There are many reasons to get excited about AI. One is that it has the potential to reshape the $600B Software industry. Another much bigger reason is that AI may also reshape multiple $1T+ services industries, unlocking new markets and business models. There’s a rise in companies that increasingly blend AI, software, and services and do so in ways that will change how investors and founders think about services.
While infusing software and services is not a new concept—many have written about “AI-first services” recently and “Full Stack Startups” as far back as 2014—the timing for a renaissance couldn’t be better. Recent advancements in LLMs and diffusion models can remedy many issues that historically plagued service businesses: they have lower margins, are tough to scale, are less sticky, non-recurring, and yield lower revenue multiples. But, by blending AI innovations and software foundations with services, margins, scalability, and more will improve.
It’s important to note that services can also benefit AI. By owning both the AI fine-tuning and the frontline activities, startups can use services to improve data collection and feedback loops, accelerate go-to-market, and fortify moats in increasingly competitive categories. This symbiotic relationship between AI and services is just starting and will be critical to a new wave of Diagonal Software companies.
Diagonal Software
Most companies fusing software, AI, and services will also combine vertical and horizontal approaches in a single business model. This mixing of both vertical and horizontal results in Diagonal Software. Diagonal Software can come in many flavors, but here are three particularly exciting examples that reposition services as a strength: 1) Core Diagonal, 2) Back Office as a Service, and 3) Robotic Services.
Core Diagonal
Pairing a vertical software core with a horizontal services distribution layer, now aided by AI, is an increasingly attractive tactic founders can leverage. By adopting this model, AI can now transform internal services and customer interactions with the help of natural language interfaces, improved chatbots, and AI Agents to perform specific tasks previously maintained by humans.
Services can also improve the GTM and TAM for these disruptors in many verticals. First, GTM can be tough in either overly fragmented markets, like hair salons, or concentrated verticals, like autos. By going Diagonal and delivering a horizontal service, you would own the distribution and therefore bypass such GTM challenges. Second, there are verticals where the software market is niche, but the end market is huge. Here, a Diagonal model will meaningfully expand the TAM.
For example, the Financial Advising industry employs 300,000 financial advisors across 15,000 registered firms (according to BLS and IAA). Vertical Software startups can find success selling to financial advisors, like Envestnet ($2.9B market cap) or Advent Software (acquired by SS&C for $2.7B). Or, a new Diagonal Software company could build a full stack financial advising firm and go after the likes of Charles Schwab ($120B market cap), Ameriprise ($40B market cap), and Fidelity ($14B market cap).
The Diagonal software model can apply to many industries, such as:
- Instead of legal software, build a new AI-first law firm
- Instead of education software, start a school that fully embraces personalized learning
- Instead of CPA and tax software, give the world a next-gen RIA
Some recent examples:
- Sora Schools — rather than selling education software to schools, Sora combined new technologies with human instructions to offer an alternative to traditional Middle Schools and High Schools for anyone.
- Metropolis — instead of only selling AI & software to parking facilities, Metropolis recently went full stack and acquired a $1.7B parking lot company to start managing 2M+ parking spaces and providing more horizontal services.
Back Office as a Service
Another opportunity in Diagonal Software is to combine a horizontal software core (i.e. Finance, HR, IT) with an AI services layer to provide these essential functions as a service. In this scenario, an AI Back Office startup will likely first pursue a particular market segment or vertical, blending vertical and horizontal approaches.
Back Office as a Service continues a trend that’s been around forever: non-core parts of a business tend to be outsourced when a third-party provider offers superior economics. A great technology-centric example is AWS—if operating infrastructure isn’t your company’s core competency, outsource it and consume Infrastructure as a Service.
However human-centric business components have been slower to take off and less attractive due to the usual services limitations. Yet there are still many successful examples, such as:
- Trinet — the outsourced HR provider has a current market cap of $6B and does over $5B in revenue.
- Genpact — outsourced Customer Support propelled Genpact to a $6B market cap.
- Wolters Kluwer — the $39B company provides outsourced finance and compliance services.
