The TMS isn’t artificial intelligence, and AI isn’t a TMS. We expect AI to be a multiplier on base layer systems like the TMS or master data management.
Not only does AI not replace the basic TMS modules: a stable TMS with good quality data is a precursor to benefiting from AI. As McKinsey reports, most companies who have embraced AI early and also had measurable benefits from it were already mature in their IT infrastructure. In other words, AI doesn't help companies leapfrog investing in basic IT systems. AI for transport planning is best suited for companies that already take data quality and analysis seriously, and who are comfortable with their TMS.
If you follow the Gartner Pace Layered Architecture you’d place the TMS as a system of record (or possibly a system of differentiation for logistics providers), and AI as a system of innovation.
We see the TMS software landscape as competing on indicators of maturity. These include having the largest footprint of features, the widest regional applicability, the lowest cost of ownership, longest list of references, and many 3rd party implementation options. Release cycles are yearly, or less, for significant technical changes.
And so the TMS is almost the definition of the system of record layer. It is well understood, fairly low cost, and changes slowly.
AI is going to come to transport on another layer. In niche, rapidly evolving add-ons that multiply the value of the base layer's data and process investments. AI will also be secretive, as it has huge potential to be a competitive weapon. In short, there are important technical and functional differences between TMS optimisation solvers and AI for trucking. But pace layer thinking explains why AI shouldn't be a TMS module for the foreseeable future.