SaaS 2.0 - Service as a Software.
Software as a service is dead. Service as a software is the future.
The agentic wave is here. And it’s here to stay. On January 23rd, 2025, OpenAI launched Operator - a browser agent that can perform routine tasks for users. Operator can control your computer, write code, book travel tickets, and browse the internet.
In a way, agents are the opposite of SaaS - it’s Service as a Software.
If SaaS changed the way software was delivered, Service as a Software will change the way labor will be delivered.
https://x.com/bhalligan/status/1781323169313222852
For most of humanity, labor belonged to the exclusive domain of humans. But with AI agents that may not hold true.
AI agents have the potential to create a new kind of labor market - the Service as a Software market.
SaaS changed the way software was delivered by bringing the technology to customers via the cloud -- one that didn't require local installation or maintenance.
As a result, software began to "eat the world", disrupting and transforming entire industries.
SaaS helped businesses and end users create specific workflows that performed a specific task. For example, users could put structured data like users’ contact info into a CRM, which would not only give insights to the business but also form the building blocks for creating workflows that could streamline specific business functions.
But software has its limitations as well - the data needed to be structured and had to be entered manually (for the most part). The success of the system relied upon the data being filled accurately & consistently.
As a result, software became a tool that employees learned to manage & use on their own. The employees were the ones responsible to achieve specific business outcomes using these tools.
But this isn't the case with AI agents.
Unlike software, AI agents are autonomous robots, not just tools. They recall, make decisions, and act without human intervention. Imagine AI-driven services that don’t just help people do the work—but replace repetitive work entirely.
Agents can automate emails, code, design, and more—doing your work without you lifting a finger.
This is the promise of Service as Software–Autonomous AI agents executing service-facing tasks wrapped in software.
As a result, the actions of these agents can be tied to specific business outcomes.
The shift to Service as Software starts with focus on workforce spend, not software spend. That is where the true leverage lies.
Salesforce makes $35 billion in revenue annually but pales in comparison to the $1.1 trillion salary paid to sales & marketing professionals.
The total addressable market for services & software is $10 trillion compared to the $3B AI market.
Sam is bullish on AI Agents for the same reason I am. AI agents are the gateway to AGI -- holding within it the potential to change forever, the way we work.
Look at this spike in Google’s search interest for the search term “AI agents”.
But what is it that makes AI agents so powerful?
The thing that makes AI agents tick is the LLM models themselves & well as the cognitive architecture that supports this.
The breakthrough behind AI agents lies in inference compute. Previously, AI models generated responses based on pre-trained data without much reasoning.
Now, with inference compute, these models can pause to “think,” simulating events to predict the best outcomes. The longer an agent thinks, the better its results.
It now “knows” what is the best way to do a specific task.
In a way, it's similar to how self-driving cars work. Self driving cars collect data from their surroundings via sensory inputs like the camera, LIDAR etc, after which, this data is processed using Machine Learning & Deep learning algorithms. The algorithm is trained across a large dataset to recognize patterns and make predictions. This gives the vehicle the best chance to react appropriately to various driving scenarios without needing specific instructions for every possible situation encountered on the road.
AI agents also take in sensory input from the users’ screens, external databases, APIs etc. Once this data is collected, AI agents use a combination of ML, inference compute & reasoning to make decisions. They can learn from previous interactions and can execute multi-step tasks, all on their own, much like the self-driving cars.
This marks a shift as AI moves from pre-trained “fast thinking” models to inference-based “slow thinking” models.
In this new Service as software paradigm shift, there's plenty of space for everyone. Just like any market, there will be outsized winners but there will also be space for billion-dollar companies to form.
And the way the next wave of agentic startups will win is by targeting automatable pools of labor. There are three major service categories ripe for disruption:
Outsourced IT Services
Vertical Labor
In-House Work
AI agents will change how humans work -- but what does that look like? Which companies will win? What are the consequences of that happening?
In this essay, we'll explore the 3 major markets AI agents will disrupt. If you are curious to learn how you can use agents in your business or better yet build a business around agents, this essay will hopefully give you some clarity on that - let’s dive in, shall we?
Brought to you by:
Market Curve is a writing studio for tech companies, VCs & founders.
Here is what I can help you with —
Website copywriting
Longform blogs, essays, newsletters
Onboarding, drip, and churn emails
Assets like decks, e-books, guides, case-studies
Product UX copywriting
Ghostwriting content for busy founders/execs
Have worked with folks from YCombinator, 20VC and more. Sounds interesting?
In-House Roles: A $2.3 trillion market waiting to be disrupted by AI agents
“The future of autonomous agents looks like everybody becoming a manager.” - Yohei Nakajima, creator of BabyAGI
Traditionally, in-house roles have involved doing repetitive, recurring tasks that must be completed each time you perform the main big task.
For example, a Sales Development Representative (SDR) spends time on repetitive tasks like prospecting, writing emails, scheduling follow-ups, and arranging demos. What if you had an agent that could do all this for you autonomously?
