Working with AI

How AI can retrieve your past trade-offs without deciding for you

Short answer

Retrieving a past trade-off and deciding are two different tasks. AI can hand back yesterday’s options, reasons and constraints, with their date and provenance. The judgment stays yours: Nawame provides context, not the decision.

Several copper lines converge at a fork; past points are lit, but the decision ring at the center stays empty and unlit.

You ask an AI to reread your past decisions, and a worry shows up quickly: what if it starts deciding for me? It is a fair worry. The answer fits in one sentence: retrieving a trade-off and making the call are two different tasks, and only the first one can safely be handed to an AI.

Why the fear of delegating is legitimate

An AI that reads your history carries a quiet authority. It quotes the choice you made in March, presents it neatly, and it becomes tempting to follow it because it is there, written down, consistent. But an old decision was made with the information, the constraints and the mood of that moment. Reusing it as is can mean repeating a mistake with great confidence.

The opposite risk exists too. An AI that knows nothing about your past offers generic options, some of which were already ruled out for good reasons. You waste time explaining again, or you walk the same road without realizing it.

The goal is an AI that knows your history without turning it into a rule.

What a trade-off is and what to keep from it

A trade-off is a choice between several options, made under constraints, at a specific moment. The outcome (“I went with option B”) is the least useful part to retrieve. What really helps, months later:

  • the options that were on the table, including the ones that were dropped;
  • the reason for the choice, and the reason for each thing given up;
  • the constraints of the time: budget, deadline, people available;
  • what happened afterwards, when it is known;
  • the questions that stayed open.

Without the reasons, a trade-off is just a verdict. With them, you can judge whether it still holds, because you can see what it rested on. A format for recording those reasons is offered in why preserve the reasoning behind a decision.

An example: revisiting a choice six months later

Imagine Claire, who runs a small communications agency. In March she picks an invoicing tool. She rules out a more complete solution because its data export is limited and she does not want to be locked in. She also notes that the choice holds for one year, long enough to see how the team adapts. In September, a colleague suggests exactly the solution she ruled out.

Claire asks her AI: “Did we already look at this tool?” A useful answer looks like this: a precedent found in March, set aside because of the export, with a date, a caveat about how long the choice was meant to last, and a question still open. An answer that decides for her would look like this: “You refused this tool, no need to revisit it.”

The first answer gives Claire her own reasons back. She can check whether the export has changed, whether the team has grown, whether the constraint still exists. The second closes the question on her behalf. In both cases the AI retrieved the same thing. How it is presented determines who decides.

How Nawame separates context from judgment

Nawame is context memory for humans and AI. It keeps decisions, reasons, changes and outcomes, then helps the AI tools you authorize find the right context when it matters. It provides context, and the judgment stays human.

A few public principles guide this separation, without needing to go into the mechanics:

  • A retrieval carries a status. Nawame speaks of a “precedent found”, a “possible connection” or “no convincing precedent”. These wordings signal how careful to be, so that a vague resemblance does not pass for a certainty. The trap is covered in when semantic search finds a precedent that never existed.
  • Found does not mean true. A memory comes with its provenance and its date, so you can see where it comes from and whether it may have aged. The questions to ask before relying on one are in why found does not mean true.
  • A replaced decision stays in the history. When you change your mind, the old position is not erased: it is marked as replaced. That is the subject of what should happen when a decision changes.
  • Your personal memory stays yours. It is private, encrypted on Nawame’s infrastructure, hosted in Europe, and never automatically becomes team memory. Details are on the security page.

The split of roles is simple. The human brings experience. The AI brings its reasoning power. Nawame brings continuity.

When a document or a simple prompt is enough

Let us be frank: if you make few decisions, or work with a single assistant, its native memory may be enough. A well-kept decision log, reread from time to time, does the job for anyone with the discipline to write it. And pasting the three useful paragraphs into a prompt is often the most direct solution.

Nawame becomes interesting when the story stretches over months, runs through several tools, and you no longer remember what you had ruled out. It does not try to remember everything either: not everything deserves to become a lasting memory, which the article on why a good memory knows when not to write explores.

Current limits

This separation between context and judgment helps, but it guarantees nothing. Here is what to keep in mind:

  • Finding the right precedent is still hard. A useful memory can be missing, or a connection can be irrelevant. This is one of the problems the Lab studies and measures.
  • What was not kept does not exist. A decision made out loud, never told to an AI or written down, stays invisible.
  • A poorly worded reason stays poorly worded. Nawame keeps what it is given, it does not reconstruct the real reason behind a choice.
  • An AI can over-interpret context that is perfectly accurate. The framework reduces this risk without removing it. Rereading dates and reasons before relying on a precedent is your part.

What this changes for you

When you make a decision that matters, keep the reason and what you ruled out, two lines are enough. When an AI brings back a precedent, read the date and the motive before believing it. And phrase your questions as “what have we already looked at, and why?” rather than “what should I choose?”. You keep the decision, the AI saves you the remembering. If you keep reopening the same question without anything new, we discuss it in detecting decision cycles without psychologizing.

If this fits the way you work, the who it’s for page describes where Nawame helps most. You can also start free and see how it goes on a real project.

Your next prompt should not have to contain your whole story.

Start building a memory your AI tools can find again when it matters.

Start free