Skip to content

The Voice Operating Layer

From document
to conversation.
And back.

Your knowledge has more to say. The Voice Operating Layer makes it accessible through conversation in everyday work – and opens a path for your team’s experience to flow back.

What makes the answer possible

Less searching. Move forward faster.

Ask. Understand.
Clarify the next step.

The work situation with VOP

Your team asks in its own words. Information backed by evidence and open questions become visible together. Existing knowledge helps directly in the work situation.

01 · The foundation

One question.
Four clear roles.

Four building blocks help Frank turn existing knowledge into a clear answer.

Frank at a machine on the factory floor.

Illustrative example · Shift handover

“What is still open since yesterday?”

KSE

Ground the answer.

Makes your sources accessible and provides relevant evidence with a traceable knowledge version.

TalkTo

Set the purpose.

Connects the task, knowledge and conversation. This is where the people responsible review new submissions and approve knowledge in their domain.

Voice

Lead the conversation.

Listens, explains and asks relevant follow-up questions – adapted to the task and the user.

Frontend

Show the status.

Shows what has been understood, which information is missing and what stage the result has reached.

02 · Voice Design

A voicethat achieves its goal.

Which information is essential to the task? What has the user already said? Which exception changes the next step? With your domain experts, we develop conversation rules for relevant follow-up questions, clear explanations and well-defined handovers.

Clarify
Ask
Explain
Guide
Summarise
Hand over

03 · The interface

The interface fits
the way you work.

Type. Speak. Work together. In your existing system.

Illustrative concept views

Illustrative concept view · Content and functions are agreed during the pilot.

Frank now knows the status. But at work he discovers something that is not yet in any source.

04 · The knowledge cycle

An observation
can become new knowledge.

Frank describes a new observation and explicitly confirms his submission. Voice can pass it on for review.

  1. Sources in KSE

    KSE ingests sources, makes knowledge accessible and preserves the evidence.

  2. Evidence & versions

    KSE provides evidence and versioned knowledge.

  3. Conversation

    Voice speaks, listens and guides the conversation through the interface provided by Frontend.

  4. Confirmed submissions

    Voice passes them on — a “correction” or “new knowledge” remains a proposal.

  5. Human review

    TalkTo handles new statements, contradictions, steward follow-up questions and approvals.

  6. … and back

    Approved knowledge flows back into KSE as a new knowledge version.

Stage 6 returns to stage 1

This is how the knowledge cycle is designed. We connect KSE with TalkTo during the project.

How a submission is reviewedFrom a confirmed proposal to a traceable knowledge version

Fictional example · Frank's submission

“The spare part has been confirmed, but this is not yet documented in the handover.”

At this point, that is Frank's statement. The person responsible checks whether it is supported by evidence and applies to the equipment concerned.

  1. 1

    Confirm the submission

    Frank explicitly confirms what he wants to submit for review. The correction includes the original question and its evidence context. Voice passes on the confirmed submission.

  2. 2

    Check the statement and evidence

    The person responsible for the knowledge checks: Who confirmed the spare part? Where is this documented? Which equipment does it apply to? Missing information leads to a follow-up question.

  3. 3

    Approval and inclusion in KSE

    The domain expert decides on approval in TalkTo. Inclusion in KSE must then be confirmed. Only then is the submission part of the confirmed knowledge version.

Possible outcomes

Follow-up question
More information or evidence needs to be added.
Rejected
The submission does not change the published knowledge.
Approved
The domain expert has approved it; inclusion in KSE is still pending.
KSE update confirmed
The submission is included in the most recently confirmed KSE knowledge version.

Frank can ask for the “Status of my submission”. We configure the review process and the connection between TalkTo and KSE during the project. Knowledge approval, conversation profile and interface design remain separate decisions.

05 · Five checkpoints

Control
you can see.

You can see what an answer is based on and who decides on knowledge and access.

A woman looks directly into your eyes.
You decide.Knowledge. Access. Accountability.
  1. Approval

    People approve knowledge, not the AI.

  2. Source

    Answers rely on your evidence.

  3. Version

    Every knowledge version is traceable.

  4. Permission

    Access is time-limited and revocable.

  5. Handover

    Limits and responsibilities are defined during the project.

New at Overmind

Voice keeps developing.

All developments
New developmentVoice

One Voice.
A sense of where it is used.

Website, app or embedded system: Voice recognises its digital environment. Three illustrative scenarios show how this context could be used in a conversation.

See the example
OutlookKnowledge extraction

What if your electrical schematic could answer questions?

Outlook: Components and connections could become structured data—a basis for questions in conversation. We explore the idea with a simplified example.

Explore the idea

06 · Working together on the project

How we set up VOP
for your team.

During the project, we agree which sources apply, how VOP speaks and where your team accesses it.

