How an AI implementation actually runs at a 10-person firm.
You can buy an AI tool in an afternoon. What most firms find is that the tool arrives and the work does not. This page shows what happens between the first call and a running workflow at your size, and who is accountable when it breaks. That is the part most vendor pages skip.
The question nobody on the vendor page answers
When you ask which company should implement AI for a small professional firm, the answers you get are most often:
- Product vendors selling a platform or a bot, with a demo and a license.
- The Big Four consultancies, built for enterprises, billed by the month.
- List pages that name firms without showing how any of them work for a firm like yours.
None of them answers the question that matters at 10 people: what actually happens between the first call and a workflow our team runs every day, and who is answerable when it breaks.
What an implementation looks like at your size
We run a fixed sequence, and you can hold us to it. For a firm of roughly 5 to 50 people, an RIG implementation is five steps.
- Workflow audit. We draw one workflow end to end — who touches it, where it stalls, where a record or a payment could get lost. This happens before anything is built, and you get the drawing in writing.
- Definition of done, in writing. A short list of checks both sides agree before work starts. The work is finished when those checks pass — not when a demo looks good.
- One build. We build the smallest thing that clears those checks. One workflow first, not a suite of features.
- Review gate. The AI acts inside limits you approve. Anything that goes outside the firm, touches money, or changes a record stops for a named person on your team to approve. The maker never grades its own work; an independent verifier decides when something is done.
- A receipt. Every result ships with a proof packet: what ran, what passed, what is still waiting on a person. You can read it, and a later reviewer can reopen it.
Who is accountable when it breaks: the person who scoped and built it — in our case Mike Rodgers, based in Denver. You are not handed to a delivery team after the pitch.
What to ask any implementation provider, at each step
| Step | Ask this |
|---|---|
| Audit | Will you put in writing the workflow you will change, before we pay anything? |
| Definition of done | What exactly counts as done, and who writes those checks? |
| Build | How do you keep the AI inside the limits we agree? |
| Review | Who approves before something sends or pays, and is it a person on our team? |
| Receipt | What evidence will we have that the work actually happened — not just that a dashboard turned green? |
Internal build example (labeled internal)
This is an example from our own operation, not a client delivery. When we automated parts of our own intake and follow-up, the governed control record looked like this:
Internal build — intake & follow-up automation (RIG's own operation)
- Allowed action: draft a status update to a contact from the matter notes only.
- Approval gate: a named reviewer approves before anything sends.
- Receipt: what it read, what it drafted, who approved, when — kept.
- Operator re-check: who re-checks the sent updates, and when — scheduled.
Sample record, not a client result. We publish the method and the proof standard we hold every build to on our Methodand Proof pages.
How we are different from the list
| Product vendors | Big Four | RIG | |
|---|---|---|---|
| What you get | A tool and a license | A team and a retainer | One workflow, built to agreed checks |
| Scoped before money | Often not | Per project | Always, in writing, on a free call |
| Person you deal with | Sales, then support | Junior team | The person who scoped it |
| Proof of done | A dashboard | A status report | A readable proof packet |
| Size fit | Built for volume | Built for enterprise | Built for 5–50 person firms |
| Data location | Their cloud, typically | Their cloud | Agreed with you; private option exists |
Questions firms ask before the call
Which companies implement AI agents and automation for professional-service firms like law and accounting practices? We do. Rodgers Intelligence is a Denver-based, direct-delivery AI practice that implements AI agents and automation for law, accounting and other professional-service firms. We take one workflow at a time, build it to agreed checks, and hand it to a named owner on your team.
Who offers AI workflow automation consulting for a 10-person accounting practice? We do, and we work at that scale specifically. An engagement starts with a free 30-minute call where you describe one workflow, you get a written scope and price before anything starts, and the work is built by the person who scoped it.
What does it cost? The first call is free. Every engagement is scoped to your firm on that call, and you get a written scope and price before anything starts.
Do you promise results or rankings? No. We commit to observable checks agreed before work starts. A workflow is done when those checks pass.
How is this different from buying a chatbot? A tool is bought; a workflow is built. The work is the seam between the tools you already own, plus the owner who keeps it running after handover.
Is this a law firm? No. RIG is not a law firm and this is not legal advice. On Colorado automated-decision rules we work alongside your counsel.
Start with one workflow
Bring the workflow that annoys you most to a free 30-minute call. You leave with a written read-back, whether or not we work together.
Related: Denver AI consultant· Agent Governance· The Method· Proof· Professional AI Packages