CONTACTUS TRACE

See how the work really flows.

Your process map says a billing call takes four steps. Your agents know it takes forty. ContactUs Trace maps the real one, straight from your calls.

THE PROBLEM

The process on paper isn’t the one your agents run.

You see it every Tuesday on the floor.

Workarounds

Fixes agents invented that never made it into the manual.

Repeat calls

The same problems keep coming back, because nothing upstream changed.

Guesswork automation

Projects scoped in a workshop, from what people remember.

78%

78% of complaints are failure demand: problems created outside the contact center.

ContactBabel, 2026 CX Guide

Your contact center is where the rest of the business’s problems show up. If anyone maps the calls.

Where are you today?

Most centers are on two.

One of these sentences is being said in your operation this quarter. Find yours. The next move is already written.

Blind

“We know handle time is up. We do not know what changed.”

WHAT’S MISSING
Volume and averages, with nothing underneath them. Every explanation is a theory somebody defends in a meeting.

WHAT TRACE DOES
Rebuilds the path of every call, so the change has a stage, a loop and a number attached to it.

Reporting

“We have dashboards. We still cannot agree on what to fix first.”

WHAT’S MISSING
Plenty of metrics, no ranking. Ten good ideas and no way to say which one is worth the most.

WHAT TRACE DOES
Prices each finding in counted hours at your loaded rate, and ranks them.

Automating

“We built the bot. It only handles the easy half.”

WHAT’S MISSING
Automation aimed at a process nobody wrote down, where agents already do it eight different ways.

WHAT TRACE DOES
Scores consistency per demand type and separates rule-shaped work from judgment work.

Agentic

“Our agents do not know how the work is actually done.”

WHAT’S MISSING
AI assistants trained on the process document, not the process. They break on the exceptions, which is most of the volume.

WHAT TRACE DOES
Supplies the real journey model — stages, exceptions, systems — as the context layer underneath.

WHY THE USUAL FIX DOESN’T STICK

Most process tools describe how the work is supposed to go. Trace watches how it actually goes.

Process documentation produces a document, and the document gets filed. In a contact center, the work is the conversation.

SCREEN CAPTURE PROCESS TOOLS

Sees the clicks.

Which apps opened, in what order, for how long. Not why the customer called.

CONTACTUS TRACE

Hears the conversation.

What the customer wanted, each stage the call moved through, the systems named, and where it looped back.

	SCREEN CAPTURE PROCESS TOOLS	CONTACTUS TRACE
WHAT IT SEES	Clicks, apps and time on screen	The customer, the agent, and every system named on the call
WHAT IT NEEDS	A capture agent on every approved desktop	Recordings or transcripts from your current platform
WHAT YOU GET	A process map and an ROI projection	Findings in agent hours, with the calls behind every one
WHO ACTS ON IT	A transformation team, eventually	The people running your floor. We run floors too.

WHAT TRACE DOES

Straight from the calls.

Maps intent and how the work actually flows.

Every call broken into ordered stages: what the customer wanted, where the call went, which systems it touched.

Bottlenecks show up. Automation candidates identify themselves.

Instead of being guessed at in a workshop. The loop that repeats. The stage that runs long. The request that takes the same path every time.

Runs on the platform you already have.

Recordings or transcripts from your current system. No migration. The ask is calls, not a switch.

Inference runs locally. No transcript text leaves the machine.

Pilots are open.

Bring us a set of your calls, and we’ll show you how the work actually flows.

One site, one question to start
Findings reviewed with the people who run your floor
Expand once it has earned it

SAMPLE OUTPUT

Four views from a live report.

Figures here are illustrative. Your report runs on your calls, your demand types, your rate.

Where the journey actually goes

Stage boxes carry the share of journeys that touch them and typical dwell time. Line weight is the share of that stage’s outflow. Dashed lines are rework.

No linear view can show a loop. A paths table lists information gathering → problem statement as one more sequence. An averaged chain drops it. Here it is 680 calls of work done twice, with a price on it.

The cost of not closing the call

Resolution rate when the agent closes the call properly, against when they do not, by demand type. Red is unclosed, green is closed, the number is the gap in points.

How repeatable each process really is

Consistency scored 0–100 per demand type, with the count of standard actions against variable ones. Demand types below the volume floor are excluded.

Variable actions outnumber standard ones roughly four to one. These processes are not written down anywhere except in what agents happen to do — which is why the same demand arrives at the same queue and gets handled a different way, and why automation stalls.

Where the screen time goes

Total minutes per system (solid bar) against minutes per touch (light bar). The busiest system and the slowest system are rarely the same one.

The order system takes the most total time because it is opened most often, not because it is slow. Hold telephony costs more than twice as much per use. Reading these two together is what stops a tooling program from optimizing the wrong screen.

WHAT LANDS IN YOUR INBOX

Six kinds of recommendation.

Each arrives with the calls behind it and the hours it is worth. Figures are illustrative.

PROCESS

Standardize the winning version

The same demand handled six ways. Trace finds the version that resolves fastest and names the actions that make it work.

41 hrs · high confidence

REWORK

Close the loop that repeats

A stage customers pass through twice. Fix the first pass and the second one disappears.

38 hrs · high confidence

COACHING

Coach one named behavior

A single action, present in the strongest agents and absent in the weakest, worth a measurable point gap.

29 hrs · high confidence

TOOLING

Fix the screen, not the agent

A system that costs more per use than any other. Usually a default, a missing field, or a lookup that should be one click.

24 hrs · medium confidence

DEFLECTION

Move a demand type off the phone

A repeated, rule-shaped request with a path already proven on the floor. The candidate for self-service, with volume attached.

36 hrs · medium confidence

AGENT ASSIST

Give the assistant real context

The stages, exceptions and systems an AI assistant needs to handle the eight per cent that is not the happy path.

Scoped, not yet priced

WHAT THE REPORTS WON’T TELL YOU

We’d rather remove a chart than publish a number we can’t stand behind.

Every Trace report states its own limits, in writing, next to the findings.

No step counts.

The measurement error is higher than we’d accept. So the view most process tools lead with isn’t in the report.

No dollar figure.

Every value is in agent hours. Money needs your loaded rate, and only you have it. Where the analysis assumes something, it says so.

A second reading.

The same calls, read again independently against your own model. We publish the gap between the two.

Its own defects, listed.

What the instrument can’t answer yet becomes a calibration agenda: what we need from you, and what each answer unlocks.

Inference runs locally. No transcript text leaves the machine.