{"canonical_url":"https://engineer.company/ai-data/","content_markdown":"# Company data ready for AI agents\n\nGive an AI workflow a dependable route to the company records it needs. We map\nthe relevant sources, clarify what each field means, connect the data and set\naccess rules with you. The first deliverable is one agreed question answered\nfrom traceable records, with checks for data quality and permissions.\n\n## Gathered: one place an agent can reach\n\nWe build the [pipelines](/glossary/#pipeline) that bring your systems together:\n[ETL](/glossary/#etl) jobs that run on a schedule, load only what changed and\nreport what they skipped, into a [database](/glossary/#database) whose\n[schema](/glossary/#schema) names your business in your words — customers,\norders, vessels, contracts, cases. Old PDFs, scans and mail archives are turned\ninto searchable text with their source kept beside them, so they can be found\nrather than only stored. The agent reaches all of it through one described door:\nan MCP server or an [API](/glossary/#api), never a copy of the database on\nsomebody's laptop.\n\n## Cleaned: one customer, not five\n\nThe same customer can be written five ways in five systems. When an agent reads conflicting versions of one fact, it needs a rule for which source to trust. We decide\nwith you which system is the source of truth for each fact, merge duplicates by\nrules you can read, and keep the merge running, so a new duplicate is caught on\nthe next load rather than at the next audit. Types and constraints in the\ndatabase refuse a bad row instead of storing it, every column says what it\nmeans, and every [migration](/glossary/#migration) is versioned and checked, so\nthe schema an agent reads is the schema that is deployed.\n\n## Permissioned: each agent sees what its user may\n\nAn agent that can read everything is a leak with a friendly voice. The rules\nbelong in the data, not in the prompt: roles in the database\n([RBAC](/glossary/#rbac)), read-only by default, row-level rules wherever one\ncustomer must never see another's, and personal data kept out of reach unless\nthe task needs it — which is also what GDPR asks. The agent acts with its user's\nprivileges, never an administrator's, and every read is\n[logged](/glossary/#logging) with the person it was made for. How we lock the\nagents themselves is [its own page](/ai-security/).\n\n## Every answer traced to its source\n\nAn answer drawn from company data is worth what its source is worth, so the\nsource travels with it. Every row keeps where it came from and when: the system,\nthe load, the file and the page. An agent can cite the record behind a figure,\nand a person can open it. A load that skipped rows says how many and why rather\nthan reporting success, and a figure that turns out wrong leads to the load that\nbrought it in, not to a week of guessing. It is the rule our own reporting is\nbuilt on: a number is never printed that the run did not measure.\n\n## The work behind it\n\nEvery claim on this page is a method we have run, and each one has its record.\nWe [made a geographical data pipeline 50 times faster](/portfolio/made-a-geographical-data-pipeline-50x-faster/)\nby fixing its SQL and its data model, and\n[designed and ran six ETL and ELT pipelines](/portfolio/designed-and-ran-6-etl-elt-pipelines/)\nacross BigQuery, MS SQL and [PostgreSQL](/glossary/#postgresql). We\n[modelled a whole trade in 348 tables across 34 schemas](/portfolio/modeled-348-tables-across-34-postgres-schemas/)\nand kept that schema honest\n[across 1,022 migrations behind a drift gate](/portfolio/1022-migrations-behind-a-schema-drift-gate/).\nWe [generated 495 pages from a read-only SQLite export](/portfolio/generated-495-pages-from-a-sqlite-export/),\nand [replaced 963 structured-data blocks with one linked graph](/portfolio/replaced-963-metadata-blocks-with-one-graph/),\nwhich is what an agent reads first on this site.\n\n## How an engagement runs\n\nIt starts with a map, made read-only: which systems hold what, who owns each,\nand which questions an agent should be able to answer. Then we take the one\nquestion that matters most and build its path end to end — the sources, the\ncleaning, the permissions and the door — with checks that stop a load when the\ndata drifts. Then the next question. Everything lives in your infrastructure and\nyour repository, through your review, and the agents that read it can be\n[built for your trade](/ai-agents/) or be the ones you already use. It is bought\nlike the rest of our work, on the terms in [what it costs](/what-it-costs/).\n\nIf the answers your people need are in more places than anyone can name,\n[tell us where your data lives](/contact/) *(read by the engineer)*.\n\n\u003chttps://engineer.company/ai-data/\u003e\n","date_modified":"2026-11-12T05:34:23Z","description":"Connect company records to an agreed AI workflow, with clear data sources, quality checks, permissions and a traceable first result.","formats":{"gemtext":"https://data.engineer.company/ai-data/index.gmi","gopher":"https://data.engineer.company/ai-data/gophermap","html":"https://data.engineer.company/ai-data/","json":"https://data.engineer.company/ai-data/index.json","markdown":"https://data.engineer.company/ai-data/index.md","pdf":"https://data.engineer.company/ai-data/index.pdf"},"kind":"page","language":"en","title":"Company data ready for AI agents","url":"https://data.engineer.company/ai-data/","version":1}