Context and Memory

Components four and five are the knowledge components. Context is what your digital worker knows right now, inside this piece of work. Memory is what it carries forward to the next one. Same AI, different context, and the output changes.

What you will be able to do

• Define context and memory in one sentence each and say how the two differ
• Turn a vague request into one that carries the information the decision depends on
• Explain why a capable model still depends on what you give it

Component 4: Context

Context is what the AI knows right now.

What counts as context

• Project details
• Goals
• Policies
• Customer information

None of that is AI terminology. It is the same read-in a new colleague needs before their first real assignment: what the work is, what it is for, what the rules are, and who it is for.

The same request, twice

Watch what changes between these two versions.

First: "Build a workforce strategy."

Then: "Build a workforce strategy using these goals, budget assumptions, labor market conditions, and stakeholder priorities."

Same AI. Different context. Huge difference.

The second version is not more cleverly worded. It is loaded with the facts the strategy depends on. The first version asks a capable worker to guess at your goals, your budget, your labor market and your stakeholders, and then judges the guess. The second version asks the same worker to do the actual job.

Context is the fuel

Poor context plus great AI equals poor results.

Context determines decision quality. Poor context can undermine results even when the model is capable.

So there is a question worth asking before you hand any serious work to a digital worker. What information do you need before making an important decision? Feed the AI that. If you would not expect a person to decide without it, do not expect the model to.

Component 5: Memory

Memory is what the AI remembers later.

What memory holds

• Preferences
• Decisions
• Lessons learned
• Historical information

Memory carries useful knowledge forward across sessions. Without durable memory, future sessions may need key preferences and decisions supplied again.

That is the practical cost of missing memory. Nothing breaks loudly. You simply re-brief, every time, on the things your organization settled months ago, and the quality of each session depends on how much of that history you remember to repeat.

Context versus memory, in one example

Component	What it holds	In a budget conversation
Context	What the worker knows right now	The current project's budget
Memory	What the worker carries forward	How leadership usually evaluates budgets

The number belongs to this project. The pattern belongs to the organization. A worker with only the first can produce a defensible budget that lands badly in the room. A worker with both can produce one that reads the way your leadership expects budgets to read.

Context and memory depend on each other. Context without memory means starting from zero every session. Memory without context means knowing your habits and nothing about today's work.

Key takeaways

• Context is what the AI knows right now: project details, goals, policies and customer information
• Poor context plus great AI equals poor results, and context determines decision quality, so ask what information you would need before making the decision yourself and feed the AI that
• Memory is what it remembers later, and it carries preferences, decisions, lessons learned and historical information forward across sessions

Next step

You now have all five components: brain, skills, tools, context and memory. The next module puts them together and answers the word everyone uses and few people define. Continue with So What Is an Agent?.