Atelier by CrowdOS · Digital Twins powered by CrowdOS

A twin model of your customers, built for your business only

Digital Twins powered by CrowdOS is the first solution from Atelier by CrowdOS: a private cohort crowd built from your customer file, a calibration layer trained on your own campaign outcomes and, where your data supports it, a small model trained only on your customers.

Apply for a pilot→How customer digital twins work

By applicationPilots by application. The privately trained model is in build; pilots start at Lite.

Tiers

What is included at each tier?

What is included at each tier?
	Lite	Standard	Private
Private cohort crowd from your customer file	✓	✓	✓
Private calibration layer trained on your outcomes	✓	✓	✓
Small model trained on your survey data and outcomes	—	✓	✓
Ledger only your organisation can see	✓	✓	✓
Isolation	Shared infrastructure, isolated data	Dedicated tenant and storage	Dedicated tenant and storage
Deployment	CrowdOS cloud	CrowdOS cloud	Your cloud (scoped separately)
Refresh	Quarterly calibration	Quarterly model and calibration	Monthly
Executive workshop and pitch-notes pack	—	✓	✓

In buildThe private calibration layer and the trained model are in build. Standard and Private are scoped now and start when they pass validation; until then a pilot runs at Lite.

Pricing is set per engagement. Talk to us for a proposal.

Engagement

How long does an engagement take?

1 week
Discovery
The decision to test, a data inventory, and a success measure agreed and written down.
1–2 weeks
Data audit and agreement
Column mapping, personal-data policy, a minimum-data check and the data-processing addendum.
2 weeks
Cohort build and baseline
Your cohorts and twins are built, a baseline study is run, and the result goes into your private ledger.
2–3 weeks
Model training and validation
Standard and Private only: the model is trained on your data and checked on a campaign it did not see.
1 week
Activation
The first live study on a real decision, the pitch-notes pack and a working session.
Ongoing
Refresh
Results from each campaign are uploaded and the calibration and model are refreshed on schedule.

Typical durations. The exact plan is agreed in discovery.

Data

What data does each tier need?

What data does each tier need?
Tier	Minimum data
Lite	At least 5,000 customer rows with at least six usable features
Standard	Lite, plus either two past campaigns with outcomes that can be attributed to cohorts, or 1,000 survey responses with demographics
Private	Standard, plus five campaigns or 5,000 survey responses

Below the Lite minimum we will say so, and suggest a market study on the public crowd instead.

How to prepare a customer file→

Ownership

Who owns what is built?

You have exclusive use of your cohorts, your calibration layer and your model. We host them; they never load for another organisation.
On exit, your datasets, cohorts, calibration layer and model are deleted, and we issue a deletion certificate.
Motif Motion keeps the pipeline, the engine and the templates. Nothing learned from your data is used for another client.
Outputs are cohort-level. No prediction about a named individual is delivered.

In buildTo be written into the engagement contract and data-processing addendum, which are being drafted with counsel.

Success measure

How do we know it worked?

Before the work begins we agree the measure, write it down and do not move it: how closely the twins' ranking of offers matches what a held-out campaign actually did, how much better the cohorts they ranked highest converted than the rest, and whether the people using the pitch notes found them useful. If the twins miss at 90 days, the engagement converts to Lite or ends, and your data is deleted.

Questions

What is Digital Twins powered by CrowdOS?

An Atelier engagement that builds a private model of your customers: cohort twins from your customer file, a calibration layer trained on your campaign outcomes and, at Standard and Private, a small model trained only on your data.

How is it different from the self-serve customer twins?

The self-serve mode builds cohort twins from your file on a CrowdOS plan. Digital Twins adds a calibration layer and a model trained on your own outcomes, a dedicated tenant at Standard and Private, a pre-agreed success measure, and our team running the work with you.

Is our data used to train models for anyone else?

No. Nothing learned from your data is used for another client, and on exit everything built from it is deleted. These terms go into the engagement contract and data-processing addendum.

Can we buy it through our agency?

Yes. Digital Twins can be delivered through an agency partner, who can handle data collection and the workshop.

What happens if it doesn't work?

The success measure is agreed before we start. If the twins miss it at 90 days, the engagement converts to Lite or ends, and your data is deleted.

Apply

Apply for a pilot

Tell us the decision you want to test and how your customer data is held today. We will reply within two business days.

What is included at each tier?

How long does an engagement take?

Discovery

Data audit and agreement

Cohort build and baseline

Model training and validation

Activation

Refresh