{"canonical_url":"https://engineer.company/ai-agents/","content_markdown":"# Custom AI agents built for your trade\n\nBuild an AI agent around one useful workflow: the systems it connects to, the\ncases it is evaluated against, the actions a person approves and the access it\nreceives. We agree that scope with your domain experts and choose a model using\nthose requirements. The integration and evaluation record remain part of your\ndeliverable when the model changes.\n\n## Connected: to the systems the work lives in\n\nAn agent that cannot reach your systems can only talk. We connect it to the ones\nthe work happens in — the ERP, the CRM, the booking system, the mailbox, the\ndocument store — each through a described tool with a narrow job: look up a\nvessel, draft a quote, file a case. Tools are offered over MCP or an\n[API](/glossary/#api), every input is typed and [validated](/glossary/#validation),\nand an agent that needs your company's own records reads them from\n[data made ready for it](/ai-data/). Memory of past cases lives in your\n[database](/glossary/#database), not in a vendor's chat history.\n\n## Measured: against cases your experts write\n\nWithout tests, \"it seems better\" is the only measurement anyone has. We sit with\nthe people who do the work and write down real cases from your trade, each with\nthe answer a senior person would give. Every change — a prompt, a tool, a model —\nruns against all of them before it ships, and the score is kept per case over\ntime, so review can reveal regressions in the agreed cases. It is the\n[quality gate](/glossary/#quality-gate) we run on our own code, applied to an\nagent's answers.\n\n## Supervised: a person decides where it matters\n\nNot every action should be the agent's alone. We write down, with you, what it\nmay do on its own, what it drafts for a person to approve, and what it never\ntouches, and the rules are enforced in the tools rather than asked for in the\nprompt. A review screen shows a person what the agent proposes and why; a\ncorrection made there becomes a new test case. Every step is traced with\n[OpenTelemetry](/glossary/#opentelemetry), so when an answer goes wrong you can\nsee which tool, which document and which model produced it.\n\n## Evaluate model changes before rollout\n\nModel choice is configuration where the integration permits it. We compare a\ncandidate against your agreed cases for answer quality, failure behavior, cost\nand latency. The evaluation informs a deployment decision; a newer model does\nnot automatically improve your workflow. Changes go through the agreed review\nand rollout process, with the previous configuration available for rollback.\n\n## The work behind it\n\nThe habits that make an agent trustworthy are the ones we already keep in our\nown engineering. We [built a guard engine of 268 commit checks](https://engineer.company/portfolio/built-a-268-check-guard-engine-for-the-repo/)\nand [a commit gate of 22 linters with no warning tier](https://engineer.company/portfolio/built-a-22-linter-commit-gate/),\nthen [wrote tests for the checkers themselves](https://engineer.company/portfolio/wrote-tests-for-the-checkers-themselves/)\nso a check that reads nothing cannot report clean. We\n[built first-party error monitoring and tracing](https://engineer.company/portfolio/built-first-party-error-monitoring-and-tracing/),\n[made six collectors state what they skipped](https://engineer.company/portfolio/made-six-collectors-state-what-they-skipped/),\nand [ran a whole development programme as 51 written initiatives](https://engineer.company/portfolio/ran-development-as-51-written-initiatives/),\neach with its measurements and the decisions it declined. This site's own\n[readiness for agents](/ai-ready/) was built the same way.\n\n## How an engagement runs\n\nIt starts with one task and its cases. We sit with the people who do it, write\ndown thirty to fifty real examples with the right answers, and agree what the\nagent may do alone. Then we build the smallest agent that passes them, behind a\nreview screen, in your infrastructure or your cloud account, with its access\n[locked to what it needs](/ai-security/). Then the next task. You own the code,\nthe cases and the data, so the agent stays yours when the model under it\nchanges. It is bought like the rest of our work, on the terms in\n[what it costs](/what-it-costs/).\n\nIf there is work in your trade that an agent should take off somebody's desk,\n[tell us what it is](/contact/) *(read by the engineer)*.\n\n\u003chttps://engineer.company/ai-agents/\u003e\n","date_modified":"2026-11-12T05:34:23Z","description":"Custom AI agents for an agreed workflow, connected to your systems and evaluated on your cases, with human review and scoped access.","formats":{"gemtext":"https://brand.engineer.company/ai-agents/index.gmi","gopher":"https://brand.engineer.company/ai-agents/gophermap","html":"https://brand.engineer.company/ai-agents/","json":"https://brand.engineer.company/ai-agents/index.json","markdown":"https://brand.engineer.company/ai-agents/index.md","pdf":"https://brand.engineer.company/ai-agents/index.pdf"},"kind":"page","language":"en","title":"Custom AI agents built for your trade","url":"https://brand.engineer.company/ai-agents/","version":1}