Advancements in AI will again improve the economics of outsourced business services through higher utilization of humans, better customer experiences, and improvement feedback loops. We expect new success stories across areas of the Back Office or non-critical business functions. There are many potential examples:
- Instead of contact center software, build a new contact center
- Instead of CFO co-pilot, build a better accounting and bookkeeping firm
- Instead of recruiting software, build a new recruiting firm
- Instead of SecOps software, build a new security operations center
Startup examples exist and Felicis has invested in several including:
- Glencoco — takes the non-core part of Sales - cold-calling and lead-gen - and provides an AI-enabled SDR service & marketplace.
- Juniper Square — offers software, full finance, and admin servicing for various asset management markets.
Robotic Services
Another opportunity is to take a Diagonal model and introduce robots to replace traditional services. Robotics is on the cusp of its “ChatGPT” moment. Recent seminal AI advancements like RT-2 coupled with cheaper commodity hardware are placing robotics companies in the limelight in 2024.
Combining AI, software, services, and robotics is an incredibly compelling approach. Over the last year, we've seen a blend of frontier models, scaled training data, and human feedback expand to different modalities, like images, audio, and video. However, training data can be tough to come by for industrial and robotic use cases. Robots don't just take visual data as input and spit out visual data as output — robots move and take action in the world. They’ve also been expensive, and it costs a lot to gather training data from one robot in one environment.For these reasons, combining a Robotic AI business with end-services the robots perform will be a more efficient way to gather new data, improving the performance of future robotic intelligence models in turn. Imitation learning can also accelerate time to value for automating net new services.
- Instead of manufacturing AI, build a new contract manufacturer
- Instead of trucking software, build a new trucking company
- Instead of architecture software, start a new home builder
- Instead of autonomous driving software, build autonomous vehicles
Examples:
- Hadrian — building highly automated full-stack factories rather than selling manufacturing software
- Monumental — selling a robotic bricklaying service instead of selling robots
- Cafe X — builds robotic-powered coffee kiosks rather than selling robotic baristas to large chains
Risks in Scale
With greater outcome potential and ambition comes greater risk. Scaling, capital intensity, and execution risks are all magnified in a Diagonal model. While we believe AI can now greatly increase the odds of success, we recognize some have tried and failed:
- Atrium — the hybrid legal software and law firm raised $75M but shut down after only 36 months.
- Convoy — after raising $260M and reaching a $3.8B valuation, the integrated tech & trucking company shut down just last year.
Louis Coppey recently summed up the risk of AI-first services, writing, “There’s no PMF risk anymore, there’s a ‘scalability risk.’”
Reasons to Believe
To conclude, here are the main reasons why combining services with software is a good idea:
Services are a massive TAM
The services market's immense size, highlighted by the professional and business services sector leading US GDP at $3.5T, significantly surpasses the global software market's $600B, presenting vast opportunities for Diagonal Software through AI-driven expansion.
Services can be a GTM accelerant and booster
A services wrapper can set your enterprise software apart and boost market share, as shown by CrowdStrike's initial managed endpoint security and Pennylane's accounting services bundling to fast-track dominance.
Customer expectations are changing
The horizontal customer service and interaction layer can be optimized with natural language interfaces (like prompts) and LLMs to create a more efficient and pleasant customer experience,
Agents are coming
AI Agents trained for specific tasks can begin to augment the cumbersome and low-margin manual components of the Services layer.
Data and feedback loops can be significant moats
By owning the frontlines, Diagonal Software companies gain industry-specific data and feedback, enabling them to refine models that emulate human services and learn from internal workflows and customer interactions.
Specific models with big potential
Diagonal Software companies are developing domain-specific models in various verticals (robotics, manufacturing, pharmaceuticals, and more), aiming to create robustly trained, fine-tuned market-redefining models that will justify their larger TAM.
Diagonal Software epitomizes the “big swings” that venture is meant to pursue, and some of the biggest success stories include companies like Tesla ($590B) and Uber ($150B). Ultimately, this AI era has the potential to deliver numerous new outlier companies across multiple verticals.
AI will fundamentally change the trajectory of many companies. For those wanting to build impressively big software companies, ignoring the services side of business no longer makes sense. If the main knock on services is that they are challenging to scale and low margin, AI will improve both.
AI is the catalyst that will elevate services to be a critical component of the next generation of great software companies.
