Skip to main content
Viithiisys
Employee Operations Automation

The internal requests that cross three teams and stall in the middle of all of them.

Onboarding, access, equipment, approvals and leaving. Each is a sequence nobody owns end to end, which is why each one is slower than any individual step in it.

One leaverAccess after the last day
Last dayHandled from memory
What was grantedState

The list of what to revoke is reconstructed each time.

Shape of an access list, illustrative. No system or count on this page is read off this drawing.

Four sequences

Four sequences that fail the same way.

These are coordination failures, not capability gaps.

  1. A new starter waits on three teams in series

    IT, facilities and the manager each act correctly and sequentially, so a two-day setup takes two weeks and nobody is at fault.

  2. Access requests have no visible state

    The requester cannot see where it is, so they chase, and the chasing is now a meaningful share of everyone's week.

  3. Approvals live in inboxes

    The approval exists as a message. When the approver is away, the process has no defined behaviour at all.

  4. Leavers are handled from memory

    Which is a security exposure rather than an inconvenience, because the list of what to revoke is reconstructed each time.

What it costs

Three costs, and one of them is a security finding.

Every individual team is doing its part correctly, which is why buying a better tool for any one of them changes very little.

  1. New people are unproductive while they wait

    The salary is being paid from day one. The capability arrives later, and the gap is pure loss.

  2. Chasing becomes a job nobody was hired for

    Managers and requesters spend their time asking where things are, which is coordination work disguised as management.

  3. Leavers keep access they should not have

    Not through negligence but because the revocation list is rebuilt from memory each time. This is the item that appears in an audit.

  4. Nobody can see where the time goes

    Without state, there is no cycle time, so the process cannot be improved and its cost stays invisible.

Before and after

The workflow afterwards.

Every block below is the same width. Only the starting point changes: strung end to end above, all starting together underneath. No width here is a duration.

BeforeIn series
  1. Requests move team to team in series.

  2. State is invisible, so people chase.

  3. Approvals sit in an inbox.

  4. Leaver revocation is remembered.

  5. Cycle time is unknown.

AfterAt the same time
  1. Steps that do not depend on each other run at the same time.

  2. The requester and the manager can see where it is without asking anyone.

  3. Approvals have a defined route, a timeout and a stated fallback when someone is away.

  4. It is a defined checklist derived from what was granted, with the closure recorded.

  5. It is measured, so the slowest step is a fact rather than an opinion.

Said plainly

Most of this is not an AI problem, and it is worth saying so.

The category is sold as AI automation, but most of the recoverable time comes from sequencing and state, which are ordinary software. Knowing the difference keeps it small.

The stepWhat it actually needsWhy
  1. Steps running in series that need not

    Ordinary software

    Ordinary workflow orchestration

    The fix is the dependency graph, not intelligence. This is usually the biggest single win.

  2. Requesters cannot see progress

    Ordinary software

    A state model and a status view

    Nothing needs interpreting. Something needs recording.

  3. Approvals with no fallback

    Ordinary software

    A rule, a timeout and a named deputy

    This is a policy decision expressed in software, not a prediction.

  4. Requests arriving as free text in messages

    Benefits from AI

    This part benefits from AI

    Classifying an unstructured request into a defined type is genuine interpretation.

  5. Answering the same policy questions

    Benefits from AI

    This part benefits from AI, over owned sources

    Retrieval with provenance. See knowledge automation for the ownership prerequisite.

  6. Deciding who should have access to what

    A person, always

    The consequence of being wrong is a security incident, not a delay.

Roughly the first three rows are where most of the time is recovered, and none of them requires a model. A proposal that puts AI in every row is selling scope rather than solving the problem.

How the work runs

How the work runs, step by step

  1. Time one real sequence end to end

    Usually onboarding, because everyone recognises it. Measure elapsed time per step, not effort, since the waiting is the cost.

  2. Separate dependency from habit

    Which steps genuinely require an earlier one to finish, and which are sequential only because that is how the email chain grew.

  3. Give the process a state model

    Defined states, one owner per state, and a visible current position. This alone removes most of the chasing.

  4. Automate the sequencing, then the interpretation

    Orchestration first because it is where the recoverable time is. Classification and retrieval afterwards, where they earn a place.

  5. Close the loop on leavers

    Revocation derived from what was granted, with completion recorded. This is the part with an audit consequence.

Evidence

Where this work has been done

Answer quality, provenance and knowing when a system should decline are the substance of this work, and they are already how Viithiisys runs AI in production.

  1. Vizitor, workplace operations at spread

    In daily use across 500+ workplaces in 15+ countries.

    Software for workplace operations that has to work the same way in many locations, which is the constraint this solution has. Multi-site consistency is the part that breaks first when a process grows by accretion rather than by design, and it is the part this work is built around.

  2. Milo, a time-critical operational sequence

    A driver-facing system where a coordination delay has immediate operational cost.

    Relevant to sequencing work under time pressure, where the fix is almost always the order in which steps run rather than the speed of any single one.

  3. Fitelo, controls in production

    Evaluation and cost controls in place rather than added after an incident.

FAQ

Questions we are actually asked

Pick a topic, or ask us directly. We answer every inbound within one business day.

Still have questions?

Talk to a senior engineer, not a bot.

Talk to us
Start here

Tell us how long onboarding actually takes.

Elapsed days, not effort. If most of it is waiting rather than working, this is the right page, and the first fix is usually sequencing rather than anything involving AI.