Service
AI implementation for operations teams
AI belongs in an operations workflow only where it does a specific job better than the alternative, and a person signs off on anything that matters. We find that job in your work, build it into a tool your crew already uses, and leave AI out where it does not earn its place.
The problem
Owners are being told to do something with AI, and the usual results are a chat window nobody opens or a pilot that never reaches the crew. The useful work is narrower: a service report drafted from the technician's notes and photos, a quote checked against the last ten like it, a pile of job tickets sorted by what is missing. Each is one step in a workflow, with a person checking the result.
Where AI belongs, and where it does not
AI earns its place where a step is mostly reading, sorting or drafting, the volume is high, and a person can check the result quickly. It does not belong where a mistake is unsafe or expensive and nobody would catch it. In the Workflow Spec we decide step by step, and the answer is often a tool with little or no AI in it.
Human sign-off
When a decision matters, we use the Council Method from Jason Santiago's book, Let the AI Be Smart: Several AI models answer the same question separately, so none can copy another. Each then sees the others' answers and must defend or correct its own, backing every claim with a source you can open, because agreement alone is not proof. One model writes a verdict that keeps only what survived, and a person signs it.
Your data
Your data stays yours. Meeting audio is transcribed by OpenAI and specifications are drafted with Anthropic, both through their business APIs, which do not train models on API data. Both may keep API data for up to 30 days for abuse monitoring, and zero-data-retention terms are not yet in place with either. More in where does our data go.
Scope and price
A Workflow Spec at $15,000 fixed, then a Production Build from $60,000, fixed after the Spec, the same as any other engagement. The last 20 percent of the Build fee depends on the crew using the tool.
What we decline
Requests to use AI everywhere without a specific workflow, tools that would act outside the company on their own, unapproved money movement, employment decisions, and binding customer commitments without a human approval step.
Questions people ask
Do you put AI in every tool?
No. AI goes into a tool only where it earns its place. It is a way of delivering the tool, not the product, and many of the most useful workflow tools need little or none of it.
Is our data used to train AI models?
No. Where client data reaches an AI model it goes through our own platform to providers named in your contract, using their business APIs, which do not train models on that data. Both providers we use may keep API data for up to 30 days for abuse monitoring; zero-data-retention terms are not in place with either.
Can the AI act on its own?
Not where it counts. We decline tools that act outside the company on their own, move money without approval, make employment decisions or commit to a customer without a human approval step.
Next step
Talk to us
Start with a free twenty-minute fit call. If the workflow, the sponsor and the path to access are real, the next step is a Workflow Spec.
Fox & Santiago, Genesee, Idaho. Home territory: the Inland Northwest, where we respond fastest. We fly anywhere in the United States for paid engagements.