How We Work
Small scope, real deadlines, and a paper trail
The engagement model
- Start with a conversation or an assessment. Most firms start with a workflow assessment. If you already know the workflow, we scope it in a single call.
- Hourly, time and materials. Senior US engineers, billed by the hour, as a project engagement or as staff augmentation inside your team. Before work starts you get a written scope note: what the agent does, what it does not do, which systems it touches, and what "done" means. You see the hours as they accrue, and changes are agreed in writing.
- Small builds, real deadlines. One senior engineer owns the build; a second reviews it. You see working software in the second week, not a slide in the fourth. A first build typically runs four to six weeks, though your systems and data set the pace.
- Parallel run. The agent runs beside the current process for at least two weeks before anyone relies on it.
- Handover and operations. Documentation an operations lead can follow, a runbook for the failure modes we expect, and ongoing support for API and model changes on the same hourly basis, so the agent keeps working when your ATS changes an API.
Who does the work
Senior US-based engineers who have run production systems for years. The person who scopes the work writes the code and answers the phone during parallel running. There is no offshore hand-off and no junior bench.
What you get at handover
- The source code, in your repository or ours with you as owner
- The rubric, rules and prompts as versioned documents you can read
- An operations runbook: what the agent does each day, what it alerts on, how to pause it
- Integration credentials issued to a named service user in each system, so access can be revoked without touching anything else
Security posture
- Agents run in a dedicated cloud environment per client; no shared tenancy of your data with another firm
- Personal data is read from your systems of record and held only as long as the workflow needs it; retention is set per agent and documented
- Model providers are used under agreements that exclude training on your data; we can run on a provider you already have a contract with
- Every write to your ATS or back-office system is made by a named integration user and logged
- Secrets live in a managed secrets store, never in code or configuration files
Compliance posture
We build for the regulatory position staffing firms are actually in: NYC Local Law 144 and Illinois HB 3773 place the duty on the employer or agency using an automated tool, and Colorado's AI Act adds developer and deployer duties on a separate timetable. That is why every agent that touches a hiring decision logs its inputs, its rationale and the human who acted, and why we produce audit exports rather than promising a tool is "compliant". Your counsel decides what satisfies each jurisdiction; we make sure the records exist.
What we will not do
- Build autonomous outreach or rejection with no human in the loop
- Claim partner or certification status we do not hold
- Sell a workflow that a feature already inside your ATS handles well enough