customer segmentation models
While the value of applied AI is generally accepted, not all buyers are convinced that applied-AI solutions are critical. Go-to-market strategies should therefore use a segmentation approach that emphasizes buyer segments in which sustaining an operating model without innovative technologies is increasingly difficult. That is, applied-AI solutions may be most valuable to buyers in industries that are highly competitive and in which technology can provide a critical advantage. Of course, this approach to buyer segmentation should complement—not replace—traditional ways of articulating companies’ value proposition.
A plan for multipersona marketing
Decision making related to adopting an AI solution is often shared among managers, business users, data scientists, and IT professionals. Such buyer personas often play different roles in the purchasing and adoption journey, such as decision maker, champion, and end user. Since the personas and roles will likely differ, AI solutions’ value proposition might be distinct for each persona to best address their pain points, needs, and goals. These considerations may require the coordinated use of diverse channels on the way to a purchasing consensus.
Captive workflows to defend against competition
The commercially and strategically strongest AI and machine-learning product companies build workflows that capture the entire user process and, crucially, create feedback loops and reinforcement learning so end users can contribute to the AI by confirming or disagreeing with its outputs.
Of course, just as with SaaS, lock-in and customer stickiness will likely come from a strong go-to-market approach, control over the postsales process, and a deep understanding of vertical application (in which a solution is designed for the specific needs of a market, industry, or company).
New efficiency levers
AI companies are seeing advantages in developing ways to optimize spending in product development. R&D partnerships and improved tooling in data and model development in product engineering, also known as MLOps, can help. Any efficiencies can be built into the product, which could help the company create revenue more efficiently.
A multiyear go-to-market road map to becoming embedded solutions
As individual AI companies gain traction and expand their market share, they would ideally become embedded in their customers’ processes and capture their customers for the long term. To do this, AI companies could plan for three key waves of adoption.
The first wave is early adoption, which is ongoing. Applied-AI solutions currently provide improvements to traditional business processes, prove their value, and manage risks en route to establishing product–market fit. In this phase, AI solutions would be easy to use, likely by providing an interface that fits into customers’ current ways of working.
The second wave is step-change innovation, which could occur over the course of about three years after a product–market fit is established. In this phase, AI becomes embedded in business processes and operates with significant human oversight.
Last, AI might disrupt conventional workflows, with AI solutions taking the place of existing processes. This phase requires mature and tested technology that is highly embedded in businesses, so much so that stakeholders would have a hard time imagining working without the AI-powered solutions. This relationship with AI-powered solutions would spur businesses to reorganize around them, becoming long-term customers.
Applied AI has attracted a surge of investor attention thanks to rocketing public awareness, but a thoughtful approach can help companies make their way toward sustainability and continued growth.