  1. Start with a clear task. Agree which task VOP should help the team clarify.

    01 · Use case

    Start with one clear task.

    Together, we define the use case: what should VOP help your team clarify?

1 / 6
1 Define the task and knowledge We define the scope of the task and choose suitable sources. You appoint the person who reviews new submissions for accuracy in your domain. 2 Try the conversation We agree the conversation's goal, language and voice. Your team tests answers and follow-up questions for the chosen use case. 3 Design the access We design an interface suited to your workflow. You decide who receives access and explicitly authorise it.

Illustrative scenario

Setup and approvals in detail
  1. Company and knowledge

    Set up the company, appoint a steward and add sources. Once processed, the knowledge is available for retrieval.

  2. TalkTo use case

    Define the name, solution and conversation goal. The use case identifies the topic and knowledge area.

  3. Configure the voice

    Configure the voice, language and conversation profile. New conversations use the active profile version.

  4. Test and start

    Test live, approve the voice sample and deliberately start the service. It is then active internally.

  5. Design the interface

    Create the design and assign it as a fixed version. Its name and version number are visible.

  6. Publish access

    Approve access and share a link or widget. Guests use a time-limited access code to speak.

Three separate approvals

Active internally

A person approves the voice sample and deliberately starts the service.

Design assigned

The design is adopted as a fixed version.

Access published

Guest access is explicitly published.

A test does not open guest access. A new design does not replace the deliberate start or the publication of access.

Find your starting point, talk to us.

What partners say

Perspectives on the
Voice Operating Layer.

“Overmind's Voice Operating Layer is a great technology that helps companies in South Africa take the fast route to automation and advanced voice AI solutions.”

Greg HegerPrincipal Overmind South Africa

“Overmind.one supports Europe's companies in becoming the most innovative and sought-after in the world. With care, critical reflection, creativity, courage and commitment. The Voice Operating Layer is closely linked to my understanding of innovation.”

Hans LercherProfessor, entrepreneur

“Our VOP gives companies the tool they urgently need to strengthen the flow of knowledge, and so to make the people in companies more confident and more competent. Not replacing people — making them better. That is our credo.”

Johannes WeissCOO at Overmind

The portraits show the people quoted. Quotations translated from German.

From the example to your company

Which question
keeps going unanswered in your team?

Frank has a clearer picture: He can pass on confirmed information, identify open questions and submit a new observation for review. Which task could this help your team with?

Frequently asked questions

What makes Overmind different – isn’t a voicebot enough?

We focus on shaping conversations around your domain. Together with your experts, we clarify which details matter, when a follow-up question helps and what needs to be available at the end. With Social Coding, we adapt the conversation, knowledge and interface to this workflow. What matters is whether it helps your team handle a specific task better.

We use Microsoft Copilot. When does Overmind make sense?

Microsoft already offers tools for voice agents, enterprise knowledge and system actions. Overmind makes sense when you want to develop the domain-specific design and rollout for your business together with us: relevant questions, your terminology, typical exceptions and a verifiable result. We clarify whether and how your Microsoft systems can be integrated for the chosen use case.

What happens when sources are missing, outdated or contradictory?

These gaps are explicitly part of testing. Together, we define which sources apply and when a follow-up question or human clarification is required. Evidence and the knowledge version make the basis of an answer verifiable. An open question must not be presented as confirmed information.

Does VOP automatically learn from every conversation?

No. A submission confirmed by the user initially remains a proposal. The domain expert responsible checks evidence, validity and possible contradictions and decides on approval. It only becomes part of the new knowledge version once its inclusion has been confirmed. We set up the review process and the connection between TalkTo and KSE for your use case.

Does VOP complete tasks – or only provide answers?

That depends on the agreed use case and the available integrations. We define whether VOP explains, captures information, prepares a draft or executes a system action. During the pilot, we explicitly distinguish between simulation, draft and actual submission. An executed action must have a verifiable result status.

Do we need to organise all our knowledge first?

We start with a clearly scoped task, selected sources and a person responsible for the domain. This reveals which content is missing or needs clarification. We develop the conversation and interface for this use case. The effort depends mainly on source quality, domain rules, desired integrations and the review and approval process required.

What happens to confidential data and access rights?

Before use, we clarify which content may be processed, which services are involved and who should have access to knowledge and results. This also includes storage, deletion and the handling of sensitive information. The specific solution must fit your domain and organisational requirements; we review these for the agreed use case.

How do we know whether a pilot really helps?

We agree a specific task and verifiable success criteria in advance. For example: Are reports more complete? Are unnecessary follow-up questions avoided? Can corrections be incorporated correctly? Does the handover reach the target system? We test the agreed cases, including errors and exceptions, and record what works and where limitations remain.

Solution concept · Availability and scope are agreed for each project.