Three phases, four named builds, three retainer tiers. Fixed-fee where it matters, hourly where it doesn’t, and the same engineer across all of it. Every engagement is anchored on Hermes — our sovereign LLM foundation — so your data never leaves your walls.
Most AI projects fail in scoping, not engineering. The Assessment is the engineering of scope: stakeholder interviews, system inventory, a scorecard that ranks every candidate use case, a sovereignty analysis, and a 12-month roadmap with real TCO. We don’t sell follow-on work from this; we earn it.
The standard engagement — a CPA-credentialed AI engineer embedded with your finance and ops leads for two to three weeks. Output is a board-ready plan you can actually fund.
Refundable against any subsequent Build engagement. If we don’t think AI fits — we’ll say so and bill nothing.
1:1s with finance, ops, IT and ERP admins; a complete map of what’s running, what it costs, and what it doesn’t do.
Every candidate workflow scored on feasibility × ROI × sovereign fit. Greenlit, deferred, killed — with reasons.
What’s clean, what’s a graveyard, where the joins fail, and which integrations need to be built before any model goes in.
What can run on-prem on Hermes; what’s safe to send to a frontier model; what your contracts and your compliance team will sign.
Sequenced phases, total cost of ownership including infra and labor, and a recommended Build package (or “don’t build yet, here’s why”).
Each Build engagement is anchored on the Foundation — managed Hermes, retrieval, the messaging gateway, observability via Langfuse, and team training. The three vertical packages stack on top, integrating with the ERP you already run. Hourly out-of-scope rate is the only soft number; everything else is contracted.
The substrate every other engagement extends. A managed Hermes instance, retrieval pipeline, messaging gateway, evals, and team enablement — sovereign by default.
Foundation + the AP/AR workflows that eat your controller’s week. Vic.ai or Stampli on the inbound side, BC or NetSuite on the outbound, and reconciliation in the middle.
Foundation + n8n agent workflows wired into your ERP for asset, maintenance, and field-ops automation. The deepest integration footprint of the four packages.
Foundation + a Retell or Vapi voice agent plus chat surfaces (Fin or Hermes) wired into your CRM and helpdesk. The fastest visible-to-customer engagement we run.
Hermes hosting passed through at cost · hourly out-of-scope billed at published rate · see Pricing for ranges.
An AI system that ships without operations decays — model drift, prompt rot, integration breakage, and the slow accretion of edge cases nobody updated the eval for. Three tiers, scoped by hours and embedded depth. Most Build clients land on Standard.
The baseline. Keeps your Hermes deployment running, monitored, and audited. Right for businesses with a stable workflow and no roadmap pressure.
The default after a Build engagement. Active roadmap work, regression management, and a new workflow or two each quarter as the business changes.
Embedded fractional AI lead. We’re in your standups, your roadmap reviews, and your vendor selection calls. Continuous roadmap execution.
For businesses that aren’t ready to build yet, but need a real AI voice in the room. Monthly retainer, no implementation, no agency-style upsell. The model is “the Director of AI you can’t justify hiring, on a few days a month.”
Talk to Doyle →