Weekly API Evangelist Governance (Guidance)
Context is one of those magical words out there right now to describe where businesses should be focusing when it comes to making the rubber meet the road with artificial intelligence if you want get the results you desire. It’s a nice word. I like it. But it is vague enough and I see it thrown around a lot. This tends to make me want to dive in and do some Thesaurus exploration and some writing to get myself to a better place when it comes to using the word in my own storytelling.

Regular Dictionary Meaning
Let’s start with the dictionary definition for what is context, to lay the foundation for what it means collectively, which may or may not be guiding the way people are wielding the word. For me, the dictionary definition is always the place to start, no matter where any trends would like us to go.
The circumstances that form the setting for an event, statement, or idea, and in terms of which it can be fully understood and assessed.
The parts of something written or spoken that immediately precede and follow a word or passage and clarify its meaning.
These definitions seem very nice and neat, but I think the slipperiness of all of this will be in the words like idea and meaning, but will end up being the shift in context (pun intended) when you move from human to a large language model—where we tend to take a lot of context for granted.

Artificial Intelligence Meaning
Here is what context means in the context of artificial intelligence—if you ask artificial intelligence.
“In artificial intelligence (AI) agents, context is the relevant, surrounding information that provides meaning and allows the agent to understand and respond accurately to a task or query, similar to human working memory or background knowledge. It includes background data, user profiles, project memory, and task-specific details, enabling the agent to connect a request to the "bigger picture," avoid errors (like hallucinations), and perform complex, multi-step tasks effectively.”
That is a lot. Much less simpler than the regular dictionary definition. There is a lot packed in there that I don’t think us people in charge of technology always have a handle on.

Where The Context Work Is Right Now
I am interested in moving from a high level definition of what context is to more of the details. The things I can put into a spec or configuration. I want specific APIs that can be used to define the context needed. So what are the areas people are investing in when it comes to context right now to support their agentic work.
Background Information - The "what, where, and why" behind each request we make to in a digital world.
User Data - Information like user profiles, purchase history, or calendar details about each moment.
Task-Specific Details - The project requirements, code structure, or previous steps in a complex problem.
Tools and Capabilities - Information about what functions and tools the AI agent has access as part of work.
Memory - A system of short-term scratchpads and long-term storage to retain past information and decisions.
I am very interested in the background information portion of this. I think the user data, task, tooling, and memory is likely getting a lot of attention from engineering efforts. The what, where, and why I know from experience does not get a lot of attention in our busy days, and I don’t think AI will get any more attention than the API realm has received over the last decade.

What Context Brings to the Table
I am looking for the reasons why we are doing this work, and understand how it informs automation. I am also interested in what context is specially needed for agentic automation versus other types of automation. After reading multiple articles on the subject, these four things stand out as what people are talking about.
Improves Accuracy - Context helps the agent interpret data and make decisions relevant to the specific situation, leading to more reliable outputs.
Minimize Hallucinations - Without context, AI agents can generate irrelevant or nonsensical answers because they lack the necessary background information or understanding of the user's intent.
Enables Complex Tasks - For multi-step problems or coding projects, context allows the agent to use previous information as "memory" to inform future actions.
Guides Action - By providing the right information at the right time, context ensures the agent can select the correct tools and actions to achieve the desired outcome.
It is all about minimizing the work needed to obtain the desired outcome of agent interactions, and minimize the downside of using LLMs to accomplish what we need. Essentially we are trying to minimize the risk of using AI to automate, which may lead us to better understand the risks we are taking by focusing so much attention on this portion of our automation toolbox.

The Human Aspects of Context
The part of this that worries me the most is that we are so heavily focused on the context that agents need, and not always adequately thinking about the difference between human user context and the agent bot context. These are the areas I am thinking about when it comes to how we are proposing to automate our lives, and what context means in this wider human space.
Linguistic Context - This is about the meta language around the language. Words are rarely meaningful in a vacuum; they take on meaning from the words, sentences, and conversations around them.
Situational Context - This is related to the context of the user, i.e. their physical environment, user current activity or device state.
Temporal Context - This involves understanding time and how it applies to user intent and expectations.
Cultural and Social Context - Culture and social norms have a strong influence on language and behavior. The same phrase or gesture can mean different things in different communities.
Emotional and Personal Context - This layer deals with user-specific signals such as emotions, preferences, past behavior, and psychological states.
Historical Context - Leveraging past interactions to make sense of the present. If you’ve previously told your AI assistant you prefer Italian restaurants, it should prioritize those when you say, “Find me a place to eat.”
Are these things always transferable from the human realm to the online realm? It feels like we aren’t always having a fair conversation when it comes to the context humans possess and enjoy and the context that platforms and agents will possess and need—which is where the challenges with AI are introduced.

The Context We Need to Understand
Whenever a word or concept is put up front and center in the debate, or more precisely, the marketing and sales machine surrounding technology I am left skeptical the word is being wielded in any meaningful way. It is hype and hustle. Cloud was this. Artificial Intelligence is this. The context needed to automate our worlds will be no different. Context is complicated and relative. It is something that is very human. Which means it is messy. I am not convinced the people who are having the discussion around context are the ones who need to be having it—me included.
It comes down to whether we are talking about the context that matters end-users and the agents that represent them, or the context that matters to those who are designing, developing, and maintaining the agents, as well as the platforms who are feeding this round of automation. Those of us in charge of this discussion do not possess the standpoint required to understand the nuance present in what is needed for personal, professional, commercial, or industrial automation. But, like with the cloud and artificial intelligence, I am convinced that this is by design. Which is why I am leaning into these important conversations surrounding artificial intelligence and agentic automation, not because I think they are the answer, but they need someone reminding us of what is being ignored, lost, and ground up as part of this approach to automation.
“If the context is lost and merely bits and pieces remain from a scattered existence, only the connection of anchor points may reinstate a distorted mental balance in an upset life story. ― Erik Pevernagie