A

agentic GIS

GIS run by an AI agent from a plain request. In a 2026 benchmark, the best agent completed about a third of realistic tasks exactly.

AI agent

An AI system that plans and carries out multi-step tasks using tools. In a simulated software company in 2025, the best agent completed 30% unaided.

AI hallucination

When an AI states something false with confidence. A national standards profile calls it confabulation, a natural result of how generative models work.

anomaly detection

Finding values or patterns that don't fit what is normal. It complements matching rules, which catch broken links between records such as a wrong rate.

API

A defined way for one program to ask another for data or actions. Whether a timber system has one, and on what terms, decides how usable its data is.

audit trail

A time-ordered record of who changed what, when and why, enough to rebuild how a figure or record reached its current state.

C

computer vision

AI that interprets images and video: finding logs, knots, cracks and grades. Its accuracy on the test set and its effect on yield are different things.

confidence score

A number an AI attaches to a result to say how sure it is. It routes work to people well only if it is calibrated, which is worth testing.

connector

A ready-made link that lets one system read from or act in another. For AI assistants, it should see only what the user is already allowed to see.

D

dashboard

A screen of the few measures that matter, updated from live data, readable at a glance. It shows what its definitions and cost pools let it show.

data connector

The piece of a data pipeline that reads from one source system. Common systems have ready-made ones; many timber systems need custom work.

data governance

Who decides what data means, who may use it and how its quality is kept. Large firms say the barrier is people; small ones may lack the tools too.

data lake

A store for raw data of every kind, including scanned documents and machine files, kept as it arrived. Without a catalogue it becomes a data swamp.

data lineage

The record of where a figure came from and what happened to it on the way: sources, joins, conversions, adjustments. It lets anyone check a number.

data model

The agreed map of the things a business records and how they relate: contracts, loads, logs, boards, orders. It decides which questions data can answer.

data pipeline

The automated route data takes from source systems to where it is used, cleaned and joined on the way. Judge it by what it does when a source changes.

data platform

Software that collects data from a business's systems and documents, cleans and joins it, and serves one consistent version to reports, people and AI.

data quality

Whether data is fit for its use. A real seasonal drop is correct data, and rules that flag it as an error teach people to ignore alerts.

data warehouse

A database built for analysis: it copies data from other systems, keeps history on one set of definitions, and must decide what to do with late records.

deterministic calculation

A calculation that gives the same answer from the same inputs every time. Reported totals need it, and a written rounding rule to go with it.

digital forestry

Using digital data and tools across forest management: remote sensing, machine data, models and records. The 2005 definition put integration at its core.

digital twin

A virtual model of a real asset kept current with live data, able to test changes before they are made. Published forestry twins are still early.

document extraction

Reading documents such as contracts, tickets and invoices and turning the facts inside them into structured records, each linked back to its page.

E

encryption at rest

Encrypting stored data so that disks, backups or files are unreadable without the key. It protects against stolen storage, not against a user with access.

ERP

Software that runs a business's core transactions, orders to invoices to ledger, on one database. It records the business well and explains it less well.

ETL

Extract, transform, load: taking data out of source systems, converting it to agreed definitions and loading it. The middle step sets meaning.

F

forecasting model

A method that predicts a future value from past data and known drivers. Judge it by its error on data it has never seen, against a naive forecast.

G

GIS

A system that stores, analyses and maps data tied to places. In timber, its value depends on shared identifiers and matching coordinate systems.

H

human-in-the-loop review

Building a person into an AI process where judgement or accountability matters. Routing by confidence helps; a random sample checks what it lets through.

K

KPI

A key performance indicator: one of the few measures a business watches to judge whether it is on track. A single KPI can hide what its parts show.

L

least-privilege access

Giving each person, system and AI tool only the access its job needs, and no more. With AI agents acting on records, it matters more than before.

LiDAR

Light detection and ranging: measuring distance with laser pulses to build 3D models of a forest. Its accuracy depends on what it is checked against.

M

machine learning

Software that learns patterns from past data to make predictions about new data. How many past records are usable often matters more than the algorithm.

master data

The shared records the business runs on, such as customers, suppliers, products, locations and accounts. They also answer questions of their own.

MCP

The Model Context Protocol, an open standard for connecting AI applications to other systems. Its spec sets consent rules; servers vary.

model training on your data

When an AI provider uses what you send it to improve its models. Commercial terms often forbid it; free tools may allow it. Check the contract.

N

natural-language query

Asking a data system a question in plain words instead of code. It works on tidy test databases; complex company warehouses are much harder.

O

OCR

Optical character recognition turns images of text into machine-readable text. Print reads well; handwriting is hard for machines and people alike.

optimisation solver

The engine that finds the best answer to a formulated planning problem. Buyers meet it inside scheduling tools; judge the model, the gap and the speed.

P

predictive maintenance

Maintaining equipment when its measured condition says it is needed, before it fails. The first question is which machine's failure costs most.

primary key

The value that uniquely identifies each record, such as a ticket or tag number. Every join and reconciliation needs keys to be unique and stable.

R

reconciliation

Checking that two records of the same thing agree, and explaining every difference. In timber the records are tickets, contracts, invoices and the ledger.

remote sensing

Measuring land from a distance, by satellite, aircraft or drone. Volume and biomass estimates rely on field plots, and how plots are measured matters.

retrieval-augmented generation

A way of making an AI model answer from documents it retrieves when asked. It reduces made-up answers; a 2024 test found it doesn't end them.

role-based access

Granting access by job role rather than person by person. Roles drift over time, and AI assistants should usually act with the asking user's access.

S

schema

The formal structure of a database: its tables, fields, types and links. It says what is stored, not what it means, and that gap costs AI accuracy.

semantic layer

The shared definitions that sit between raw data and reports, so that margin, cost per thousand board feet and volume mean the same in every tool.

single source of truth

One agreed, referenced place for each fact, so every report reads the same value. In timber it usually means one owner per fact, not one system.

T

tenant isolation

How a shared software service keeps each customer's data walled off from every other customer's. Ask what isolation covers, including AI, and for evidence.

U

unit-of-measure normalisation

Converting quantities into one agreed unit, with the rule written down. In timber, the conversion factor itself can vary with tree size.

W

WMS

Software that records where stock is and how it moves: receiving, putaway, picking, counts, shipping. Its records age unless counts correct them.