Skip to content
Governed AI workforce

Digital collaborators that actually work inside your processes.

Not another chat window to open. With Genesis, AI agents receive tasks, use authorized company knowledge, work with email and documents, use tools, collaborate with each other and ask for approval when it matters. Every run stays traceable.

  • Tasks assigned, not prompts
  • Every run tracked and approvable
  • Cloud or inside your own infrastructure

Why AI delivers so little value in most companies today.

It isn't a model problem: it's a problem of how work is organized. Here's what changes when AI stops being a chat and becomes a collaborator.

  • AI is a chat window a few people open, when they remember to.

    Tasks assigned to digital collaborators who work every day, with no need for reminders.

  • Everyone writes different prompts and gets different results.

    A way of working defined once and repeated identically on every case.

  • No one knows what the AI did, with which data, or why.

    A complete history of actions, sources used and decisions made.

  • Information is scattered across email, attachments, documents and internal knowledge.

    Agents that read authorized sources where the data already lives, with no migrations and no new repositories.

  • Confidential data can't leave the company.

    Local or on-premise execution for sensitive tasks, cloud only where it's actually needed.

  • AI costs are unpredictable and hard to justify.

    Usage and cost measured per task, per agent and per process.

The shift in perspective

From a tool you use to work that gets done.

The difference between an AI assistant and a digital collaborator isn't the technology: it's who carries the weight of the work.

AI assistant

  • Someone has to open it and explain the context
  • The result depends on how the request was worded
  • No memory of the business process
  • The work still sits with the person
  • No record of what was actually done

Digital collaborator

  • Receives tasks, not prompts
  • Knows the company's process, rules and documents
  • Works autonomously on repetitive steps
  • Asks for confirmation when a decision is sensitive
  • Leaves a verifiable record of every action

People stay at the centre: they decide, approve and manage exceptions.

How it works, in four steps.

No organizational revolution: a task is assigned, the digital collaborator carries it out, the company checks the result.

  1. 01

    You assign a task

    Describe a recurring job: sorting incoming requests, extracting data from orders, drafting a reply.

  2. 02

    The collaborator understands the context

    It only accesses authorized sources — email, documents, company knowledge and integrated systems — and applies the rules of your process.

  3. 03

    It does the work

    Reads, classifies, extracts, compares, fills in and produces a result that's ready to use, not a generic piece of text.

  4. 04

    You check and approve

    Critical steps go through a person. Everything else is executed and logged, with a history you can always review.

Every step is configurable: you can start with approval on everything and relax it as trust grows.

Three digital collaborators to get started.

Operational profiles already built around the tasks that, in most companies, take up the most time while delivering the least value.

Email

Correspondence management

Watches over shared inboxes — info, orders, support — and turns messages into organized tasks instead of an endless queue.

  • Reads and classifies incoming messages
  • Recognizes urgent items, requests and relevant attachments
  • Drafts replies consistent with your company's tone
  • Assigns each case to the right person
Documents

Document processing

Works on orders, invoices, delivery notes, specifications, contracts and data sheets: extracts the data and returns it in the format your systems expect.

  • Extracts structured fields from PDFs and attachments
  • Compares documents and flags differences
  • Summarizes long contracts and specifications
  • Proposes structured changes to DOCX and XLSX, for approval
Operations

Support for internal processes

Covers the connective tasks between systems and people: checks, controls, reminders and reports that someone currently does by hand.

  • Checks that case files are complete
  • Retrieves information from internal archives
  • Generates periodic reports and summaries
  • Flags exceptions and cases that need attention

Profiles adapt to your process: we start with one and extend it once it's proven to work.

Agents collaborate and delegate tasks to each other.

An agent doesn't have to do everything on its own. Genesis can assign parts of a task to other agents with the required skills, while keeping the relationship between the main task and the delegated ones fully tracked.

  1. 01

    Lead agent

    Receives the task and defines its goal and the steps required.

  2. 02

    Delegates the task

    Assigns parts of the work to the agents with the required skills.

  3. 03

    Administration agent

    Checks documents, data and the conditions set by the process.

  4. 04

    Research agent

    Retrieves information from authorized company knowledge.

  5. 05

    Results

    Every contribution returns to the lead agent, with its source and outcome logged.

  6. 06

    Lead agent

    Aggregates the work, prepares the result and submits it for approval.

  • Every delegation is explicit and logged
  • No agent operates beyond the permissions you've assigned
  • The chain of work can be reconstructed step by step
  • The human checkpoint is always defined in advance

The real work runs through email and documents.

