Three reasons AI agents will be the biggest software market yet
The AI agent market is emerging, and it will be very big
AI agents, artificially intelligent programs that use large language models and a variety of APIs to reason, act, and accomplish a goal, are in their infancy. Initial projects like AutoGPT and BabyAGI are giving way towards more completely packaged tools like Reworkd, Fleetworks, and others. The emergence of these tools will cause discussions about pricing, value, and market size. AI agents benefit from three things that will help them grow into the largest software market yet.
The evolution of software usage and pricing
AI agents can have a higher price point since they do work, not just enable it
AI agents will replicate employee skillsets, meaning there could be hundreds of AI agents within an organization
Evolution of software pricing
Software pricing has moved from custom development to per mainframe license to per-employee-per month subscriptions. The change in pricing mechanism was driven by hardware improvements (more capable machines at a lower price, meant more machines) and greater software usage overall. This allowed software costs to be spread over a wider user base and the per machine price to drop significantly.
1960s programming services - Fixed fee or staffing services model
1970s software for mainframes - Per mainframe, charging between $5k and $100k1
1980s through 90s software for PCs - With the development of PCs and client/server model of computers, software was sold on a per terminal/machine model. For business market this could be $500 to $5k per terminal, leading to 7 and 8 figure sales contracts.
Modern software pricing: per employee per month or usage based
Two major buckets of software pricing today are per employee per month subscription based pricing and usage based pricing. Subscription based pricing on a per employee basis is a continued development from the traditional per machine based pricing of the past. The more users of the software, the more a company pays.
The software market has grown propelled by the number of computers and the number of applications per company
In 1980, the size of the software market was estimated at ~$6B2 and there were an estimated 327,000 PCs sold3 (most of the software sales were to mainframe and minicomputer users). Today, the software market is estimated at over $600B4 and over 200M PCs are sold annually.5 The number of software applications used by a company continues to grow as well. The growth in machines and the growth in applications per company drove a large increase in the overall size of the software market.
The software market provides a comparison for the potential AI agent market, but since AI Agents do work, not just enable it. The labor market provides another comparison for the AI agent market size.
The labor market is way bigger than the software market
Software in a business environment is used to help workers be more productive. This means that software is often priced as a fraction/percentage of the types of workers they are helping. General AI, and AI agents, allow startups to sell work, not software. Looking at the labor market, we get a sense of how big the market for paying people to do work is:
The size of US payroll is $9.4T (no, this is not a good comp, but its a comp)
The size of the US staffing market, how much people other companies for work is $218B6
The size of the US business process outsourcing market is ~$90b7
People pay a lot to get work done. People will not pay nearly as much for software as they do for people, so these comps aren’t reasonable. However, there is a segment between software and people, tech enabled services, that is more like AI agents.
Tech enabled services: bridging software and people
AI agents can be sold as doing work, not just providing a store of information or information retrieval. This means that pricing can be done more in line with services. According to Bessemer Venture Partners, tech enabled services address a market opportunity 10x that of SaaS alone.8
Companies will eventually have hundreds of AI agents
AI agents will initially carve out small functions done by an employee, but will eventually replicate specific tasks and skills done employee. Employee skills taxonomy efforts show thousands of skills. Similar to how companies have hundreds of software applications, they will have hundreds to thousands of AI agents replicating the specific skill sets of employees.
The initial entry markets for AI agents will be similar to those for business process outsourcing: areas that companies are already comfortable outsourcing and can be broken down into smaller tasks.
AI agents will get priced on a per task/skill set basis, likely with a usage based component. Because these agents will take over tasks that are done by outsources, AI agents will be priced higher than software alone, leading to high contract values.
Key Takeaways
Historical pricing and evolution: The software market has moved from pricing software on a one-off basis to per machine and then to per user. This opens up a move to per AI agent task basis.
Market Size Comparison: Because AI agents perform work, they can be priced higher than traditional software, and more like a services.
AI agent proliferation and usage: Companies will likely have hundreds to thousands of AI agents, with usage based pricing, leading to a very large market for AI agents.
Conclusion
AI agents stand poised to transform not just the software industry but potentially many aspects of our day-to-day lives. As their capabilities grow, so too will the market opportunities they offer.
From Airline Reservations to Sonic the Hedgehog, A History of the Software Industry, By Martin Campbell-Kelly
From Airline Reservations to Sonic the Hedgehog, A History of the Software Industry, By Martin Campbell-Kelly, Table 1.1
https://www.nytimes.com/1981/08/13/business/big-ibm-s-little-computer.html#:~:text=International%20Data%20estimates%20that%20327%2C000,to%201.3%20million%20by%201985.
https://www.statista.com/outlook/cmo/consumer-electronics/computing/worldwide
https://www.statista.com/statistics/273495/global-shipments-of-personal-computers-since-2006/
https://pgcgroup.com/blog/2023-us-staffing-forecast#:~:text=The%20US%20staffing%20industry%20had,%24212%20billion%202022%20revenue%20projections.
https://www.grandviewresearch.com/industry-analysis/business-process-outsourcing-bpo-market#:~:text=The%20global%20business%20process%20outsourcing%20market%20size%20was%20estimated%20at,business%20process%20outsourcing%20market%20growth%3F
https://www.bvp.com/atlas/benchmarks-for-growing-health-tech-businesses#Insights-on-tech-enabled-services