Exagion

AI-native planning & optimization

Optimization that speaks human.

Describe your fleet, schedule or supply chain in plain English. Exagion’s AI turns it into a rigorous optimization model, our solvers find the best plan, and you can re-plan in seconds when something changes.

Request a demo See how it works

Built for

  • Routing & dispatch
  • Workforce & production scheduling
  • Supply chain & inventory
  • Forecasting & what-if planning

Example: routing · Truck 12, Port of Oakland → Redwood City

OriginalRe-planned

Map of the San Francisco Bay Area. The original route, dashed, crosses the Bay Bridge and runs down US-101. The re-planned route runs south on I-880 and west on CA-237, crossing no bridges. Warning markers sit on the Bay, San Mateo and Dumbarton bridges.

Dispatcher

Wind advisory just went up on the Bay bridges. Keep Truck 12 off all of them.

Exagion

Done. I moved Truck 12 off the Bay Bridge and onto I-880 South and CA-237, with no bridge crossings. That’s about 20 mi and 23 min more. Downstream stops are updated.

News

Selected for a Navy SBIR Phase I

Exagion has been selected for a Department of the Navy SBIR Phase I under FY26 Release 1, topic DON26BZ01-NV027 (Office of Naval Research): Automated Ice Detection and Polar Navigation Tool.

Navy SBIR selections

The gap

Solvers are rigid. Real operations aren’t.

Optimization models sit at the center of serious planning: routes, shift schedules, production runs, inventory, budgets. But across commercial, industrial and defense settings, the plan has to survive a world the model never saw coming.

01

Constraints change hourly

Breakdowns, absences, outages, demand swings and new regulations.

02

Tribal knowledge rules

The rules that matter most live in planners’ heads, not in the model.

03

Errors are expensive

Mistakes and delays carry a high operational cost.

04

Interfaces don’t fit

Solver inputs rarely match how people actually plan.

The product

One interface for every plan

Exagion puts a natural-language interface on industrial-strength optimization. Planners can:

  • Describe goals and rules in plain English, like “no one works three nights in a row” or “keep two trucks in reserve”
  • Re-plan in seconds when exceptions occur
  • Pull in data from the systems they already run: TMS, ERP, scheduling tools and spreadsheets
  • Explore tradeoffs interactively instead of accepting a single “optimal” plan
  • Ask why, and get summaries, plots and explanations of any plan
Exagion planning console for a Dallas last-mile delivery day: the dispatcher's heat-warning instruction, the model change Exagion drafted, six van routes on a map, and a schedule showing every van back by 1:58 pm

Technology

From static solver to decision support.

Exagion bridges formal optimization models and real-world decisions, so planning tools work the way people actually plan and adapt.

Plain-English constraints

Our AI translates guidance in plain English into model rules and states its assumptions, with an emphasis on avoiding hallucinations.

The optimizer decides

The AI never draws the plan itself. Every route, shift and assignment comes from a mathematical solver.

Connected to your stack

A flexible API pulls in data from the systems you already use: transportation management, ERP, scheduling and more.

Auditable by design

Interrogate any optimization output to get summaries, plots and explanations of why the plan looks the way it does.

A planner at a multi-monitor workstation, on the phone, surrounded by whiteboards of schedules

Use cases

Built for the people who plan.

Routing & dispatch

Re-optimize in real time when trucks break down, roads close or orders change mid-shift.

Workforce & production scheduling

Build shift, crew and production schedules from rules people actually say, and repair them when someone calls out.

Supply chain & inventory

Decide where to stock, what to move and when to reorder, and see the cost and service tradeoffs side by side.

Forecasting & what-if planning

Turn demand or economic forecasts into plans, then ask what changes if fuel rises 20% or a supplier goes offline.

Demos

Plain-English guidance in, new plan out.

Three short narrated demos built on synthetic data: one commercial delivery day and two defense logistics exercises. In each, a dispatcher or planner types guidance in ordinary language. Exagion’s AI writes it into the optimization model as new rules and states its assumptions. The planner applies the change, and Exagion’s solver builds the new plan. The AI never draws the plan itself: every route and assignment comes from the optimizer.

Commercial · 0:34

Beating the heat on a Dallas delivery day

“Excessive heat warning from 2 pm. Get every van back to the depot by 2:00, keep every delivery inside its window, and use a spare van if you need one.”

Exagion writes a 2:00 pm return deadline for every van, keeps all 60 delivery windows unchanged, and allows a spare van. The optimizer re-plans the day: the last van is now back at 1:58 pm instead of 3:17 pm.

60/60

deliveries still on time

1:58 pm

last van back, before the 2:00 deadline

+1 van

and 6% more miles

Defense exercise · 0:34

Cutting fuel in northern Sumatra

“JP-8 resupply to Banda Aceh is disrupted. Cut fuel use as far as it will go and keep a CH-53K on strip alert for CASEVAC, but every request still has to be delivered in full.”

Exagion turns this into a fuel objective, a full-delivery requirement and one helicopter held back from all flights. Cargo that was flown by three CH-53Ks moves to road convoys of MTVRs and JLTVs.

12/12

requests delivered in full

−77%

JP-8 fuel use

1 of 3

CH-53Ks flying; two stay available

Defense exercise · 0:34

Convoy rules under an IED threat in Borneo

“IED threat on the Banjarmasin–Syamsudin Noor road. No truck runs it alone: minimum 4-vehicle convoys on that road. Keep two MTVRs at Banjarmasin as a QRF, and everything still has to arrive.”

Exagion writes a four-truck minimum for any convoy on that road, a two-truck QRF that stays at Banjarmasin, and full delivery. The threatened road is now run by a 4-truck convoy, up from 2.

68/68

requests arrive

4 trucks

convoy on the threatened road

2 MTVRs

held at Banjarmasin as QRF

Our mission

To bridge the gap between complex optimization models and real-world decision-making, so teams in logistics, scheduling and supply chains can handle fast-changing constraints through a natural-language interface that works with the software they already use.

Partners

Who we are

The team

John Carlsson

John Carlsson

Founder & CEO

USC Professor of Engineering, 15+ years in vehicle routing. “The man who reroutes the world” — Popular Science.

LinkedIn →
Carol Sexton

Carol Sexton

Founder & CTO

20+ years at Oracle scaling database and AI products; 15 years as head of Oracle Security Program Management.

LinkedIn →

Contact

Let’s talk about your plan.

Tell us what you need to plan, from routes and schedules to a whole supply chain, and the rules you work within. All information is kept strictly confidential.

jgc@exagion.ai

    All information is kept strictly confidential.