Glossary. The words we use, defined plainly.
Every term on this site, in the sense we mean it. If a definition here differs from a vendor's, ours is the one we build to.
Process maturity level
A rating of how reliably a business process runs and how much of it can safely be automated. Cantera Works uses five: Below L1 (undocumented), L1 Predictable, L2 Automated, L3 Agentic and L4 Looped. You can't skip a level.
L1: Predictable
A process that's standardized and written down so that a new person could follow it and two people produce the same result. No automation yet. The prerequisite for everything above it.
L2: Automated
A process where deterministic rules handle the routine steps on systems the company already owns. Anything outside the rules halts loudly for a person.
L3: Agentic
A process where an AI agent handles real variance, like messy documents or open-ended messages, within a bounded scope. Every decision type has a confidence threshold and a named human owner.
L4: Looped
A process where the system measures its own output and surfaces recurring exceptions. The people who own it adjust thresholds and rules in a regular review.
Agentic workflow
A workflow in which an AI model decides among several actions to complete a multi-step task, rather than following a fixed script. Safe agentic workflows start read-only and earn autonomy with a record of being right.
Confidence threshold
The score below which an automated decision stops and asks a person. Set per decision type and approved by the client before anything is built.
Fail closed
A design rule: when a step can't complete, the system halts and raises rather than guessing or partially succeeding. Silent partial success is the expensive failure mode.
Posting service
The single component allowed to write to a system of record. Agents propose; the posting service executes; every write path has a tested rollback. One place to audit, one place to stop.
Runbook
The document that lets someone who wasn't in the build run, break and repair a system. Written as the system is built and reviewed by a person outside the build.
Handover
The scheduled phase at the end of an engagement. The client's own people are trained to run the system, documentation is delivered, and the consultant's access ends on a date in the contract.
Survey
Cantera Works' three-week, fixed-fee diagnostic: process maps, a systems audit, maturity scoring and a costed roadmap. Ends with a decision you can act on.
Pilot
One workflow, built and measured over four to six weeks against acceptance criteria the client sets first. It ends in a go or no-go with the reasoning shown.
Deterministic automation
Automation whose output is fully determined by its inputs and rules, the same input always produces the same result. Cheaper, faster and easier to audit than a model; the default where variance is low.
Human in the loop
A design where a person reviews or approves an automated decision before it takes effect. In Cantera Works engagements, where AI is in the loop a named human is in the loop above it.
Thirty minutes. Worst case, you lose half an hour.
Bring one process. You leave knowing which level it's on, what to fix first, and whether it's worth doing anything at all.