Skip to the questions
Twinorg AI
  1. 1Why
  2. 2The twin
  3. 3Try it
  4. 4Decide
  5. 5Questions
Sign inBuild your twin

Change your organization on a copy first.

Twinorg models how your work flows, so you can try a change and see what it does before anyone's job changes.

Try one now. Give Request intake to an AI agent. Requests delivered per week will:

The work took a day. The waiting took a week.

Departments, layers, sign-offs: a shape built for a world where people did all the work. Companies that redesign around AI will out-deliver it.

We added the AI agent. Delivery went down.

The agent finishes faster than the one person who has to check its work, so the pile moves to approval. Adding AI to the old structure only moves the bottleneck.

So make a copy you can change.

A twin of your organization, laid out the way the work flows: what you deliver, the steps that produce it, who does each one, and where it waits.

Now it is your turn.

The agent is on and the work has jammed. Change the design and watch output, lead time and cost. Your real organization does not change.

Then decide.

Roll out the design that worked. AI does the routine work, your people do the rest, and a person answers for every outcome.

More delivered, sooner, at a lower cost per unit.

The same people, with the work redesigned around them.

Your real organization: unchanged
Delivered / week
97–101
today 105
Lead time, days
8.0–12
today 7.8
Cost / request
$427–446
today $398
Jammed: work arrives faster than it leaves.
Decision record
Question

Should request intake move to an AI agent?

Your call

You did not make one.

First test

The agent alone, nothing else changed: requests delivered fell from 105 to 98 a week, and lead time rose from 7.8 to 10 days.

Changed

Request intake: AI agent, checked by a dedicated reviewer.

Approval: only above a threshold.

Planning and core work: people, assisted.

Billing and handover: automated, exceptions to a person.

People released from paperwork move to core work.

Unchanged

Core work and review stay with people. Nobody leaves.

Result

Delivered a week: 123 to 132 (was 103 to 108).

Lead time, days: 0.8 to 1.0 (was 6.0 to 9.2).

Cost per request: $326 to 350 (was $387 to 407).

Depends on

12 simulated weeks at 130 requests a week. Ranges are the 5th to 95th percentile of the runs.

Accountable

A named person, not a team. That includes the work the agent does.

Reviewed by

Someone who did not propose it.

Do this with your own numbers.

This is what Twinorg leaves you with: the change, what it did, what the result depends on, and who answers for it.

Leave a work email and the Twinorg team sets up an isolated organization for you. You start with one value stream and one decision.

Twinorg in plain terms

What is Twinorg?

Twinorg is software for redesigning an organization around AI. It builds an organization twin: a working model of what the organization delivers, the steps that produce it, who or what does each step, and what it costs. Leaders change the twin, simulate the effect, and decide with evidence before anything changes in the real organization.

What is an organization twin?

An organization twin is a digital twin of an organization, built around its work instead of its org chart. It holds the outcomes the organization produces, the value streams that produce them, the activities in each stream, the people, agents and automations that perform them, the human position accountable for each one, and the cost of it all. Every number is marked by how well it is known.

Why not just add AI to the organization we already have?

AI takes over tasks, not whole jobs. Speeding up one step often moves the queue to the next person who has to review or approve the work, so output barely changes. The gain comes from redesigning the flow of work around what AI can do, and that is a design question an org chart cannot express.

What can we test in a simulation?

Moving an activity to an assisted, agent or automated mode. Removing or narrowing an approval. Reassigning work, merging teams or removing a layer. Adding or removing capacity. Changes in demand or cost. Twinorg reports the effect on lead time, capacity, cost and risk as ranges, and states what each result depends on.

Does Twinorg replace people or make the decision for us?

No. Twinorg shows what each option does and records the decision; your leaders make it. A named person stays accountable for every outcome, including work done by agents. Hours freed by AI show up as capacity you can put to work where it counts.

Who is Twinorg for?

Owners and operations leaders who decide how the organization is shaped, and the people who do the modelling with them: transformation and organization design leads, heads of function, and finance.

Is Twinorg only for one industry?

No. An organization twin is built from your own outcomes, steps and roles, so it is not tied to an industry. The example on this page uses general step names: request intake, approval, planning, core work, review, then billing and handover. A visitor can switch them to the names used in their own industry (manufacturing; professional services; software and IT; healthcare; banking and lending; insurance; public sector). Only the names change; the model underneath is the same.

How do we start?

With one value stream and one decision, not a model of the whole company. Leave a work email and the Twinorg team will set up an isolated organization for you. From there you map the steps of one stream, publish a baseline, and test your first change.

Where do the numbers on this page come from?

The scene plays a small model of one six-step value stream, from a request coming in to the result being delivered, simulated for 12 weeks at 130 requests a week. In the old shape it delivers 103 to 108 requests a week, with a lead time of 6.0 to 9.2 working days and $387 to 407 of labour and tools per request. With an AI agent on request intake and nothing else changed, it delivers 96 to 100 a week: fewer, because one approver now also checks the agent's work. In the redesigned shape, with the same people, it delivers 123 to 132 requests a week, with a lead time of 0.8 to 1.0 working days and $326 to 350 per request. The core work stays with people. Request intake moves to an AI agent whose output a person reviews, approval applies only above a threshold, billing and handover is automated, and the people released from paperwork move to the core work.

Start with one value stream.