Services

Five fixed steps, from first use case to production

Fuchs & Agents offers five fixed-scope AI services: an Executive AI Diagnostic, an Architecture Sprint, a Prototype Sprint, a Production Readiness Review and a Fractional AI Lead. Each has a written scope and ends at a decision point.

Every engagement starts small, has an hours cap, and you choose whether to take the next step.

Consulting

From the first use case to production

Most of our work is for businesses where sensitive data, quality and operating cost all matter: financial firms, professional services, software companies and other regulated teams. Smaller businesses usually start with the Fridays back offer instead. If you are a regulated team, start here.

Executive AI Diagnostic

A short look at your use cases, data, architecture, governance and operating model.

You get a ranked opportunity map, a recommended architecture and a next-step roadmap with the unit costs worked out.

See it on fictional data: cost per approved document, governance policy builder.
Architecture Sprint

Reference architecture, deployment model, model choice, agent boundaries, infrastructure and security.

You get an implementation plan and a cost model: capex against opex, break-even and sensitivity.

See it on fictional data: model comparison scorecard, RFP compliance matrix.
Prototype Sprint

A working prototype built around one valuable workflow, on your own material.

You get a demo plus before-and-after measurements: reviewer minutes per draft, edit rate, turnaround.

See it on fictional data: receipt and invoice extraction.
Production Readiness Review

Security, monitoring, cost, model updates, logging, scale, human approvals and failure modes.

You get a readiness scorecard and a list of fixes.

See it on fictional data: isolation leak test, approval log viewer.

Fractional AI lead

An AI lead, a few days a month

For teams that need senior AI judgment before they are ready for a full-time Chief AI Officer or CIO hire. Typical work: reviewing vendor proposals, setting architecture rules, sitting in on product and customer calls, and keeping the AI roadmap tied to the business plan.

Two levels

Advisory: a set number of hours each month for reviews, calls and decisions.
Embedded: roughly a day a week working alongside your team.

Three-month minimum. [PLACEHOLDER: monthly fees and hour allowances, from rates/ v2 once Joe approves them for publication]

Local-first private AI

Keep the sensitive work in-house

Most organizations don't have to choose between cloud AI and private AI. They need a rule for which work goes where. We set up that rule and the private side of it, using FoxGrade, our private AI deployment pattern, on hardware you own, inside your own cloud account, or both.

Workload placement: what stays on your existing cloud tools and what moves to the private lane
Hardware sizing and cost per document, measured before you buy anything
Agents with scoped permissions, approval steps and an audit trail
Separate workspaces per client or matter, with leak testing

For the plain version of where data lives and who can see it, read how we handle your data.

How engagements work

Fixed scope, fixed fee, a decision at each gate

Engagement ladder
Step	Typical length	How it's priced	Fee
Executive AI Diagnostic	1 to 2 weeks	Fixed fee, hours cap	[PRICE TBD]
Architecture Sprint	2 to 3 weeks	Fixed fee, hours cap	[PRICE TBD]
Prototype Sprint	4 to 8 weeks	Fixed fee, paid by milestone	[PRICE TBD]
Production Readiness Review	1 to 2 weeks	Fixed fee, hours cap	[PRICE TBD]
Fractional AI lead	Monthly, 3-month minimum	Monthly retainer	[PRICE TBD]

Ground rules

The Diagnostic is the entry point. Nothing bigger is sold before it.
New use cases, data sources or deliverables go through a written change order before work starts.
You own the deliverables once they're paid for.
A mutual NDA is signed before the Diagnostic. Your data stays in your environment by default, and it is never used to train models.
Hardware, GPUs and model licences dedicated to you are bought by you or billed at cost.

[PLACEHOLDER: terms above are the proposed defaults from rates/ v2; confirm with counsel before publishing]

Ask about a Diagnostic

When we're not the right call

Who should not hire us

Your AI work is general office work. If no client or proprietary material is involved, the AI tools you already license are probably enough.
Your volume is low and your data is allowed to leave your network. At low volume, metered cloud AI is often cheaper on raw token cost, and we'll say so. The worked example.
You want a large program sold up front. We start with the Diagnostic and a decision gate. Nothing bigger is sold before it.
You want AI to act on its own. Every output here has a named human approver.