That's where hours of manual work pile up. Digital collaborators work directly on these channels, with no need for the company to change tools.

Email

Each agent can have its own company mailbox: email received → analysis → business context → task → optional delegation → draft → approval → reply.

  • Reads messages and classifies requests
  • Uses authorized knowledge and analyses attachments
  • Creates tasks and brings in other agents when needed
  • Prepares the reply and waits for approval before sending

Documents

An agent reads a document, understands its content, extracts the information and uses it to carry out the task. It can also propose structured changes to DOCX and XLSX files.

  • PDF, DOCX, XLSX, CSV, JSON, TXT and images with OCR
  • Extraction of fields, tables and key information
  • Comparisons, consistency checks and anomaly flagging
  • Original document → proposal → verification → approval → new version

The original file is never modified: every change goes through a new, approved version.

Autonomy with a handbrake always within reach.

Delegating work to AI only makes sense if it stays clear who can do what, with which data, and with what ability to step in.

Per-agent permissions

Each digital collaborator only accesses the sources, inboxes and tools you've assigned to it.

Human approval

You define which steps require confirmation before they're executed or sent externally.

Verifiable history

Every action logs its input, sources consulted, outcome and time of execution.

Process rules

Exceptions and operating limits are defined by the company, not left to the model.

Data boundaries

You decide which information can reach an external service and which stays internal.

Instant interruption

Any agent can be suspended or restricted at any time, without stopping the others.

Control isn't an add-on feature: it's the precondition for using AI on real processes.

Your data stays where you decide.

For many companies the constraint isn't cost, it's confidentiality. That's why execution can take place inside your own infrastructure.

Local execution

Models run on company servers or dedicated hardware, with no content sent externally.

Hybrid mode

Sensitive tasks run locally, cloud capacity is used only where extra processing power is needed.

Control over data destination

You define which tasks must stay within your infrastructure and which can use external services, according to your organization's policies.

Defined perimeter

You know exactly which flows stay internal and which can leave, for every single task.

Designed for companies handling industrial, contractual or personal data.

The infrastructure behind 93labs

Genesis: the infrastructure of the digital workforce.

Genesis is the system that makes digital collaborators possible: it assigns tasks, coordinates agents, controls permissions and measures what gets done.

A single agent isn't enough: you need rules, memory, controlled access to data, delegation between agents and a reviewable history. Genesis is the layer that holds all of this together, deciding for every task which model to use, which tools to activate, and whether to run locally or in the cloud.

Genesis

Orchestration

  • Agents
  • Email
  • Documents
  • Company data
  • Tools
  • Permissions
  • Memory
  • Local AI
  • Cloud AI
Genesis assigns every task to the most suitable resource and returns a tracked result, ready to use in your business processes.

You know what the AI is using. And what it costs.

Every AI call can be attributed to the task that triggered it. Genesis makes it possible to see which models are used, what they cost and where local execution can reduce reliance on the cloud.

Measure your AI workforce
Cost per task
Measures the cost of running a process with AI.
Local vs. cloud
Shows where the work is being executed.
Cost per agent
Understand the usage associated with each digital collaborator.
Human intervention
Identify which workflows still require approval.
Providers and models
See which models are actually carrying out the tasks.
Cloud cost avoided
Assess the impact of local-first strategies, where applicable.

These are the metrics Genesis can measure in your own installation: the values depend on your processes.

Product status

Genesis is already operational. Now we're taking it into the field.

The core of Genesis is up and running: agents, delegation, email, documents, memory, governance, local/cloud execution and cost control. We're consolidating workflows on real use cases, progressively extending integrations, and measuring reliability and results across business processes.

  • Agents and capabilities
  • Traceable executions
  • Delegation between agents
  • Mailboxes and email pipelines
  • Documents and OCR
  • Knowledge and memory
  • Human approvals
  • Scheduled tasks
  • Model routing
  • Local and cloud execution
  • Cost accounting
  • Audit and observability

We're looking for companies with a real process to improve.

We don't want to start from a demo built in a vacuum. We want to start from a task that currently takes up time for the people in your company.

  • Managing a shared inbox
  • Analysing documents
  • Repetitive administrative tasks
  • Researching and gathering information
  • Managing internal knowledge
  • Preparing documents
  • Coordinating tasks across different roles

What's the first task you'd delegate?

Tell us how you work today. Together we'll assess whether there's a digital collaborator that would deliver a real, measurable and sustainable benefit.