# Twinorg AI > Twinorg models an organization around its work and lets leaders test changes, such as moving work to AI, before they make them. ## The argument - **Your organization was built for a world where people did all the work.** Departments, layers, sign-offs. Every request climbs the hierarchy, waits for a signature, and comes back down. - **That world is over.** AI now does pieces of almost every job. Companies that redesign around it will out-deliver the ones that do not. - **Adding AI to the old structure only moves the bottleneck.** One step gets faster. The work piles up at the next person who has to check it, and you deliver no more than before. - **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. - **Try the change before you make it.** Move work to an AI agent, remove an approval, reshape a team. See what it does to output, lead time and cost before anyone's job changes. - **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.** Start with one value stream. ## Questions and answers ### 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 as observed, estimated or assumed. ### 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, with the assumptions each result depends on. There is no single score. ### Does Twinorg replace people or make the decision for us? No. Twinorg informs decisions and records them; people make them. A named human position stays accountable for every outcome, including work done by agents. Hours released by AI are reported as capacity, not as savings, until someone decides what to do with them. Twinorg models roles and teams and does not monitor individuals. ### 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? They are an illustrative example, not a customer result. 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. 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. Choosing an industry on the page renames the steps and the unit of work. It does not change the model or these numbers. ## The example, in the words of each industry The scene plays one six-step stream of work. Only the step names and the name of the unit of work change by industry; the model and its numbers are the same. Open the page in one industry's words with `/?industry=`. - **Any industry** (`general`, unit of work: request): Request intake → Approval → Planning → Core work → Review → Billing and handover - **Manufacturing** (`manufacturing`, unit of work: order): Order intake → Credit approval → Scheduling → Production → Quality check → Invoice and dispatch - **Professional services** (`services`, unit of work: job): Client intake → Scope approval → Staffing → Client work → Partner review → Billing - **Software and IT** (`software`, unit of work: ticket): Ticket intake → Change approval → Sprint planning → Development → Code review → Release - **Healthcare** (`healthcare`, unit of work: referral): Referral intake → Prior authorization → Scheduling → Patient care → Clinical review → Billing and coding - **Banking and lending** (`banking`, unit of work: loan): Application intake → Compliance approval → Risk assessment → Underwriting → Final review → Funding - **Insurance** (`insurance`, unit of work: claim): Claim intake → Coverage approval → Assignment → Claim assessment → Quality audit → Payment - **Public sector** (`public`, unit of work: case): Application intake → Eligibility approval → Case assignment → Case work → Supervisor review → Decision notice ## Links - [Home](https://twinorgai.com/): the full story, as a scroll-driven scene with the same text - [Request an organization](https://twinorgai.com/auth/onboard): leave a work email to get started - [Sign in](https://twinorgai.com/auth)