AI and automation don’t fail because of the technology—they fail because of the data. Pilots succeed on curated datasets, while production environments rely on decades of fragmented, inconsistent network information. AI won’t tell you the data is wrong; it will confidently act on it. That’s why trusted network data is the foundation of every successful AI deployment.
AI needs trustworthy data.
This is why MarlinDT exists: to bridge the gap between a digital map and a true digital twin. We create one trusted, validated and up-to-date model of your entire network. That foundation unlocks the next stages: intelligent insights, predictive capabilities and, ultimately, autonomous operations.
See the four maturity levels →
We use AI a lot — quietly, under human supervision, where it genuinely works.
A coax migration that a decade ago meant a 24-month project and a large team, we now deliver in under nine months. AI flags the records it can’t confidently migrate, finds similar cases, watches how an expert corrects one, and applies what it learns — human-in-the-loop, at scale.
From LiDAR, mobile mapping or a technician with a camera, AI detects objects and assigns state — number of floors and apartments from a lobby photo, grass route versus concrete, a damaged asset — and captures locations and depth from on-site images during as-built. Built on FME Flow and the ArcGIS ecosystem.
Schematics trapped in PDFs, Visio and scanned diagrams? AI extracts the information and maps it into your new digital twin, turning legacy documentation into live inventory.
Adding a new tool — an AI survey app, a new sensor — is the easy part. The part that matters is integrating its dataflow into your digital twin. Bolt on a clever AI tool that produces yet another standalone dataset, and you’ve recreated the digital map: data that can’t be used for root-cause analysis or planned maintenance, sitting in its own silo, costing money to keep. New data has to be linked into the twin, not parked next to it.
Merkator innovates continuously — testing new tools and building our own — and we partner with AI specialists who build the next-level use-cases. Our job is to make sure the network is modelled correctly first, so that when the AI runs, it runs on the truth.
