Decision Engineering
From spreadsheet planning
to better decisions.
Shift rosters, delivery routes, order quantities: in most companies, someone works these out by hand, from scratch every week. It works, but it takes time, and nobody knows how good the plan really is. At Anexia, we combine AI with mathematical optimisation: models that predict what is coming and compute the best plan for it, with all the rules that apply in your business, and tools your planners can work with day to day.
Typical problems
We hear sentences like these a lot. Behind each one is a forecasting question and an optimisation problem.
Scheduling & dispatch
“Scheduling eats half of every Monday morning. By Wednesday the plan is outdated again.”
Tours & routing
“Our drivers are on the road a lot, but nobody knows whether the routes are any good.”
Inventory & ordering
“Either the wrong things are in stock, or the right ones are missing.”
Staff & shift planning
“Preferences, qualifications, rest periods, holidays: nobody likes doing the roster.”
Resource allocation
“We have more requests than capacity and decide by gut feeling what comes first.”
Pricing & capacity
“We can’t tell whether we’re leaving money on the table with utilisation and prices.”
How a project runs
We start small and only continue once it’s clear that it pays off.
- 01
Decision audit
2–3 daysWhat happens
We look at the decision together: what should be achieved, which rules apply, how is it planned today?
What you bring
The people who plan today and some sample data. Spreadsheets are fine.
What you get
A clear description of the problem, an assessment of whether optimisation helps here, and a first estimate of what it is worth.
- 02
Model & prototype
1–3 weeksWhat happens
Where needed, we train AI models that forecast demand, durations or absences. The decision itself we formulate as an optimisation model, and build a prototype that computes plans from your real data.
What you bring
Data from a typical period and quick feedback on the first plans.
What you get
Plans for your real cases, and you can see why they look the way they do.
- 03
Comparison & ROI
alongside the prototypeWhat happens
We put the computed plans next to today’s: cost, time, utilisation, planning effort.
What you bring
The metrics that matter to you.
What you get
Numbers you can use to decide whether implementation is worth it.
- 04
Implementation & operation
as neededWhat happens
Integration with existing systems and an interface for planners, optionally with an AI assistant they can ask in plain language. Plus operation and further development.
What you bring
Access to the systems and someone internally who owns it.
What you get
A tool in daily use. The final decision stays with your people.
Why AI and optimisation together?
AI is strong where uncertainty is involved: machine learning forecasts demand, travel times or absences from historical data, and language models understand text and answer questions. But neither reliably respects a roster with fifty rules. And a plan that breaks one rule is useless in operations.
An optimisation model respects every rule and searches all valid plans for a very good one. At Anexia, we use models like these in railway planning, among other areas.
In practice we combine three building blocks: machine learning predicts what is coming, optimisation computes the plan, and a language model takes requests and explains the results.
Machine learning: predicts
- Demand, durations, absences
- Learns from historical data
- Says what is coming, not what to do
Optimisation: decides
- Respects every rule
- Finds provably good plans
- Builds on the forecasts
Language model: explains
- Requests in plain language
- Explains why a plan looks the way it does
- People make the final call
Would it pay off for you?
One conversation is usually enough for a rough estimate. Projects are delivered by our team at Anexia.
Let’s talk