MS DuplexSolving new problemsShow allShow less
Solving new problems.
Works out new cases from what is known, and checks first.
MS Duplex works out an approach to a new case from the methods and results your team has already confirmed, then checks it before anyone relies on it. Once the approach is verified, it becomes a method for the next case like it.
Plan and decide · one of the fourteen capabilities of MS Duplex


Your team gets
- Faster starts on unfamiliar problems
- New approaches checked before they are trusted
- Every solved problem adds to what the team can do
What solving new problems means for your team.
Solving new problems means your AI has a place to start when a case is unfamiliar. MS Duplex looks at what your team has already confirmed, the methods that worked, the results they produced and the conditions they hold under, and assembles an approach to the new case from them.
A new approach is a proposal until it is checked. MS Duplex checks it before anyone relies on it, and your researchers can see which confirmed methods and results it was built from, each with its source.
Once the approach is verified, it is kept as a method. The next case like it starts from that method, so every problem your team solves adds to what the team can do.
Why research needs AI that can start on the unfamiliar.
Research is the work of meeting cases nobody has solved yet. The routine is only part of the job.
A new design target, an unusual set of inputs, an analysis nobody has set up before: these are the cases where an agent could help most and where it is most likely to go wrong. Starting from a blank page means slow progress, and guessing means an approach nobody can vouch for.
Most new cases are related to something the team has already done. The methods, results and corrections built up over earlier work hold much of the answer, once the AI can draw on them and check what it builds before it is trusted.
New cases are the job
Research keeps meeting cases unlike the ones before, and each one needs an approach.
Experience is the starting point
The methods and results your team has already confirmed are the best material for an approach to a related case.
Trust needs a check
An approach to a new case should be checked before anyone relies on it, however plausible it looks.
Solutions should compound
A problem solved once should make the next case like it a matter of applying a known method.


How solving new problems works in your team’s day.
Your agent meets a case it has not seen. MS Duplex helps it start from what the team already knows.
- 1
A new case arrives
Your agent meets a case that no confirmed method covers directly, with its own goal and conditions.
- 2
An approach is assembled
MS Duplex puts together an approach from the methods and results your team has already confirmed, each with its source.
- 3
The approach is checked
Before anyone relies on it, the approach is checked, and your researchers can see what it is built from.
- 4
It becomes a method
Once verified, the approach is kept as a method, so the next case like it starts from there.
An example, in any field where new cases keep arriving.
A team’s agent runs a standard analysis on incoming data. One day a request arrives that combines two things the team has only ever handled separately.
- 1
The new case
The request needs data from two different sources treated in one analysis. No confirmed method covers both at once.
- 2
The approach
MS Duplex assembles an approach from two confirmed methods, one for each source, together with the conditions each was confirmed under, and shows where each part came from.
- 3
The check
The approach is checked on a case with a known answer before it is used. A researcher reviews it, adjusts one step and approves it.
- 4
The next time
The approach is kept as a method. When a similar request arrives later, the agent applies it directly, and the method is re-verified where it is used.
Solving new problems works with the other capabilities.
A new approach is built from what the rest of the layer has kept, and once verified it adds to it.
Reusable methods
The methods your team has verified are the building blocks of each new approach, and a verified approach joins them as a new method.
Learning cause and effect
What your team has learned about which action led to which outcome shows which parts of earlier work carry over to the new case.
Decisions that know their limits
Where a case is unlike anything seen before, the decision goes to your model or a researcher, with the reason.
Provenance, review and unlearning
Every new approach keeps the sources it was built from, so a researcher can check what it rests on.
What your team gets.
A faster start on unfamiliar problems, with every new approach checked before it is trusted.
Faster starts on unfamiliar problems
A new case starts from the methods and results your team has already confirmed, instead of from a blank page.
Approaches checked before they are trusted
A new approach is checked before anyone relies on it, and your researchers decide whether it is adopted.
Visible building blocks
Every approach shows the confirmed methods and results it was built from, each with its source.
Every solved problem adds up
A verified approach becomes a method, so the next case like it is handled directly.
Experts where they matter
Your researchers spend their review time on the new and the unusual, while familiar cases follow the methods the team trusts.
Know-how that grows
The range of cases your team’s AI can handle grows with every problem the team solves.
Common questions.
Make the next run benefit from the last.
Bring us the research your AI should get better at. We will plan a pilot on your own workflows, with your experts as the judges.

