Experience is more than retrieval.
Own what your AI learns.
Retrieval finds text that looks similar. Experience knows what worked, where it holds and what has changed. MS Duplex turns your team’s experience and knowledge into an asset you own, not one locked inside an AI model.
Where retrieval stops.
Retrieval is good at finding documents. It was never built to learn from them.
Similar is not right
It returns passages that look like the question, whether or not they apply to the case at hand.
Old and new come back together
When a fact changes, both versions return and the model has to guess. In our controlled tests, standard retrieval gave stale answers 57% of the time. MS Duplex: 0%.
It never learns
It finds more as documents are added, but no result and no expert correction ever makes it better.
What experience adds.
Experience is knowledge tested by results. MS Duplex keeps it in a form your agents act on, not just look up.
What worked, with its evidence
Confirmed outcomes, methods and decisions, each pointing to the result or expert judgment behind it.
Where it holds
Each lesson keeps the conditions it was confirmed under and applies only where they match.
What has changed
The newest confirmed value applies, and the history of what changed is kept.
What to do next
Experience changes what your agents check, try and decide, not only what they can find.
| Memory files | Retrieval (RAG) | MS Duplex | |
|---|---|---|---|
| What it keeps | Memory files: Notes someone wrote by hand | Retrieval (RAG): Passages that look relevant | MS Duplex: Confirmed outcomes, methods and decisions, with their evidence |
| How it learns | Memory files: When someone edits the file | Retrieval (RAG): When documents are added | MS Duplex: From every confirmed result and correction, as you work |
| When a fact changes | Memory files: The old line stays until someone deletes it | Retrieval (RAG): Old and new versions both come back | MS Duplex: The newest confirmed value applies, with the history kept |
| Deciding what applies | Memory files: The model rereads it all and judges | Retrieval (RAG): Similar text, whether it applies or not | MS Duplex: Checks whether the conditions match, then applies, adapts or flags |
| Improving the agent | Memory files: Only as good as the last edit | Retrieval (RAG): Finds more, learns nothing | MS Duplex: Changes what your agents check, try and decide next |
- stale answers after facts changed: MS Duplex beside standard retrieval, in our controlled tests
- 0% vs 57%
Your asset, not the model’s.
Knowledge trained into a model’s weights belongs to the model: hard to see, hard to change and left behind when you switch. Experience in MS Duplex belongs to your organization.
Owned by you
Governed by your rules and kept by your organization, whichever model or vendor you use.
Carried across models
Change the model, the agent or the vendor, and your experience comes along.
Exact and editable
Inspect, correct or remove any lesson on its own, and nothing else changes.
Value that compounds
Every confirmed result adds to it, so the asset grows with every week of work.
Start owning what your AI learns.
A pilot of MS Duplex runs on your own recurring work, beside your current setup, and your experts judge the results.

