Skip to content
Sentynel

#opinion3 min readBy Sentynel

Speeding up a step does not speed up the work

Most AI pilots make one task faster and leave the queue untouched. The time a case spends waiting between steps is where an operation is won or lost.

Share

When a company adds AI to a process, it usually picks a task and makes it faster: drafting the reply, summarizing the file, classifying the ticket. The task gets quicker and the customer waits just as long. A case spends most of its life waiting between steps, not being worked on, and a faster step does nothing about the wait.

Where the time actually goes

Follow one case from the moment it arrives to the moment it is resolved. Mark every minute as either work (someone or something is acting on it) or wait (it sits in an inbox, a thread, a spreadsheet or someone's memory).

In most operations the waiting dominates. A message arrives on WhatsApp, someone copies it into the CRM later, a person decides who owns it when they get to it, and the owner replies after finishing something else. Each step might take two minutes. The case takes two days.

An AI that drafts the reply in ten seconds instead of two minutes has improved the part of the timeline that was already short.

The wait has an owner problem

Waiting is rarely a speed problem. It is an ownership problem:

  • Nobody is assigned to the case until someone notices it.
  • The next step needs information that lives in another tool.
  • A person has to approve, and the approval is a message in a chat.
  • The case is blocked, but nothing records that it is blocked.

None of these is fixed by a faster model. They are fixed by deciding, in advance, who or what picks up each case, from where, and by when. That is operating design, and it comes before the agent.

Where a case waits
  1. 01ArrivesWhatsApp, web, social
  2. 02RecordedOne queue, one record
  3. 03AssignedOwner named automatically
  4. 04WorkedAgent or person
  5. 05ResolvedState changes, timer stops

Every arrow between two boxes is a place a case can sit unattended.

What changes when you design the handoffs first

In the Rodo case, WhatsApp, social and web enquiries were brought into one service operation with a visible path to resolution. Rodo's own operational report reports that 82% of 604 live-created enquiries reached Resolved, with a median of about one hour. That is a first measured window, covering 31 August to 24 September 2026, with no comparable earlier baseline, and the figures describe pipeline records, not all conversations.

The point is not the percentage. It is that the operation was built so that every enquiry had a place to be and a next state to reach. Once that exists, an agent working inside it has something to speed up. Without it, the agent speeds up a step inside a pile.

Rodo · first measured window
of live-created enquiries resolved
~1 hmedian time to resolved

Source: Source: Rodo operational report, 24 September 2026. Resolution cohort covers 31 August–24 September. Figures describe pipeline records, not all conversations. First measured window, no comparable earlier baseline.

A test you can run this week

Pick one process and pull twenty recent cases. For each, write down the timestamp of every state change, then split the total time into work and wait. Most teams find that work is a small fraction of the total.

Then ask where the longest waits happen and who could have acted on the case during them. If the answer is "nobody was assigned", the next investment is not a better model. It is a rule for who owns the case at that moment.

If you removed every wait in your busiest process, how much of the remaining time would an agent still be able to save?

Share
← All sygnals