Artificial intelligence may have access to vast amounts of information, but information alone does not guarantee a useful answer. The quality of an AI-generated response often depends on whether the system has the right context: the background, history, goals, preferences and constraints surrounding a request.
Jason Balser, Senior Director of AI & Data Strategy RELI Group, recently spoke with Angela Marcato, Senior Director of Health, about the role context plays in artificial intelligence. Their conversation explored how context improves AI outputs, supports knowledge retention, increases efficiency, and helps organizations turn AI from a general-purpose technology into a more useful and intentional tool.
Jason Balser: When you think about context, how do you define it?
Angela Marcato: My background is in human-centered design, web content management, and user experience, so I approach context from a human-first perspective.
I think of context as the connective tissue that explains why the information you are entering into a tool matters. It includes the circumstances surrounding the request, the reason you are making it, and the outcome you are trying to achieve.
A prompt may tell an AI system what you want it to do, but context helps it understand why you want it done and what a useful response should look like.
Jason: Why has context become so important in conversations about AI?
Angela: People frequently focus on prompts. They want to know which prompt someone used or how they should phrase a particular request. Prompts are valuable, but I learned through my own use of large language models that the background information surrounding a prompt is often where the real value lies.
As a tool learns more about your goals, preferences, and ways of working, its outputs can become much more relevant. That experience made me think about the opportunities we have to apply the same concept to the work we perform for our customers.
There is a lot of interest in adopting AI, but simply adding AI is not enough. We need to understand how it can help solve a problem, what information it needs, and which context will allow it to produce a meaningful result.
Jason: Is more context always better, or is the goal to provide the right context?
Angela: I think it is the right combination.
I saw this firsthand while using AI to help prepare recurring customer reports. Each week, I entered my team’s updates and asked the tool to consolidate the information. Over time, the tool accumulated a history of the team’s work. It could use that history to connect individual updates and tell a more complete story about the team’s progress.
When I went on vacation, I showed a colleague the prompts and inputs I typically used. The report they produced was technically correct, but it did not contain the same historical perspective or continue the broader narrative.
My colleague knew how to enter the information, but they did not have the ongoing context or know which threads to pull together. That distinction demonstrated why context is more than a set of instructions.
Jason: What does that example tell us about individual and organizational uses of AI?
Angela: It highlights the difference between personal empowerment and team or organizational empowerment.
The reporting process empowered me because the context lived within my ongoing interaction with the tool. However, it did not automatically empower the rest of the team.
One potential solution is to move appropriate information into a shared context, such as an approved shared repository that team members can access. This could allow other people to benefit from the same history and supporting information.
Organizations must be thoughtful about what is shared, however. Context may contain information that should remain private or restricted. The challenge is identifying what people want to share, what they need to share, and what should not be shared.
Jason: How can context help organizations preserve institutional knowledge?
Angela: A significant amount of organizational knowledge exists in people’s heads. When someone leaves a company, changes roles, or moves to another project, that knowledge can leave with them.
New employees or project leaders may find it easier to start over than to search through years of disconnected files and documentation. Unfortunately, starting over can mean losing the reasoning behind earlier decisions, including what was attempted, why a certain approach was selected, and which lessons were learned.
Structured knowledge-capture interviews could help. An organization might identify the information it wants to preserve, prepare questions in advance, record interviews with experienced employees, and securely incorporate the resulting transcripts into an approved knowledge system.
That system would not replace the people who left, but it could provide future employees with a starting point and help preserve important context about their experience and decisions.
Jason: How would access to that history help a program or project leader?
Angela: Program leaders frequently inherit work from other managers or teams. Consultants also enter organizations specifically to help solve problems.
In either situation, knowing what previous teams attempted, why they made certain decisions, and how they arrived at an existing solution would be extremely valuable. Without that information, people may repeat earlier work or pursue an approach that has already been tested.
Preserving context gives new leaders a better foundation. Instead of only seeing the final decision, they can understand the reasoning and history behind it.
Jason: Organizations already manage enormous amounts of information. How should they determine which context is important?
Angela: That is one of the central challenges. Organizations first need to understand what they want to capture and then determine where that information currently exists.
A problem-focused approach can help. Start by asking:
- What are the organization’s most significant challenges?
- Where are teams experiencing inefficiencies?
- Which decisions require information that is difficult to find?
- What knowledge would be most damaging to lose?
- Which recurring tasks could benefit from a clearer history?
AI is a tool, not a solution by itself. The goal should not be to collect information simply because it exists. The goal should be to identify the information that can help people solve problems, make decisions, increase efficiency, or improve continuity.
Jason: How does context improve AI in everyday situations?
Angela: Context allows an AI tool to filter possibilities according to the needs of the person using it.
For example, while traveling with my family, I might ask an AI assistant where we should stop to eat. A generic search could return a long list of nearby restaurants. A context-aware response could account for my location, my family’s preferences, and my children’s dietary restrictions.
Instead of simply returning the closest option, the tool can help identify the option most likely to work for everyone.
That is a simple example, but it illustrates the larger point. Context can help AI narrow a large amount of available information into a response that is relevant to the situation.
Jason: Does that context develop all at once?
Angela: In my experience, it accumulates over time.
When I first began using an AI assistant, my prompts were basic and the experience felt like using a more advanced search engine. As I provided feedback and corrected responses, the results improved.
Over time, the tool developed a better understanding of my goals, preferences, philosophies, and communication style. That did not mean I wanted it to agree with everything I said. I specifically wanted it to challenge my ideas when appropriate.
That kind of relationship requires active feedback. If a response is inaccurate or unhelpful, correcting it is just as important as writing the initial prompt.
Jason: What should people keep in mind as AI systems learn more about them?
Angela: They should be intentional.
AI systems may learn from both the information people deliberately provide and the patterns that emerge through repeated interactions. That can improve the usefulness of the tool, but it also raises important questions about privacy, governance, and appropriate information sharing.
The same considerations become even more important at the organizational or government level. Organizations need to determine what information a system may use, how that information should be protected, and where human oversight remains necessary.
AI can help identify connections that people might not recognize on their own. However, organizations still need clear boundaries around the information being used and the purpose it is intended to serve.
Jason: Do you view AI as a replacement for people?
Angela: No. I view AI as an augmentation for people.
AI can synthesize information quickly, which can help people move faster, increase efficiency, and expand their capacity. The value comes from combining that speed with human experience, judgment, and intention.
The technology does not replace the need to understand the problem. It does not eliminate the need for governance, and it does not remove human responsibility for evaluating the result.
Jason: What final advice would you give to people using AI?
Angela: Be thoughtful and intentional in your engagement with AI.
Provide useful context, but also pay attention to the feedback you give the tool. Correct it when it is wrong. Ask it to challenge your assumptions rather than simply agree with you. Help it understand your goals and the standards you expect it to follow.
The greatest value does not necessarily come from finding one perfect prompt. It develops through an ongoing process of instruction, feedback, evaluation, and refinement.
When used responsibly, context can provide an intentional benefit by helping AI address a known need. It can also provide a serendipitous benefit by helping people uncover connections or questions they might not have identified on their own.
