Decide what happens to every role and every person when AI arrives.

Role: the position as it exists in your company, with its own name and people. For example, “Contact center agent”, in Customer service, with 4 people. It comes from the CSV.

ISCO-08 occupation: the role's international label, one per role. For example, 4222 “Contact centre information clerks”. It's not a task: it tells the AI what kind of work it is, so it suggests fitting tasks, and lets you compare with other companies and countries. It's shown and changed in the role header on the Map.

Tasks: what the role does each week, several per role, each with hours and exposure. They determine whether the role needs a plan.

HR or an admin loads the structure (areas, roles and people) by importing a CSV under “Import”.

HR or an admin registers AI and automation projects under “Projects”.

The system proposes each role's tasks from the ISCO-08 occupation catalog, and the area lead reviews, corrects, adds or deletes them.

Under “Import”, download the sample CSV, replace the rows with your company's (one row per person) and upload it. The columns are: area, puesto_codigo, puesto, isco, persona_id, pais, antiguedad_anios and email (the last two are optional; with the email, the person gets their plan and answers it in Transita).

Commas or semicolons both work. If a single row has an error, nothing is imported and you're told which line to fix. If a person already exists, their role, country and tenure are updated.

Each new role gets suggested tasks based on its ISCO-08 code, waiting for validation.

No. Each person is matched by employee ID. New people are added; existing ones are updated (role, country, tenure) and keep their plan, signatures and history. People missing from the file are not deleted.

When it finishes, the import reports how many people are new, how many changed and how many were missing. You can record the missing ones as leavers right then, all at once, or review them one by one in “People”. If someone with a plan changed role, it flags them so you can check the destination still fits.

Record it in “People”, with the date and reason (resignation, retirement, end of contract, dismissal, supported exit or other). Nothing is deleted: their plan and the audit log are kept.

From then on they stop counting in the Map, the headcount and the exposure by area. An unfinished plan is kept as a record and cannot continue. A Transita account they had is deactivated.

If their plan was completed (for example a supported exit or finished reskilling), it still counts in the Dashboard and report results, because it is an outcome of the transition. If it was a mistake, reinstate them with one button.

The registration date is saved automatically. The one you enter is the planned start: when the technology will be deployed. It's entered by hand because projects should ideally be registered before they start, to plan the transition early.

A project entered by mistake can be deleted from “Projects”, and its tasks are left without a project. If it already justifies transition plans it can't be deleted, because those plans are evidence: mark it as “Discarded” or “Closed”.

By invitation. An admin (or HR, for leads and employees) invites each person from “Users” with their email and role. They receive an email with a link valid for 7 days.

After that, each person signs in with their Google or Microsoft account, or by requesting a sign-in link by email. There are no passwords to remember. If someone leaves the company, deactivate them and they're locked out immediately.

Each role has an occupation from the ILO's ISCO-08 international classification (436 occupations). With “Suggest tasks with AI”, a language model (Claude, by Anthropic) reads that occupation's official description, the role and area names and the company's AI projects, and proposes 4 to 7 tasks with weekly hours, exposure and reasoning.

No personal data is sent: no names, employee IDs or plans. Suggested tasks are labeled “AI” and stay pending until the area lead validates or corrects them. Without AI configured, the catalog suggestion is available.

They're today's human hours: what a typical person in the role spends on each task per week, working the current way. It's not the total across everyone in the role.

A task's exposure says how much of those hours technology can absorb: 10 h at 0.80 exposure means about 8 h removed and 2 h that stay human (supervising, handling exceptions). The Map shows the total human hours removed for each role.

How long the AI or robot takes to do the task is not determined: for the transition, what matters is how much of the person's time is no longer needed, not the machine's speed.

A role's hours shouldn't exceed the company's working week (40 h by default; administrators change it in “Settings”). If they do, the Map marks it in red and warns you, and the AI suggests tasks within that total.

It calculates the role's exposure: how much of a person's working week in that role technology can absorb. It ranges from 0 (nothing) to 1 (everything).

h is each task's weekly hours and e its exposure (0 to 1). Σ means "add up all the tasks". The top adds the hours technology can absorb (hours × exposure for each task); the bottom adds all of the role's hours. The result is a weighted average: a 16-hour task weighs more than a 2-hour one.

Example, a contact-center agent with 40 h: answering frequent queries 16 h × 0.80 = 12.8 h; logging complaints 10 h × 0.75 = 7.5 h; solving complex cases 8 h × 0.25 = 2 h; in-person service 6 h × 0.10 = 0.6 h. That adds up to 22.9 h technology can absorb, out of 40 h: E = 22.9 ÷ 40 = 0.57.

