Sandy was the second agent I hired. Her job was harder to explain in one sentence because it was less about motion and more about awareness. Shelley helped work move. Sandy helped the team understand what was already happening.
A lot of organizations have the same problem: information is scattered. Some of it is in Slack, the CRM, or email. Some of it is sitting in customer channels nobody has looked at closely enough. Some of it lives only in somebody’s head because they noticed something and never got around to writing it down. By the time that context gets assembled into anything useful, the moment has often passed. Sandy’s job was to close that gap.
She focused on customer intelligence and account awareness. In practice that meant reading across the noisy surfaces where customer truth hides and turning that into something a team could actually use. Not a dashboard. Not a surveillance tool. A layer of understanding that could surface risk, opportunity, and themes that deserved attention before they got lost in the background.
That is easy to underrate if you think of it as just “monitoring,” but it is more than that. Customers don’t churn unexpectedly because you lack data. They churn because you miss the pattern while it is still early enough to intervene. Sandy’s value was making it easier to know what needed human attention.
She could notice patterns, pull together fragments from different systems, and generate briefings covering the entire customer base. The work was interpretive, not just mechanical. A lot of early AI monitoring work collapses because it floods people with technically correct but operationally useless activity. Sandy only became useful to the degree that she could be selective, restrained, and context-aware. That took work.
The challenge was not just collecting the data. It was learning what deserved a tap on the shoulder. What counts as a risk signal? What actually matters in an account thread? What sort of silence is meaningful, and what sort is just normal? Those are not trivial questions, and the answer is rarely a static rule.
That is why Sandy is such a good example of the difference between prompts and persistent agents. A prompt can summarize a thread. A persistent agent can become part of the team’s awareness model. Sandy did not matter because she could produce text. She mattered because she could sit in the background, accumulate context, and make it less likely that important customer truth would stay buried until it was too late.
That changed the shape of the work for the humans around her. They still had to decide what to do. They still had to respond with judgment, context, and empathy. But they were no longer relying entirely on memory, luck, or someone happening to check the right place at the right time. If Shelley made the front of the pipeline feel more owned, Sandy made the middle of the customer reality feel more legible. And in a small team, legibility is leverage.
On the name
Sandy is named after Sandy Olsson from Grease—the character who shows up transformed when the situation calls for it. The name fits the job. Sandy’s value comes from reading scattered, noisy signal and surfacing something useful: a risk worth acting on, a pattern worth naming, an account that needs attention before it needs rescue. Seeing what the situation calls for and showing up ready is exactly the work.