Jon CursiAsk an AI company what their product does and you get adjectives. Ask what it did last month and the room goes quiet.
So here is the second answer. Two clients, two very different agents, one month of actual output. No projections, no "up to" numbers.
Client one: 196 conversations nobody had to answer
Making Waves Swim School runs classes, camps, and private lessons. Before their agent, every inquiry funneled to one place: the owner's phone and inbox. Questions came in at 9pm, during classes, on weekends. Many of them never got a reply.
Their customer-facing agent's first 30 days:
- 196 conversations handled start to finish
- 13 booking-link clicks driven directly from chat
- 32+ hours of staff time saved on repeat questions
- An estimated $1,000 to $6,000 in new revenue from leads that would have gone unanswered
The revenue number is the one to sit with. Those weren't new leads. They were leads the business was already generating and then losing, because someone landed on the site, asked about pricing or schedules, got silence, and moved on. The agent didn't create demand. It stopped throwing it away.
Client two: a full marketing operation for a company that had none
Boxwood Home Construction is the opposite problem. Great work, steady referrals, and essentially zero digital presence. The company had no real web presence and nobody whose job it was to fix that.
Their Internal AI agent now runs, month after month:
- Full website creation and management from scratch
- Blog publishing twice per week, every week
- A social content pipeline across multiple platforms
- Monthly website audits for SEO and performance
- Estimate generation to speed up quoting
Staffing this the traditional way means a marketing person, a web developer, and probably a content writer. For a company Boxwood's size, that's $5,000 to $10,000 per month in salaries or contractor fees. One agent replaced the whole stack, and the work actually ships on schedule, which is more than most small marketing teams can say.
The common thread
A swim school and a construction company don't look alike on paper. Their problem was identical: important work that nobody had capacity to do.
For Making Waves, that work was answering inquiries fast enough to win them. For Boxwood, it was marketing that had been "on the list" for years. In both cases the work wasn't hard. It was constant. And constant work is exactly what falls off the plate when your team is busy delivering the thing customers actually pay for.
That's the honest pitch for AI agents. Capacity matters more than intelligence. An agent doesn't get pulled into a crisis, doesn't skip the blog post because a project ran long, and doesn't leave the 9pm inquiry until morning.
How to read numbers like these
A few things worth being straight about.
First, your numbers will not match these. Conversation volume depends on your traffic. Revenue recovered depends on your close rate and ticket size. What transfers is the mechanism: instant responses convert leads that silence loses, and recurring work done on schedule compounds where sporadic work doesn't.
Second, month one is the floor, not the ceiling. Boxwood's blog and SEO work builds over time. Making Waves' agent gets better as it learns which questions come up and what answers move people to book.
Third, compare against the real alternative. For most teams the alternative is a hire, and the hiring math includes salary, benefits, ramp time, and the risk they leave in a year. An agent is a flat monthly cost that works every day it's live.
The pattern applies past small business
These two examples are small companies because small companies feel the capacity gap first. But the gap doesn't disappear at 50 or 500 employees. It just changes shape: reports rebuilt every week, engineering maintenance postponed for quarters, handoffs where work dies between teams.
Bigger teams have specialists. They still have more recurring work than the specialists can absorb, and the overflow either lands on senior people or doesn't happen at all. Same problem, larger dollar figures attached.
Run the same exercise on your business
Here's the useful version of this post. Take thirty seconds and answer two questions:
- What inquiries or requests go unanswered, or answered slowly, because nobody is free when they arrive?
- What recurring work has been "on the list" for more than three months?
Whatever you just thought of is your version of the 196 conversations or the twice-weekly blog. That's where an agent earns its cost, usually in the first month.
If you want to see what that looks like with your actual workflows instead of someone else's case study, book a live demo. Bring the two answers above. We'll map an agent to them on the call.