WEEKLY NEWS “Our LLMs are trained by us, on our own servers,


with our own dataset,” he explains. “That ensures customer data is never shared, never exposed, and never used to train public models.” The dataset in question spans millions of searches,


quotes, and bookings from more than 150 countries. Combined with anonymised industry


inputs and


upcoming integrations with IATA ONE Record, the result is a rare blend: transaction-level data at global scale. This data allows the platform to make predictions


AI IN THE INBOX BY Anastasiya SIMSEK


AS the air cargo industry wrestles with digitisation at scale, artificial intelligence is shifting from hype to implementation. And according to Matt Petot, founder and CEO of CargoAi, that shift is happening directly inside the inbox. While many


forwarders navigating basic and e-booking airlines adoption, are still CargoAi’s


latest push embeds AI tools into quoting, pricing, and customer interaction — automating everything from rate prediction to natural language email replies. If


it works as claimed, it could significantly raise productivity without overhauling legacy systems.


Conversational AI CargoAi’s approach puts AI front and centre in its


10


booking platform, CargoMART, but avoids separating it into standalone apps. Instead, AI operates under the hood — predicting rates, identifying the best routing options, and now responding to emails. “Our


new CargoCopilot AI Agent is a full


conversational interface,” Petot says. “Forwarders can simply type ‘Book 300 kg from SIN to LHR next Tuesday’ or ‘Show me rates to JFK’, and get instant results.” Airlines and GSAs, he adds, are using the same tool to triage thousands of customer emails daily — from availability checks to delivery instructions. While digital quoting tools are not new, Petot


argues that AI turns static interfaces into dynamic decision-makers.


“It’s not just about speed. It’s


about replacing repetition with reasoning.” One of the key differentiators Petot highlights is


that CargoAi trains its models in-house using its own proprietary dataset — not third-party or public LLMs.


that, in Petot’s words, “no single airline’s RMS can make.” For instance, it can forecast with up to 90 percent accuracy which offer a forwarder is likely to accept — giving carriers a chance to optimise offers in real time. CargoAi claims its tools deliver measurable


returns, especially for forwarders caught between growing volumes and limited resources. “For mid-sized forwarders, we see 30–50


percent faster quote response times and a 10–15 percent uplift in conversion rates,” Petot says. “One Southeast Asian forwarder processed four times more quotes per


rep within 90 days — without


adding headcount.” On the airline side, the automation scale is more


aggressive. “One GSA now processes over 10,000 quotes per month automatically. That’s not proof of concept. That’s production.” Beyond volume, Petot is keen to stress that the


models prioritise profitability, not just filling space. “Our models are built to protect yield. In volatile markets, the system adapts its priorities in real time — volume never comes at the expense of margin.” That might be the most important distinction. AI in air cargo often leans toward speed and automation,


AIR CARG O WEEK


but ignoring revenue management nuances would risk cutting into the very profits it promises to boost. What may appear at first as an email plugin is, in


fact, where Petot sees the real strategic shift — using email content, CRM history, and booking data to personalise offers and automate replies without compromising on context. “The Agent understands cargo type, service level,


even past negotiation patterns,” he says. “When confidence is high, it responds instantly. When it’s low, it flags for human review.” The decision to build in confidence scoring — and


keep a “human-in-the-loop” for sensitive or complex requests — reflects a practical understanding of how AI tools should operate in regulated, high-value logistics. Surprisingly, this is also attracting smaller


forwarders who aren’t running full TMS platforms. “Over 40 percent of our plugin users don’t use a TMS,” Petot notes. “They connect via email or CargoMART directly — no system overhaul needed.” That’s no small detail. With much of the industry


still running on email, spreadsheets, and semi- structured workflows, AI adoption needs to meet users where they are — not demand a digital leap they’re not ready for. Despite the progress, Petot admits the industry


remains divided. A few airlines are leaning into full integration — seeing clear gains in quoting speed, conversion, and load factor management. But many remain stuck in pilot mode. “Adoption speed matters more than ever,” he


warns. “AI compounds its advantages: faster learning, better data, more market share. The performance gap between early adopters and latecomers will be huge.”


ACW 01 DECEMBER 2025


www.aircargoweek.com


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18