YC-backed AI-SDR does exactly this. It warms up unlimited email boxes and learns from your data. SDRs can now just set up campaign goals, outreach channels, and sync their calendar.
Even Salesforce introduced two fully autonomous, AI-powered sales agents: The Einstein SDR Agent that can engage with inbound leads 24/7 and the Einstein Sales Coach Agent that offers real-time sales suggestions during calls.
Customer success teams are undergoing AI-driven transformations too. Sierra uses AI for customer interactions, enabling support teams to handle more queries with fewer human resources. AI agents like the one Zendesk has, are taking on the bulk of the work, while human employees focus on managing and supervising these agents.
AI recruiter agents like Converz can make calls, while AI sales agents like Docket can answer buyer questions . You can also create custom AI employees that handle multiple tasks together.
Imagine a system of AI agents working and interacting together across teams. You can set these up with tools like Lindy, Relevance AI, Gumloop or Beam.
As a result, there will be increased cross-pollination between teams—marketing will speak with sales, sales with product, product with marketing, etc. This in turn will create better products along with better data capture & attribution. More data means better AI learning.
If you're building for in-house roles, the playbook would look like this:
Identify repeatable, outcome-driven workflows: Automate predictable tasks with clear, measurable outcomes.
Start at the source of data creation: Capture raw data at the source for accurate insights and actions.
Partner with humans as copilots: Initially, work alongside human operators, then transition to full autonomy as the AI learns.
Focus on workforce Spend, not software spend: AI should replace labor costs, not just add to software budgets.
Shift to outcome-based pricing: Charge based on delivered results, aligning value with outcomes.
AI agents won’t just assist – they’ll fundamentally reshape how work is done at companies.
Vertical Industries are ripe for disruption by AI agents
“Autonomous agents will allow everyone to live like a head of state! Need something done? Just ask, and your agents will take care of the rest. Never again will you have to waste brainpower on the routine or mundane.”
Vertical industries like law, healthcare, and cybersecurity are primed for AI disruption. These fields thrive on language and unstructured text data—making them perfect for AI agents powered by LLMs.
Unlike traditional software, which relies on structured inputs, AI agents can reason through complex scenarios, make decisions, and take autonomous actions.
This report by Goldman Sachs shows the industries most likely to be impacted by AI & automation.
Take Harvey AI. It’s not just automating legal workflows; it’s transforming law firms by taking over tasks like contract review, due diligence, and litigation prep. Imagine a junior associate that never sleeps, misses a deadline, and constantly improves.
Similarly, EvenUp focuses on personal injury law, automating settlement demand letters and allowing attorneys to focus on strategy instead of paperwork.
In healthcare, Freed and Deepscribe help doctors reclaim their time. AI voice agents schedule appointments, transcribe patient interactions, and suggest diagnoses. AI Agent Tennr helps healthcare professionals read faxes and enter data.
Cybersecurity is transforming too. XBOW has developed AI-driven penetration testing that simulates cyberattacks to identify vulnerabilities faster and cheaper than human testers. What once required high-cost specialists is now done at scale, with better accuracy.
Building a vertical solution offers tremendous value–you can capture a dominant market, establish credibility, and build a moat around your community.
Your models will improve as you fine-tune them on domain knowledge. Since agents use reinforcement learning, the more tasks they do, the better they learn and the faster they perform. Coupled with the decreasing cost of inference computing, you wonder–what are these agents ultimately capable of?
Here’s the playbook on how to build in vertical AI:
Target language-heavy industries: Identify a language-heavy industry: Look for fields where unstructured text data—like contracts or medical notes—is a bottleneck for productivity.
Focus on a core pain point: Identify the most frustrating, repetitive task your audience faces and solve it first.
Leverage LLMs for workflows: Build agents that process unstructured data, reason through complex problems, and improve over time.
Bundle solutions over time: Start narrow, but bundle by adding products: Once you’ve automated a workflow, look for complementary tasks to create an integrated compound product to build long-term technical & distribution moats.
Fine-tune with domain-specific data: Use reinforcement learning to make your model an expert. The more tasks your agent handles, the smarter it becomes.
Start with a niche vertical use case to wedge into the market; add products layer on top and bundle adjacent use cases to build a compound product that’s defensible.
AI Agents will disrupt the $2.3 Trillion outsourced services market
“[Intelligent] autonomous agents are the natural endpoint of automation in general. In principle, an agent could be used to automate any other process. Once these agents become highly sophisticated and reliable, it is easy to imagine an exponential growth in automation across fields and industries.” - Bojan Tunguz, Machine Learning at NVIDIA
IT services is a $933 billion market, growing at a CAGR of 8%. The amount spent on outsourced IT services and business process services is $2.3 trillion.
But they’ve been constrained by slow delivery and low margins due to an over-reliance on manpower.
AI agents can transform this market by accelerating execution, creating software outcomes efficiently with faster delivery, increasing gross margins.
The trick here is to focus on workforce spend and not software spend.
