From Calibration to Go-Live

Both, and they answer different questions. A demo runs in a TetriXX AI environment with realistic freight data and shows how verification and the reports work. A calibration runs on your invoices, under NDA, and is free of charge. It is the step that tells you something you do not already know.

If you only do one, do the calibration. A demo shows what the platform does with data that was formatted for it.

For a calibration on your own data: details for one or two logistics providers (modalities, trade lanes, rate cards), and two to three months of invoices. That is enough to establish both the error profile and whether your invoice data supports continuous verification at all.

From a freya by TetriXX AI calibration you receive an integrated dashboard with your structured data fully accessible, a detailed calibration report with findings, recommendations and identified opportunities, and a GLEC-based CO₂e emissions baseline across the shipments whose data supports it. It runs under a signed NDA.

For a calibration, a few hours of a logistics analyst’s time to collect invoices and rate cards, plus whoever can authorise the NDA. The work is collection, not analysis.

For a full deployment the two real asks are provider data access and a named internal owner for the exception queue, someone who reviews flagged charges weekly. Implementations that stall almost always stall on one of those two, not on technology.

You receive an integrated dashboard with your structured data fully accessible, a detailed calibration report with findings and quantified opportunities, and a GLEC-based CO₂e baseline across the shipments whose data supports it. Then you decide, with a number in hand rather than a projection.

There is no obligation attached to the calibration, and the report is yours whether or not anything follows.

Yes, and it is the recommended way in. One or two providers is enough to prove the mechanism, tune the exception queue against your real contract structures, and produce a number that justifies widening scope. Starting with every provider and mode at once is one of the four common implementation failure modes.

The objective is always the same: in production in under three months. That is what we plan against and what we have delivered.

What moves it is circumstance rather than effort: the scope you choose to start with, the modes and markets in play, and how many service providers have to be brought into the loop and how fast each grants access to invoice data and rate cards. The first provider takes longer than the ones after it, which is why scope is normally phased in waves rather than switched on at once.

Implementations run long for three reasons, and none of them is the technology: rate cards that turn out not to exist in usable form, a provider slow to grant invoice access, or an internal security review starting from scratch because the vendor’s evidence was never published. TetriXX AI publishes its framework at tetrixx.ai/legal for that reason.

A real plan needs your real numbers. Send us your spend, modes, geographies and provider count and we will come back with a phased plan for your scope, with the milestones and the decision owners on both sides written down for it.

A named team, and the same names throughout, not a queue and not a rotating bench.

Through setup you work with a dedicated Implementation Lead, supported by data modelling specialists and integration engineers. At go-live the account moves to a dedicated Customer Success Manager who stays with it: the same person through hypercare, steady state and quarterly business reviews, with a direct escalation path into product and engineering rather than a support ticket.

Delivery and customer success are staffed functions with a single owner on our executive team. Both are amplified by the same agent infrastructure that runs the platform: routine modelling, reconciliation and exception triage are handled by agents, which is what lets a named human stay on your account rather than being spread thin across a portfolio.

You will meet the founders, and they own the commercial relationship. Delivery does not depend on their calendar.

Four failure modes, in rough order of frequency: rate cards that do not exist in usable, current form; invoice data missing the fields needed to match a charge to a contract; no internal owner for the exception queue, so findings pile up unactioned; and a scope that starts too broad, every provider and mode at once.

All four are diagnosable in advance, which is the argument for a calibration on real data rather than a demo on the vendor’s.

You do. A well-designed audit platform is human-in-the-loop: deviations are raised for review with the reason attached, and a person approves, rejects or corrects each one. Corrections feed back so classification accuracy improves against your contracts.

Be specific with any vendor about this. Ask whether the platform disputes charges with carriers automatically, and whether that behaviour can be turned off. freya by TetriXX AI raises everything for review and disputes nothing on your behalf automatically.

That is a decision you make rather than a number the software fixes. The share of charge lines that reaches a person is set by how the audit workflow is configured, and across live deployments it runs from under one line in five hundred to roughly one in three. The spread is not a difference in accuracy. It is three different answers to the same question.

One operation wanted the fewest possible items reaching its operations managers, so price differences below a set money threshold clear automatically. Fewer than two charge lines in every thousand now reach a person.

A second maintains its own rate cards and knows there is drift between what it negotiated and what the system holds. It approves no overcharge automatically, because each deviation is what finds that drift. About one line in eight is reviewed, and that is the purpose of the deployment rather than a cost of it.

A third buys largely from providers who work without rate cards, so those movements arrive with no contracted price to verify against, and the team reviews them deliberately.

So a straight-through percentage tells you very little on its own. Ask whether the threshold is yours to set, ask which unit the percentage is measured in, per invoice or per charge line, and ask what happens when you want to see more rather than less.