AI Advisory for Private Equity Portfolios in Logistics.

Freight brokers spend 8 to 15 minutes per load researching carriers, comparing DAT rates, and calling first-choice carriers. Margin per load gets squeezed on the spot market when the broker takes the first acceptable rate instead of the right one. Most brokers cover 8 to 15 loads per day; the per-load research time is the binding constraint on throughput.

AI carrier-matching (Loadsmart, Uber Freight enterprise APIs, in-house builds on DAT/Truckstop rate feeds plus the broker's own carrier history) predicts the right carrier from rate, lane history, on-time performance, and current capacity in under 60 seconds. Margin improves 1 to 3 percentage points per load. Broker throughput lifts 30 to 50%. The broker keeps the relationship-call override on every load that matters.

Dock door utilization at mid-market 3PL warehouses sits at 55 to 70%. The dock manager runs the schedule on a clipboard and a phone. Carriers wait at the gate for 1 to 4 hours when the schedule slips. Throughput is the binding constraint on the entire facility's labor productivity.

AI dock scheduling (Shipwell, FourKites, Trimble) replaces the clipboard with carrier-preference learning, predictive ETA matching, dynamic re-slotting when ETAs shift, and integration to the WMS for inbound prep. Throughput lifts 20 to 30%. Carrier dwell time drops 15 to 25%. The warehouse manager keeps the override on exceptions; the model carries the cognitive load on the normal day.

50 to 70% of inbound CSR contact volume in 3PL and trucking is trace, ETA, or status. The data already exists in the TMS and the visibility platform. The CSR's job is mostly to read it back to the customer over the phone or email. Headcount scales with order volume in a business that should run flatter.

A retrieval-grounded agent over FourKites, project44, Trimble Visibility, or Shipwell tracking plus the customer's order context resolves 60 to 80% of trace and ETA queries without escalation. The redeployed CSR capacity moves to exception management. On-time performance lifts measurably in the first 60 days because the CSR team starts solving problems instead of describing them.

Yard management at most mid-market 3PLs is undocumented tribal knowledge. The yard jockey knows which trailers are where, which are loaded, which are empty, and which need to move. When the jockey is sick, the operation slows by 25 to 40% for the day. Yard inventory accuracy sits at 70 to 85% on a good day.

Computer vision yard management (Outrider, Phantom Auto, Trimble PeopleNet, custom builds on overhead cameras plus a vision LLM) tracks trailers in real time, calculates dwell time, and flags trailers approaching detention. Yard inventory accuracy goes to 95 to 99%. The jockey workflow becomes documented instead of tribal. Detention exposure drops 30 to 50% in the first quarter.

Asset-based trucking dispatch is built on a whiteboard and a 30-year-old dispatcher who knows every driver's preference. When she retires (and she's planning to), the operation drops 20 to 30% in productivity for the first six months. HOS compliance, customer SLAs, driver preferences, fuel cost, and equipment availability all factor in; the dispatcher carries it all mentally.

An AI dispatch agent that respects HOS rules, customer SLAs, driver preferences, fuel cost, and equipment availability, then proposes the next-best assignment with the trade-off explained, makes the dispatcher's institutional knowledge transferable. Magnus Technologies and Carrier Logistics (post-STG) ship this pattern. The dispatcher keeps the override. The model carries the cognitive load. Productivity lifts 8 to 15%; succession risk drops to near zero.