GLOSSARY OF TERMSTerminology demystified
A lot of what we do at Sightline is dig around in the weeds behind AI systems, looking at things the product designers knew were critical but hard to solve at scale. It's unglamorous ground — but these hidden challenges are the difference between people trusting an expensive AI system and quietly carrying on as they were. This page covers the core terms we keep using.
Data.
Is the information feeding an AI product sound, traceable, and properly sourced?
Quality
HR data can be complex, and information on work itself can be thin. It can also be complete, current, well formatted, and plain wrong. We look across completeness, accuracy, timeliness, and validity. Solving each is different, and how we help put it right draws on multiple techincal partners and analyst skills.
Lineage
Ask an enterprise leader where AI's answers come from, and the trail can be hard to follow, because the same people, skills, or tasks often sit in multiple systems, with many definitions and disagreements. We support clients to trace data back to its source, surface where duplicates collide, and define which is authoritative, so you can defend an output.
Provenance
Lineage is the path data has travelled. Provenance is where it began, and whether that beginning can be trusted. A figure from the HRIS and a number typed manually into a spreadsheet may look identical in the field, yet they do not carry the same authority. We map and register the data sources, so your AI draws from a trusted system of record.
Context.
Does the system understand how your organisation actually works, decides, and describes itself?
Workflow Context
Many automated and agentic products ship with assumed users and assumed business workflows: who acts, what triggers each step , where people are in the loop. Real organisations rarely match those assumptions.
We understand the workflow a product is built to expect, then use experience and process-design methods to find where humans and the product fall out of step.
Work Ontologies
Most workforce and talent AI products reason over vocabularies of tasks, skills and roles, which should match across systems, at a designed or chosen level of fidelity.
We work with specialist partners where needed to simplify data structures and support the mapping and inference of people, agents and work (demand).
Knowledge Catalogues
Many AI features, like HR agents, are only as good as the corpus they pull from. In reality, this material inside an enterprise is often scattered, contradictory and out of date, so a confident answer might come from the wrong source.
Data platform and governance tools exist in most IT organisations. We bridge the world between HR and IT, ensuring workforce products retrieve from well-catalogued information, flagging and removing what might lead an answer in the wrong direction
Governance.
Can you show the system is operating within the rules — to regulators, executives, and the people it affects?
Regulatory Risk
The EU AI Act classes most decisions about hiring, promotion and work-allocation as high-risk, which brings obligations aroudn documentation, transparency and human-oversight (for good reasons).
We work out which obligations your specific deployment triggers and what evidence each one needs, then build that evidence as you go. The alternative is assembling it under audit.
Executive Oversight
When automated systems make decisions about people, accountability sits across several leadership roles. Today, that ownership is often opaque or loosely defined.
We support clients to map internal and external accountability for each layer of AI-enabled decisions, from source data, to model, learning loops and user-level output. This is a first response to emerging regulatory and board oversight expectations.
Worker Voice
Involving the people affected by automated systems is increasingly both a legal requirement and a source of better design. Works councils and unions can hold co-determination rights over AI that touches personnel decisions, and rollouts that skip them face risk. The people closest to the work are also the best guide to whether a system fits how the work is really done. We help clients build that involvement into a source of strength rather than a late obstacle.
