Are you leveraging APIs to
The era of agentic AI is here. As these autonomous agents rely on APIs, we are seeing a surge in API usage and a new wave of API-driven partner ecosystems.
This means now's your chance to deliver more value to your customers and grow API revenue-but it also means your APIs may quickly become obsolete.
Here's how to sell API services in the agentic AI era—and how to prepare for what's next.
- Provide client services?
- Build partner ecosystems?
- Power AI agents?
The era of agentic AI is here. As these autonomous agents rely on APIs, we are seeing a surge in API usage and a new wave of API-driven partner ecosystems.
This means now's your chance to deliver more value to your customers and grow API revenue-but it also means your APIs may quickly become obsolete.
Here's how to sell API services in the agentic AI era—and how to prepare for what's next.
1000x API Growth
As mentioned in my recent blog post on agentic AI, client agents act on behalf of a human user. They contain API calls to other services such as "make a payment" or "get invoice status."
(Increasingly these API services are contained inside service agents powered by MCP servers. Anthropic's Model Context Protocol (MCP) specifies the structure for how agents call one another to request services. When agents are built using an agentic framework, such as LangGraph or CrewAI; the framework enables agents to call these MCP servers.)
The more agents we build, the more API services they'll consume—whether directly or through MCP servers. This is why the API services provided by software vendors are even more important in the era of agentic AI.
As Aaron Levie CEO of Box said, "the de facto model of software integrations in AI is one primary AI Agent interacting with the APIs of another system. This is a great model, and we will see 1,000X growth of API usage like this in the future."
Selling APIs to Agents
As we accelerate into the agentic AI era, API service vendors (which all software companies should be) need to craft strategies for API products to be more easily utilized by Large Language Models (LLMs) and agents. Like many other companies, Stripe has implemented an “Agent SDK” which natively supports agentic frameworks.
These SDKs often expose higher-level capabilities more easily used by an agent-for example, a single function like "clients_last_actions()" instead of multiple API calls. They could also provide natural language "context" which is useful for an LLM; for example "this function provides the most recent client actions such as support calls, product inquiries, or payments."
Using Agents to Sell APIs
Companies can also use LLMs to make their APIs easier for developers to use - whether they are building agents or traditional applications. As is becoming increasingly true for support across a variety of industries, software companies who are providing APIs to their partners can create an AI support agent. LLMs power these agents which answer developer questions and provide real-time support. Companies such as VoPay have created developer support agents by pre-populating an instance of ChatGPT with FAQs, code samples, and other documentation.
Of course, it doesn't stop there. Software companies can use LLMs to help generate code snippets or even most of a working application tailored to the developer's needs. Stripe has an AI Dev assistant and Google Cloud has Gemini Code Assist which integrates into IDEs and provides not only chat-based assistance for developers using their APIs but also contextual code completion and even code generation.
Partner Ecosystems: More Revenue, More Client Offerings
If you are selling API-based services you are providing them to developers or really other companies which means you are building your partner application ecosystem. Not only does this generate more revenue, but it also expands the offerings available to your clients. Look at all the applications available on Salesforce's AppExchange or SAP Concur's App Center (and I happened to have worked on both of these).
With the boom in agentic AI and its need for APIs, the potential growth for your partner program is accelerating. As I have written about in the past, you need to consider "How to Become a Platform Company", the "what" and the "why" of becoming a platform company, and how platform and partnership programs grow your company. Further, as I wrote about in "SaaS Partnership"; software companies need to distinguish between different type of partnerships and partnership business models; and what organizations and functions are required to grow partner revenue.
However in order to survive and thrive, you not only must consider how you are selling your APIs to partners; but also what type of APIs or really services you are selling. Otherwise you won't fulfill the potential (and see the direct and indirect revenue growth) of your partner ecosystem.
UI-Driven Agents
Today agents will be easier to build and more reliable if they engage with services that provide a explicit contract about how they behave, an API. However some agents are being designed to watch users work and mimic their actions. This is a very powerful paradigm as it will augment human work by doing many of the repetitive tasks we do daily. Vendors providing "agentic robotic process automation (RPA)" or "LLM-enhanced process agents" include RPA companies such as UiPath and Automation Anywhere.
These agents aren't using APIs because they are acting like a human user and interacting with a software company's traditional application graphical user interfaces (UI). UIs change and their usage isn't always clear (even to us humans), but as agents evolve they will be more capable of navigating a traditional software UI.
Since these type of agents don't use APIs, so this is one way that agents may render your APIs obsolete.
UI-Driven Agents
Today agents will be easier to build and more reliable if they engage with services that provide a explicit contract about how they behave, an API. However some agents are being designed to watch users work and mimic their actions. This is a very powerful paradigm as it will augment human work by doing many of the repetitive tasks we do daily. Vendors providing "agentic robotic process automation (RPA)" or "LLM-enhanced process agents" include RPA companies such as UiPath and Automation Anywhere.
These agents aren't using APIs because they are acting like a human user and interacting with a software company's traditional application graphical user interfaces (UI). UIs change and their usage isn't always clear (even to us humans), but as agents evolve they will be more capable of navigating a traditional software UI.
Since these type of agents don't use APIs, so this is one way that agents may render your APIs obsolete.
From API Calls to Conversations
However the most effective and impactful agents will continue to use APIs. Although as vendors' services start becoming more agentic; traditional JSON formatted APIs may give way to more conversational or agent-native interfaces.
Also, as mentioned in my earlier blog, the revolution will take another leap when Agent-to-Agent communication becomes more prevalent. This will fuel a further shift towards conversational interfaces.
Google's new Agent2Agent protocol prescribes agent-to-agent communication (A2A) to use JSON-formatted messages. These messages can trigger underlying services that are either traditional JSON-based APIs or natural language-based interfaces.
As the providers and the consumers of APIs become increasingly agentic, these powerful programs will increasingly be using their natural language interfaces for their interactions. This is especially true with A2A communication which enables more sophisticated and powerful capabilities.
=> Will agents and agent-to-agent communication make today's application APIs obsolete?
=> How are you evolving your API services for the agentic AI era?
(Views are my own and do not represent those of any current or former employer.)
