Employee Digital Twins
A work assistant built around each employee's role.
Help people find answers, prepare work and complete routine tasks using the knowledge and systems they are authorized to use.
An account manager asks for a meeting brief
Your team
Employee makes a request
“Brief me for tomorrow's renewal call.”
NGB agent
Assistant retrieves permitted context
Only documents and records this employee can already access. Sources are cited.
Connected system
Brief and draft prepared
A brief, the relevant actions and a drafted follow-up — all in a pending state.
Your team · approval point
Employee reviews and approves
Nothing is sent externally or written to a system of record before review.
An illustrative sequence built from the deployed workflow design.
What work it does
Give employees a role-based AI work assistant that understands their approved context and helps execute defined tasks.
- Available to configure
Prepare a meeting brief
Assemble a brief from records the employee is already permitted to see, with sources cited.
- Available to configure
Draft a follow-up for review
Produce the draft; the employee reviews and sends. Nothing external goes out unreviewed.
- Available to configure
Retrieve a policy with its source
Answer from approved documents and show where the answer came from.
- Available to configure
Create an action list
Turn a conversation or document into a list of concrete next actions.
- Pilot only
Update an allowed system field
Write back to a specific, permitted field — with the change visible before it happens.
- Available to configure
Answer a new joiner's role questions
Give someone new to the role a grounded answer on process and policy, with the source attached. Structured training programmes remain Potential.com's territory.
Who it is for
- Buyer
- Function head, COO, CIO
- Budget
- Departmental productivity and transformation
- Trigger
- Employees repeatedly search, summarize, prepare and update across systems
- Best first step
- One role, a small set of repeatable tasks, and clear permissions
Where it stops
“Digital twin” here means a configured assistant grounded in a person's role, permissions, knowledge and work processes. It does not mean a replica of their identity, unrestricted access to their communications, or permission to act as them.
Capability scope
What this can do today.
Each capability carries its availability status and its deployment scope. Planned items are labelled as planned. Nothing here is a promise we have not scoped.
| Capability | Scope | Status |
|---|---|---|
Source-grounded answers Answers with citations, inside the employee's existing permissions. Per document store connector. | Per document store connector. | Available to configure |
Role packs Sales account manager, customer-service supervisor, operations coordinator, HR or program administrator. Each pack starts from a few repeatable tasks, not an open-ended assistant. | Each pack starts from a few repeatable tasks, not an open-ended assistant. | Available to configure |
Draft preparation Briefs, summaries and follow-up drafts for human review. Review before send is enforced, not optional. | Review before send is enforced, not optional. | Available to configure |
Permitted system writeback Updating a specific allowed field in a connected system. Field-level allowlist, per deployment. | Field-level allowlist, per deployment. | Pilot only |
Acting on the employee's behalf externally Sending external messages without review. Not offered. Consequential actions require approval. | Not offered. Consequential actions require approval. | Planned |
What you provide
What it needs from you.
- One role, with a named business sponsor
- A small set of repeatable tasks to start from
- The knowledge sources that role is permitted to use
- Existing permission boundaries, which we inherit rather than widen
- A baseline task, so time saved can be compared like-for-like
How it is controlled
Who decides what.
- The assistant inherits the employee's permissions — it does not widen them.
- Permitted data and permitted actions are visible to the employee.
- External messages and consequential updates wait for human approval.
- Activity is recorded, including whether the output was actually useful.
Demonstration
Configure a role assistant, then ask it a policy question.
What you see
Knowledge, skills, channels and voice are configured, then the twin answers from the uploaded operations handbook with the relevant rule applied.
Completion evidence
A source-grounded answer, with the rule it applied visible.
What a scoped demo adds
A scoped demo uses one of your roles, your handbook and your existing permission boundaries.
This is a recording of the product against a synthetic business. No client data, records or systems appear in it. In a live deployment we always tell the person that they are interacting with AI.
Digital twin: a role assistant answering from the handbook
An executive twin is configured with its knowledge, skills, channels and voice, then answers a policy question from the uploaded operations handbook with the relevant rule applied.
28 seconds · captions are burned into the recording · the business, contacts and records shown are synthetic
A walkthrough of the product, not a recording of a client deployment. Want to see it against your own workflow? Ask for a scoped demo.
Evidence
Voice screening and qualification support for a large hiring round
A voice screening and qualification workflow that gathered structured information from candidates and passed it to human recruiters for follow-up.
Figures in review
Figures from this engagement are withheld until their definitions, reporting window and client permission are confirmed. The qualitative case is published; the numbers follow once they can stand on their own.
Measurement
How we will know it worked.
Agreed before launch, with a baseline. Speed on its own is not a result.
- Task completion
- Completed to a reviewed standard, not merely attempted.
- Quality-reviewed time saved
- Compared against a baseline task, like-for-like.
- Rework
- The honest counterweight to time saved.
- Source accuracy
- Did the citation support the answer?
- Weekly active use
- Seat activation alone is not productivity.
- Employee satisfaction
- Asked of the people actually using it.
Getting started
How a first deployment runs.
- 1
Choose one role
With a named sponsor who owns the outcome.
- 2
Define three to five tasks
Repeatable, describable, and worth measuring.
- 3
Confirm permissions
We map to existing access, and we show the employee what is in scope.
- 4
Set the approval line
Which outputs are drafts, and which may be completed.
- 5
Baseline and review
Compare completion quality before expanding beyond the first team.
Commercial model
Role configuration and integration, then a team or active-user subscription.
A shared usage allowance across the team, rather than per-seat metering surprises.
Start with one team and expand after an adoption and outcome review.
We do not publish a price list. Scope drives cost, and we would rather quote your workflow than a generic package.
Questions
Before you book.
How is this different from the assistant we already have?
Generic assistants answer questions. A role pack completes a defined set of tasks using your systems and your permissions, and reports whether the task was completed to standard. If you cannot name the tasks, it is too early to buy.
Can it see everything about an employee?
No. It works inside the permissions that employee already has, and the permitted sources are visible to them.
Will it send emails as me?
It prepares drafts. External messages and consequential updates require your approval.
How do we know it saved time?
We baseline a specific task first and compare completion quality, not just speed. Seat counts are not evidence.
Next step
Map one workflow with us.
Tell us where the work slows down. We will come back with the agent scope, the systems it needs, and how we would measure the result.
Not sure this is the right offer? The other two are Customer Service AI Agents and Autonomous Revenue Agent.