From 0.20 the role changes and its people need a plan; from 0.50 it changes so much that it is worth assessing whether positions are freed (both thresholds can be changed in "Settings"). Which plan each person gets comes from the decision in "Freed hours". In the example, 0.57: each agent frees 22.9 h per week, and the company decides whether to reduce positions, share out those hours, or do a mix.

Each task's exposure is proposed by the AI or the ISCO-08 catalog, but the area lead validates it: the equation is only used once all of the role's tasks are validated. It does not say how many people to replace or how fast the machine works: it says how much the job changes.

They're starting points designed for a 40 h week. Each company can change them in “Settings”; before saving, Transita shows how many roles change level and which numbers are recalculated.

Automatable task, from 0.70: technology absorbs at least 7 out of 10 hours. What's left is mostly supervising and handling exceptions, so the task stops being the core of the person's work.

Augmentable task, from 0.40: the tool does a significant part, but the person stays in charge. Below 0.40 the task changes little.

Role in adaptation, from 0.20: a fifth of the week, about 8 h of 40, is almost a working day that changes. From there it's worth planning training and where those hours go. Below it, the change is absorbed without a plan.

Role in transition, from 0.50: half the week, about 20 h of 40. The role as it was no longer fills the workday, so it's worth assessing whether to reduce positions. Still, who leaves the role is always the company's decision in “Freed hours”.

It follows the logic of AI exposure studies, which separate low, medium and high exposure, but the values are a Transita design choice, not a standard. If your sector justifies it, adjust them and record why.

Freed hours per person are Σ(hours × exposure): what technology can absorb from their week. In the “Freed hours” tab, the company decides for each role: first, whether those hours can be pooled (whole tasks disappear) or are fragmented (small pieces in each person's week); then, whether it reduces positions, keeps everyone, or does a mix.

When positions are reduced, those who stay absorb the non-automated work of those positions, so they have fewer free hours. Those hours, and everyone's if everyone is kept, go to new tasks, training or a shorter workweek. Whatever is not assigned stays "without destination". With fragmented hours, positions cannot be freed.

It is not an automatic conversion: savings arrive once the tool works and the team has learned to use it, AI also needs supervision time, and the hours are usually spread across the week. That's why it is a company decision, consulted with the people.

Shortening the workweek while keeping pay means sharing with the person the productivity technology brings. Shortening it with lower pay changes the employment terms and generally requires the person's agreement.

The team's unassigned hours are the team check: if they add up to a full workweek or more, the decision is incomplete and no plans are generated until they are assigned or positions are freed. The certification standard (criterion 2.3) requires every changing role to have that decision complete.

They stay in their area, but the job changes. For example, a contact center agent becomes an AI customer experience specialist after 100 hours of courses and 8 weeks of supervised practice.

If the role has no defined destination, the plan proposes the same job “augmented with AI”. If the person moves to a different job, the path is redeployment.

It's an estimate based on the company's hourly training cost (USD 20 by default):

Reskilling: course hours × hourly cost + USD 600. Redeployment: 40 onboarding hours × hourly cost + USD 400. Supported exit: USD 3,500 outplacement + 60 training hours × hourly cost + USD 1,800 health coverage.

Severance is not included.

Because the plan is about their job: no plan is approved unless the person knows it and accepts it.

The person always answers, in Transita. Whoever manages the plan taps “Request a plan review” and the person gets an email with a link that signs them straight into “My plan”, to read it and accept it or ask for changes with a comment. The date is recorded. The conversation can happen in person, but the answer is given by the person in the system.

For that the person needs a user: it is created automatically when their email is added, in the import ("email" column) or under “People”. If they ask for changes, the plan goes back to draft with their comment, to be revised and sent again.

A vacancy is human work that will need people: each area lead enters, under “Vacancies”, which position does the tasks (an existing or a new one) and how many hours per week they take. Transita calculates the places, that is, how many people are needed. It can be in an area that technology changes or in one that it doesn't.

A reduction of places is the opposite: how many fewer people a position needs because technology absorbs part of its work. It is decided in “Freed hours”.

Vacancies are the destinations of plans: reskilling if in the same area, redeployment if in another. Each plan reserves a place until it is signed, so a vacancy is not used twice. Without vacancies, the only possible plan is a supported exit, unless the company creates a new area.

In the “Progress” tab, by area, position, plan type and person. For each person you see the planned and actual start, the due date, what has been done for each action (training hours, weeks of practice…) and, for reskilling and redeployment, whether they are already in the new position.

The person records what they do from “My plan”; their area lead or HR can record it too and also enters the dates. The progress % is calculated automatically as the average of the actions, and plans past their due date are flagged as late.

