Glossary
Glossary.The terms we use, defined once.
Data and AI vocabulary moves fast. These are the terms used across this site, what each one means, and where Akashic puts it to work.
- AI agent
Software that uses a language model to plan and carry out steps of a task, such as retrieving data, reading documents or checking citations.
In Akashic Agents are in preview. They gather evidence and explain; a person approves every consequential step. See where
- Append-only audit log
A record of actions that can be added to but never edited or deleted, so the history of who did what stays intact.
In Akashic Always on, under every module. See where
- Bring your own model
Running each AI task on a model you choose: a commercial provider with your own key, an open model, or one hosted inside your boundary.
In Akashic Ask Akashic works with the model providers you enable. See where
- Change data capture (CDC)
Reading inserts, updates and deletes from a source database as they happen, instead of copying whole tables on a schedule.
In Akashic One of three ingestion modes in Akashic Data Pipelines, beside streaming and batch. See where
- Citation
A pointer from an answer to the records, documents or pages it was drawn from, so anyone can check it.
In Akashic Ask Akashic cites its sources on every answer. See where
- Column-level lineage
The trail from a figure back through every transformation to the source columns it came from.
In Akashic Recorded automatically, across every layer. See where
- Data contract
An agreed, versioned description of the shape and quality of the data a source will deliver, checked when the data arrives.
In Akashic Versioned in Akashic Data Pipelines; failures are quarantined, retried and backfilled. See where
- Data residency
Keeping data storage, processing and AI inference inside a chosen country or boundary.
In Akashic SaaS in your region, your own cloud account, or on-premise. See where
- Data steward
The person accountable for the quality of a set of records, who reviews proposed matches and merges.
In Akashic Stewards approve matches in the Master Data review queue. See where
- Document intelligence
Reading contracts, certificates, reports and scanned forms, extracting their facts and linking them to the entities they mention.
In Akashic PDF, Office, email and scanned intake, with page-level citations. See where
- Entity resolution
Recognising that records in different systems describe the same customer, supplier, asset or person, and resolving them into one record with the lineage of what was merged.
In Akashic Exact and fuzzy matching with survivorship rules in Akashic Master Data. See where
- Governed metric
A metric defined once, with its calculation and choices written down, and served the same way to every dashboard, model and answer.
In Akashic Defined in the semantic layer of the Data Warehouse module. See where
- Grounding
Making an AI answer draw on specific, governed sources rather than on what a model remembers, so the answer can cite them.
In Akashic Ask Akashic grounds every answer in the governed foundation. See where
- Knowledge layer
The governed layer that holds an organisation's entities, the relationships between them and the meaning of its terms, for people and AI to reason over.
In Akashic Built from Master Data's resolved entities and governed by Akashic Governance. See where
- Maker-checker
A control in which one person prepares an action and a second person approves it before it takes effect.
In Akashic Available in every plan. See where
- PII masking
Detecting personal data such as names, identity numbers and addresses, and hiding it from the people and answers not permitted to see it.
In Akashic Detected on arrival, masked in answers. See where
- Row and column security
Access rules that limit which rows and columns of a table a person may see, applied at the moment the query runs.
In Akashic Permissions follow the person from the records to the answer. See where
- Semantic layer
Shared definitions of metrics and business terms, so that “active customer” or “margin” means the same thing everywhere.
In Akashic Part of the Data Warehouse module, read by Business Intelligence, Machine Learning and Ask Akashic. See where
- Sovereign AI
AI that runs on models and infrastructure inside a country's or organisation's own boundary, under its own control.
In Akashic Models hosted inside your boundary in on-premise and private cloud deployments. See where
- Tenant isolation
Keeping each customer's data, storage, identity and network separate inside a shared service.
In Akashic Multi-tenant SaaS isolates every tenant at all four levels. See where
- Text to SQL
Turning a plain-language question into a database query.
In Akashic Ask Akashic shows the answer with its sources alongside. See where
Talk to our team
Missing a term?Ask the people who use it.
If a word on this site needs more than a definition, a solutions architect will walk you through it on your own data.