I remember creating a financial dashboard for a plastic surgery clinic. It became something the owners shared with their buddies. The numbers gave them plenty to talk about, but the dashboard wasn’t doing much to help them run the clinic.
That experience has stayed with me. A dashboard can look impressive and accurately describe a business while leaving you with all the work of figuring out what to do next.
I think we should expect more from it now. But I don’t think the answer is to replace every chart with a chat box.
Take a monthly review of accounts receivable. You want to see what’s outstanding, how long it has been sitting and whether anything has changed. A familiar chart is useful here. You shouldn’t have to start a conversation with AI just to get the same overview you need every month.
The more interesting question is what happens after you notice something.
From seeing a change to making a decision
Suppose more invoices are overdue than last month. That tells you where to look. It doesn’t tell you why. Perhaps the clinic was busier. Perhaps payments haven’t been recorded yet. Perhaps follow-up has slipped. Those explanations call for different responses.
This is where I want AI to help: bring together the relevant records, show me what’s changed and let me ask questions without rebuilding the report each time. I should be able to check the evidence behind an answer, including how current it is. If the information is missing, I want to know that too.
Once you understand the situation, you can decide what to do. Some invoices may need a follow-up. Others may need a correction or a conversation with someone who understands the history. You might decide to wait.
The software should help you make that distinction and carry the decision forward. Who owns the next step? What have they already tried? When should they check again?
Those are ordinary questions. But when the answers live in someone’s memory, an email thread or a separate spreadsheet, the team has to piece the situation together again every time.
What happens at the next review?
When you return a month later, the dashboard should remember what you decided.
You should be able to see which follow-ups happened, which are still waiting and where something went wrong. If an automated step failed, it should be clear what remains unfinished. You shouldn’t have to discover the failure by noticing that nothing changed.
Then comes the question that matters: did your response help?
A reminder being sent is different from an invoice being paid. Even if payment arrives, you can’t assume the reminder caused it. Still, keeping the action and the result together gives you something useful to review.
Perhaps the same missing information keeps delaying payment. Chasing each invoice may be necessary, but you also have a reason to examine how that information gets collected in the first place. The next improvement could be a clearer intake step, a different handoff or a better weekly routine.
Where AI fits
Attio connects context with actions; Sequoia argues for selling completed work. Both raise the question of what people still need to decide.
An agent can prepare the next step. That doesn’t mean it should make every decision or take over every conversation. Nor does a longer list of AI recommendations help if nobody has time to assess them.
This shift has changed how I think about Caretrics. We started with the problem of manual reporting. Understanding the data mattered, but the questions went further: what needs attention, what should we do, and how do we make this easier next month?
The dashboard I want brings those questions together. A chart when you need an overview. A conversation when you need to investigate. A clear next step when you’re ready to act. And a record you can return to.
What you learn from doing the work should help you improve how the work gets done.