Take Realfast AI, for example. It builds AI agents on Salesforce, automating workflows that require hours of manual input. Similarly, AirMDR uses AI assistants to manage detection and response in cybersecurity, cutting response times while improving accuracy.
Business process services like data entry or audio transcription involve repetitive, rule-based tasks. AI agents excel at these tasks. They can quickly understand context, process information, and take action. ,
AI agents don’t just reduce costs—they create a flywheel effect:
Faster delivery: By cutting development and maintenance time, AI agents reduce costs and increase efficiency.
Lower costs: Cheaper services attract more customers, driving demand for automation.
Greater efficiency: Increased adoption feeds data back into the system, improving performance.
Compounding growth: As agents become smarter, the market for AI-enabled services expands rapidly.
If you’re building AI agents for this space, here’s how to start:
Build LLM agents: Use pre-trained models to process unstructured data and perform complex tasks.
Integrate workflows: Ensure agents work seamlessly with human teams.
Record actions and decisions: Instrument interactions to create a rich dataset for future improvements.
Train with real-world data: Use these datasets to fine-tune your agents to adapt to unique workflows.
Adopt outcome-based pricing: Charge clients based on measurable results, creating clear ROI incentives.
Start by building on top of SOTA models & have a human copilot. Collaborate & collect data to fine tune the agents until it becomes autonomous. Expand to other IT services.
Who will win BIG?
Like always, the companies owning distribution will emerge on top. Google & Microsoft have an unfair advantage here.
With access to over 92% of global search traffic and a seamless ecosystem, Google is launching Jarvis that can create AI agents that streamline workflows across Search, Chrome, Gmail, Calendar, and Drive.
While Google focuses on consumers, Microsoft’s strength lies in the enterprise. Leveraging Microsoft 365’s adoption among businesses, Microsoft’s AI agents can assist in sending emails, managing user records, and enhancing productivity.
Apple is claiming its stake with Apple Intelligence, transforming Siri into a true AI agent managing apps and executing complex actions across the iPhone ecosystem. This positions Apple to add immense value to its user experience, making Siri more than a voice assistant.
Several LLM providers & infrastructure products will also be key players:
Anthropic’s Claude: Equipped with an AI agent that can navigate the web, create content, and write code, offering users a powerful AI toolkit.
Langchain and LlamaIndex: These frameworks enable developers to build sophisticated AI agent applications, allowing agents to retrieve and apply context from real-time data.
MemGPT: With advanced memory retention, MemGPT acts as a personal assistant that improves with each interaction, retaining past context to provide smarter, adaptive responses over time.
Mistral: Mistral also has an agent building feature where users can create custom agents.
https://x.com/i/status/1822309514948808802
https://x.com/i/status/1851338227925631102
Beyond platform-based AI giants, horizontal products like Notion, Airtable, and Slack are well-positioned to lead the hyper-bundling trend, integrating AI agents that support user workflows.
Notion’s AI-powered workflow capabilities and Slack’s in-app features indicate this shift.
Meanwhile, compound platforms like Rippling can gain a defensive edge by embedding intelligent agents within their products.
If AI agents are the next gold rush, “shovel-sellers” like Gumloop, CrewAI, Glean and Composio will win big by enabling users to create custom AI agents that automate workflows.
Zapier and Make hold a distinct advantage due to their established distribution, strong brand, and ability to capture significant market share by scaling and adapting to AI demand.
Zapier’s AI allows users to create automated, bespoke workflows across apps. You could create a workflow that would automatically take contacts from your email and populate your personal CRM, including up-to-date information about each contact.
If you really wanna go big, your best bet is to create a platform and sell shovels — aka enable users to create AI agents using your product.
Conclusion:
AI agents are here, and they are changing the way we work.
By automating repetitive tasks, improving efficiencies, and enabling new ways to collaborate, AI agents promise to disrupt labor at every level.
This transformation isn’t just about cutting costs – it’s about unlocking human potential.
AI agents free people from mundane tasks, allowing them to focus on creative, strategic, and high-value work.
In the short term, I suspect agents will interact with humans to create polished outputs before becoming autonomous.
It will free up time for people to pursue their passions and expand the creator economy.
Employees can now quickly start side projects that weren't possible before. The creator economy, previously confined to music, content, and videos, will now embrace tech.
There are people building million-dollar apps with Cursor, Replit, Claude & Bolt without knowing how to code.
https://x.com/rom1trs/status/1838189197011616005
https://x.com/uxaiagency/status/1856713703116165307
The pie is clearly expanding. Building is being democratized.
Distribution, speed, and a community-first approach are the only advantages.
But success in this space won’t come easy.
Whether you’re targeting verticals, in-house roles, or outsourced services, the playbook remains the same: start small, build trust, and scale with precision.
The next era of innovation is upon us. Whether you're building, investing, or adopting AI agents, the opportunity is huge. The only question is: how will you seize it?
If you liked this essay, consider sharing it on your socials or inviting a friend to read this essay. Feel free to say hi to me on Twitter or on LinkedIn or email.
That’s all for today! See you soon!
Amor Fati Amor.