What AI Agents Do While You Sleep: Two Setups From Our Clients
One client wakes up to a plan for more sales. Another gets fresh leads in the CRM every day. How both setups work, and what they cost to run.
One of our clients has a whole team of AI agents. When he goes to bed, they start working: they hand tasks to each other, go through the CRM and the sales numbers, look over the website. In the morning he opens his chat and there's a plan waiting. What to change, and how, to turn more visitors into customers.
He didn't hire a night shift, and nobody stays up to watch dashboards. The work that used to wait for "when I have a free afternoon" now happens while the office is dark.
Below are two setups we've built for clients, how they work, and the part people usually skip: what goes wrong and what it costs.
Setup one: a team of agents that reviews the business overnight
Most owners know their CRM holds answers. Which deals stall, where leads go cold, which pages on the site people leave from. The problem is that digging through it takes hours, and those hours never show up in the calendar.
In this client's setup the agents split the job between them. One goes through the CRM and sales. Another checks the website. They pass findings and new tasks to each other, the same way people on a team would say "take a look at this, it might be connected." By morning it all comes together in one message to the director: a short plan for raising conversion.
The important word there is plan. The agents don't change the site at night and they don't email customers. They prepare. The director reads it over coffee and decides what actually gets done.
We wrote about the CRM side of this in more detail in How AI Agents Can Find Where You Are Losing Customers. The overnight team is that same idea, running every night instead of once.
Setup two: new leads in the CRM every morning
Another client needed a steady flow of leads that fit specific criteria. Done by hand, that means someone sits in a lead database, filters, checks contacts and copies the good ones into the CRM. It's the kind of job everyone agrees is important and nobody wants to do on a Tuesday.
We connected an agent by API to a service that searches and validates contacts, and set up tasks on a schedule. Now the agent goes through the service every day, picks the leads that match the client's criteria and saves them to the client's CRM. The sales team starts the day with a list instead of a search.
Nothing magic happens here. The agent does exactly what a careful assistant would do, the same way every day, and it doesn't skip Fridays.
And sometimes it doesn't wait for the night
Sergey, our co-founder, had a version of this happen in a taxi. All he had with him was his phone, and on it his own agent, Jarvis. Stuck in traffic, they talked through a strategy for finding new leads. While Sergey was in his meeting, Jarvis found the first ones and set the whole process up to run on its own. When the meeting ended, all Sergey had to do was read the plan and approve it.
That's the pattern in all three cases. The agent does the legwork. A person makes the call.
What we're honest about
The plan is only as good as the data. If deals in your CRM aren't updated, or half the leads have no source, the overnight analysis will be built on that. Agents are good at finding patterns. They can't find patterns in data that isn't there.
Agents need access, so access has to be planned. To review a CRM, the agent has to read it. We start with read-only access and add more only when there's a reason. Nothing gets sent to customers without a person saying yes.
It costs money every month, not just once. There's setup work, then hosting, AI usage (tokens) and in the lead case the contact service itself, which usually has its own subscription. On our side, a hosted agent setup typically runs from $99 to $799 a month plus a one-time setup, depending on how much it does. We've also learned the hard way that an agent can burn tokens doing nothing useful: We Spent $30+ on an AI Agent Doing Nothing. Limits and monitoring are part of every setup now.
Someone still has to read the plan. If the morning message goes unread for two weeks, the agents are working for nobody. This works best for owners who already look at their numbers and just never have time to do the digging.
Where to start if you want this
- Pick the report you keep postponing. The weekly look at sales, the "why did leads drop this month" question. That's your first overnight task.
- Check what's in your CRM. Open the last 20 deals. If the stages, sources and notes are mostly filled in, an agent can work with it. If not, fix that first. It's cheaper than any agent.
- Write down what a good morning message would look like. Three lines? A table? A list of leads with a reason for each? Knowing what you want to read is half of the setup.
If you're not sure which process should go first, this article walks through it step by step: How to Identify Where AI Actually Belongs in Your Business.
Want to see what this would look like for you?
Book a free 30-minute discovery call at evolutionai.biz/contact. Tell us which job you'd hand to a night shift if you had one, and we'll tell you honestly whether an agent is the right tool for it, or whether a simpler fix will do.