talk to us
How it works

from a question to a decision you can defend.

Five stages take one decision from the first data export to a plan your team can defend. Planners bring the knowledge of the operation; kint brings the mathematics and keeps the record. Tap or click a step to jump to it.

describe a decision
Stages
01/ 05connect data

your data, versioned from the first upload.

kint starts from read-only exports of the ERP, the TOS or a spreadsheet. Every upload is kept as a version, so a plan always knows what it was built from. Approved plans go back only through an API you control.

TOS exportmaintenance work ordersberth schedule.csvcrane registerREAD-ONLY EXPORTSberth schedule.csv · row 214: date missingsent back to its owner · kept out of the storeCHECKED STOREv18c21e0v23f9a1cv3b7d402v4e15f9acrane 3 offlineplan week 14plan week 15PLANSv4 arrives · week 14 stays on v2

Illustrative example. Values come from your data.

Start with the exports your teams already produce. You do not need a data project first.

02/ 05describe the decision

your planners explain it. in their own words.

Planners explain it in plain language, with no modelling language to learn. kint's decision agent listens, asks what is missing and writes the rules down. It proposes each step, and a person approves it before the work goes on.

PLANNERUse the fewest cranes and carriers per shift,but every ship must leave within 48 hours.DECISION AGENTMust every ship make it, or is a share enough?PLANNERAt least nine in ten.DECISIONcranes and carriers per shiftGOALfewest machinesRULEat least 90 % of ships leave within 48 hDATAvessel calls + handling times · v2 · 3f9a1cRULES ON RECORDreviewed by planningcranes and carriers per shiftfewestwithin 48 hoursnine in tenready to build the model

Illustrative example. Values come from your data.

The planner describes the decision as they would to a colleague, and says what a good plan must achieve.

03/ 05build the model

the agent builds the model. the solver closes the gap.

The agent turns the reviewed cards into mathematics planners can trace line by line. The solver then searches until the gap to the best possible plan is zero, or time runs out.

DECISIONmachines per shiftGOALfewest machinesRULE90 % of ships ≤ 48 hDATAv2 · 3f9a1cREVIEWED CARDSMODELcranes s1-s4 · carriers s1-s4min≤ 48 h≤ 48 h≤ 48 h≤ 48 h≥ 90 %solver · MIP · data v2 · 3f9a1cSOLVER443522machines over search timebest plan foundlower boundGAP TO OPTIMAL0.0 %optimal · 35 machinesbinding: 90 % of ships within 48 hthis rule sets the machine count, so this is where to negotiate

Illustrative example. Values come from your data.

The checked cards are the only brief the model gets.

04/ 05compare scenarios

test the options before you choose one.

What if we add a spare crane, move a maintenance window, close a berth for works? An idea becomes a scenario: the same model, data v2 · 3f9a1c pinned, one parameter changed.

CHECKED MODELsame solverdata v2 · 3f9a1cas todayoptimal · gap 0 %10091 %add a spare craneoptimal · gap 0 %11997 %move maintenance to mayoptimal · gap 0 %8592 %close berth 2 for worksinfeasiblebreaks 90 % ≤ 48 hno planCOST · TODAY = 100≤ 48 HRULESPLANall rules keptrule brokenadopted · lowest cost that keeps every rule

Illustrative example. Values come from your data.

Scenarios start from the model planners already confirmed, with the same data and the same rules.

05/ 05decide with a record

every decision on record, with its history.

Each change to the plan is saved as a version that records who made it, when and why. Scenarios run on their own track until someone approves them into the record. Open any version, re-check it against the record, or start a what-if from there.

PLAN ATapr 2027merge of scenario/maintenance-may · 6 shifts changedre-checked against the recordSTATEdata v2 · rules r2 · plan p3LOGapr 2027 · head of planning · mainmerged the scenario as plan p3why: accepted the 6 changed shiftsapr 2027 · agent · scenariorebuilt the plan, 6 shifts changedwhy: rule r2 changedapr 2027 · planner · scenariomoved the maintenance window to maywhy: crew shortage in aprildata officeagentplannerhead of planningmain · recordscenario/maintenance-maywhat-if/budget-plus-5janfebmaraprmayjuntodayopen · not mergedmain unchanged

Illustrative example. Values come from your data.

Every change is a commit: uploads, rule changes and approvals, each with its author and a reason.

bring one decision. see the five stages on your own data.

In the pilot you get a first model on your data that planners can read, a scenario sheet and the record of each run.

describe a decision
Next
the platform
Models
whitebox · blackbox · graybox · simulation
Contact
hello@kint.pro