So nobody can approve a plan alone. The person's area lead (or an admin) and HR sign, and they must be two different people. With both signatures the plan moves to “in progress”.

Transita does not replace legal advice or calculate severance. Its workflow follows three principles found in labor and AI rules across countries: inform and consult before deciding, train before letting people go, and keep a record of every step.

Consult before deciding: ILO Convention 158 (arts. 13 and 14) requires informing and consulting workers' representatives, and notifying the authorities, when contracts end for economic, technological or structural reasons. In the European Union, Directive 98/59/EC requires the same for collective redundancies, and the AI Act (Regulation (EU) 2024/1689), art. 26(7), requires informing workers' representatives and affected workers before using a high-risk AI system at work (for example, to select, evaluate or assign tasks).

Train before letting go: in Spain, the Workers' Statute (art. 52.b) only allows dismissal for failing to adapt to technical changes if the employer first offered a paid course during working time and at least two months have passed. That's why reskilling and redeployment come first in Transita.

Keep a record: the consultation, the two signatures and the audit log serve as evidence for labor authorities, workers' representatives or a certification.

No. Severance is set by each country's law and paid separately; the cost Transita shows doesn't include it. Some examples:

Argentina: seniority-based severance under art. 245 of the Employment Contract Law (as replaced by Law 27,802 of 2026). If cuts for economic or technological reasons affect a significant share of staff, the crisis-prevention procedure of Law 24,013 (arts. 98 to 105) applies.

Mexico: the Federal Labor Law (art. 439) provides that when new machinery or procedures reduce staff, the employer must seek the labor court's authorization and pay four months' salary, twenty days per year of service and the seniority premium.

Spain: objective dismissal (arts. 52 and 53 of the Workers' Statute) carries 20 days' pay per year, capped at 12 months. Chile: dismissal for business needs, including modernization (Labor Code art. 161), carries 30 days per year, capped at 11 years (art. 163). United States: the WARN Act requires 60 days' notice of mass layoffs at employers with 100 or more employees.

The supported exit adds what the law generally doesn't require: an outplacement program, employability training and extended health coverage, in line with the ILO's just transition principles.

Mexico: in June 2026 a bill was filed with the Permanent Commission to raise severance for dismissals due to AI adoption, up to five months' salary. It has not yet been debated in plenary.

Brazil: the AI legal framework (PL 2338/2023) was approved by the Senate in December 2024 and is in the Chamber of Deputies. The Constitution (art. 7, XXVII) also provides for “protection against automation”, which was never regulated.

European Union: obligations for high-risk AI systems in employment apply from 2 December 2027, under the postponement adopted by Regulation (EU) 2026/1744.

Rules change. This information is current as of October 2026; for a specific case, consult a labor lawyer in your country.

It is a self-assessment against the Transita standard for responsible workforce transition (draft v0.2): 15 criteria in five pillars — measure before deciding, a plan for every person, consult and agree, follow through, and governance and transparency.

Eleven criteria are calculated automatically from the company's data (for example, how many tasks a person validated or how many consultations the person answered themselves). The other four are declarations answered by administrators or HR, with the evidence that supports them.

The PDF evidence package summarizes everything and carries a digital fingerprint stored in the audit log. It is not the seal: the seal is granted by an external review.

When the company meets all 15 criteria, administrators request the external review from “Certification”. The review can grant the seal, leave comments to fix, or not approve it. The seal is valid for 12 months, has a public code (for example TR-ABCD-1234) and a page where anyone can check whether it is valid; there is also a badge for the company's website.

It records every change: who made it, when, and the value before and after. You can filter by type, by who made the change, by dates or by employee ID (to see a person's full history and plan) and export it to CSV.

Events cannot be edited or deleted. They are the basis of the evidence for a certification or for labor authorities.

Yes. The data the company loads belongs to it. From “Settings”, administrators download everything loaded (areas, roles, tasks, projects, people, plans, users and audit log) to a file, and can delete the company with all its data.

Deletion is permanent and asks you to type the company name to confirm. Details are in the Terms of use and Privacy policy, linked at the bottom of every screen.

Only their own plan: path, destination, actions, consultation, signatures and progress. When a proposal is sent to them, they accept it or ask for changes there. They don't see the cost or anyone else's data.

On the dashboard, groups of fewer than 5 people are shown as “<5”, so nobody can be identified by elimination. The dashboard and the report never show names either: a person's name, if added, is only visible to HR, the administration and their area lead.

Every change: who made it, when, on what (task, project, plan or import), and the old and new values. Records can't be deleted from the application.