Confirmed examples, ready for training.
When you do want to train a model, Duplex hands over confirmed examples ready to use.
Forgetting and data control. It runs On an authorised request.
A training set built from what your team confirmed.
Duplex hands over the replies your team confirmed, each as it was last corrected. A set built by hand starts from old conversations and guesses.
First, the replies have to be gathered.
You choose what goes in.
A training set goes out only on an authorised request, in the scope you set, and holds only examples your team confirmed.
The request comes from someone allowed to make it.
Train when you choose, and keep learning in between.
A model trained on the set becomes one more tool Duplex can hand work to. Duplex keeps learning beside it, so the next set is ready whenever you want one.
Your team keeps confirming support replies, and Duplex learns each one.
Manages what it learns over time.
You decide when new learning goes live, and every later change is tracked under your retention rules.
- New learning goes live in one stepApproved learning goes live in one tracked step, all at once or not at all.
- What it learned, exported in your formatEverything Duplex learned, and its history, can be exported in the scope and format you choose.
- Confirmed examples, ready for trainingWhen you do want to train a model, Duplex hands over confirmed examples ready to use.
- A whole batch removed at onceApproved subjects, examples, skills or sources are removed together, and the rest is checked again.
- Your retention rules, on scheduleYour retention rules run on schedule, with a preview when you want one.
- Many corrections applied togetherA set of corrections is applied together, and what depended on them is checked again.
- The progress of every changeA large change shows its progress as it runs, and its receipt when it is done.
See it on your own work.
From your own assistant to every agent your company runs, a pilot puts Duplex on your own work, with the measures agreed before it begins.