About

We Run on What
We Sell

Service Robotics is a builder-led consultancy. Our own operations run on a group of AI agents with a human holding the decisions, using the same architecture we install for clients, with the same controls and the same bills arriving every month.

Our Story

Service Robotics started as an experiment. Could a small company actually be operated by AI agents? Not agents that draft a marketing post now and then, but agents holding real roles, doing recurring operational work, spending money that comes out of the bank.

It turned out to be possible, though for reasons nobody expects. Getting an agent to produce good output is the easy half, and it's the half every demo shows you. What decides whether a system survives is duller stuff: knowing when the output can be trusted, deciding what a machine may do without asking first, capping what it can spend before it spends it, and being able to switch the whole thing off in a hurry.

We built that layer for ourselves out of necessity. Then we watched businesses around us run pilot after pilot that died in exactly the places ours would have, had we skipped it. That gap between a system that demos and one that runs is what we sell now.

None of the argument we make to clients is theoretical. The approval queue, the budget caps, the stop switch, the model routing: we lean on all of it daily. When we tell you what production asks of you, it's because we've paid for the version that skipped it.

Who You'll Work With

Kyle Rice has spent his 20 year career updating systems and improving workflows. With the advent of useful AI systems entire industries are going to need to change the way they do business.

Every engagement is delivered by the person you speak to on that first call. There's no account manager, no handoff to a delivery team, and no junior consultant learning the trade on your budget.

That puts a hard limit on how much work we take at once, and it's why the pricing on this site can stay fixed. The person scoping the job is the person doing it.

What We Do

We build AI and agent systems that reach production, rescue the ones that didn't get there, and run them for you afterwards if you'd rather not staff that job yourself.

What We Don't Do

No financial or CFO services. No general AI strategy decks. Where the honest answer to your problem is a spreadsheet or something you can buy off the shelf, we'll say so rather than sell you a build.

How We Work

  • Published Pricing Every engagement price sits on this site instead of behind a call
  • Fixed Scope The price is agreed before work starts, and we don't bill by the hour
  • You Own It Systems, documentation and runbooks are yours to take in-house
The stack

The Agent Org That Runs This Company

Ours, not yours. It's the reference build for what we install for clients.

ROLES

Scoped Agent Roles

Defined responsibility, limited permissions

Agents hold specific roles with explicit remits and only the tool access their job calls for, rather than one general assistant holding the keys to everything. Financial systems are read-only to all of them.

GATES

Owner Approval Queue

The system drafts, a person decides

Anything heading out of the building waits in a queue for human approval. Published content, outbound messages, commitments of any kind. Nothing gets transmitted on its own initiative.

LIMITS

Budget Caps & Routing

Predictive gating and a stop switch

Daily spend ceilings that block a run before it breaches rather than reporting it afterwards, model routing per job so cheap work runs on cheap models, and a switch that halts everything.

Why this matters to you: these are the pieces most pilots are missing, and they're why those pilots stall at the moment someone has to trust the output or sign off on the bill. We install this layer as part of a build rather than selling it back to you later.

What we learned

Running It Ourselves Taught Us Four Things

All of them the expensive way.

Evaluation Comes First

Where you can't tell a good output from a confidently wrong one, what you have isn't a system. It's a thing people quietly stop using. We build the measurement before the feature.

Autonomy Is a Dial

The question is never whether it can act on its own. It's which specific actions, under which conditions, reviewed by whom. Getting that wrong is what makes agents frightening instead of useful.

Cost Is a Design Decision

Which model handles which job moves the bill by an order of magnitude. Routing and caps belong in the architecture from day one, well before anyone runs a cost-cutting exercise six months in.

Boring Beats Clever

Systems that survive have dull, legible failure modes. We build for something your team can reason about at 4pm on a bad day.

Next step

See Whether This Fits

A free half hour, and a straight answer on whether an assessment is worth your money. If it isn't, that's a fine result and we'll tell you so.

Book a Free Consultation