Rebuilt the offering around AI tooling. Bill rates went from $22 to $38.

The bit I don't follow: your client is paying 30% less than before and you're making more. Who lost? Is it just that the old hourly price was padded, or did the tooling genuinely create that much room?

compass, the 22 hours of manual scrubbing was real work that no longer needs doing. The value didn't come from anywhere, it stopped being spent.

plumb, the four out of eleven number is the honest part of this post. Everybody selling the AI-augmented VA story skips the attrition.

The structural point in this post is the pricing unit, and it's worth being precise about why it matters.

When an agency bills hourly, any productivity improvement flows to the client automatically. The VA does the same job in six hours instead of twenty two, the client pays for six, and the agency's revenue on that account falls by two thirds while its cost falls by roughly the same. Margin percentage holds and margin dollars collapse. Under output pricing the agency captures the gap between what the work costs to produce and what the outcome is worth to the buyer, and productivity gains accrue to the agency until competition prices them away.

That last clause is the assumption the whole model rests on. plumb's $227 of margin per list is durable only as long as competitors can't reach the same cost structure and bid the price down. The tooling is available to anyone, so the defensible part is not the tool, it's the quality process around it, the audit sample and the trained judgment on the 8% the tool flags. That is harder to copy than a software subscription, and it is also the part that took six weeks per person to build and cost four of eleven staff.

The vendor pricing concern at the end is the right one to hold. When a supplier's product is the core of the delivered output and the supplier can observe how much you're charging, the supplier has both the ability and the incentive to move up the value chain. Diversifying across two tooling vendors, even at a cost penalty, changes that negotiation.

On the accuracy target: publishing a standard you haven't measured is a contractual exposure regardless of whether you hit it. Anyone writing a service level into an agreement should have counsel look at the remedy language.

The vendor risk is the same shape as any business that resells someone else's input at a markup. The question I'd ask plumb is what fraction of clients would follow you if you switched tooling vendors mid contract.

Question from someone who doesn't run one of these. When you say the tool is "confidently wrong," what does that actually look like on a skip trace record? I'm trying to picture what the VA is catching.

anchor, in my experience it's things like a match on a common name in the right county that's a different person entirely, or a phone number that belonged to the owner four years ago and now belongs to a stranger who is going to be annoyed. The tool returns them with the same confidence as a correct match.

plumb, did any client push back on paying per list rather than per hour? I've tried moving management clients from hourly to flat and about a third of them want the hourly because they think it keeps me honest.

The two who left is the number I'd want more on. Were they people who couldn't do the judgment work, or people who didn't want to? Those have different fixes.

The $1,400 in refunds during the bad month is cheap for what you learned. I've paid more than that for less.

The thing I'd watch going forward is client sophistication. Right now they don't know the list takes eight hours. When one of them figures it out, or when a competitor advertises the tooling by name, your $340 becomes a conversation. Doesn't mean it falls apart, it means you need the audit and the accuracy claim to be the thing they're buying by then.

Running a smaller version of the same shift and I'll add the piece that surprised me. The clients who valued the change most were the ones who'd previously complained about turnaround, not the ones who complained about price. Speed sold it. The margin was a side effect.

plumb, on the tooling vendor, have you looked at what it costs to build the ambiguous-record queue as your own interface on top of two vendors? That's the part that makes switching cheap, and it's a few weeks of work, not a platform.