
- Malleable Hub
What is an agentic operation?
Most AI at work tells you what needs doing. An agentic operation does the work, brings you in for the calls that need your judgment, and keeps going.
Eric Pelz
Founder & CEO · Malleable
Most teams already use AI at work. The most ambitious ones have it running in the background, flagging what needs their attention. Few realize it can go a step further and orchestrate the work itself: it does the steps, brings you in for the calls that need your judgment, and then keeps going, working across your team until the job is done.
For each step below, ask one question: who does the work?
1. Dashboards. AI builds the report. You read it and decide what to do. You do the work.
2. Chat assistants. You ask for help. It can spend minutes calling tools, and even make the change for you. But every chat starts from scratch: you explain the job each time, and nothing happens until you ask. You still do the work.
3. Proactive AI. It runs in the background and tells you what needs doing, like a supplier cost that went up or a lab report missing a test. This is useful, and it's usually where things stop. You still do the work.
4. Agentic operations. Every time an invoice comes in, it reads it, updates the price list, and brings you the one change that needs your judgment. Once you answer, it keeps going: it updates Salesforce and lets Sales Ops know. It does the work, and you steer.
Two margins dropped. Why?
- Find what changed
- Update the price list
- Ask Sales Ops to update Salesforce
Steps three and four can look alike, since both run in the background. The difference is what's waiting for you when you get back. Proactive AI leaves you with a to-do list: three costs went up, and now you have to reprice them and ask Sales Ops to update Salesforce. An agentic operation uses your rules to reprice them, asks for an approval, and then updates Salesforce and keeps Sales Ops in the loop.
It also has to keep up with the business. Promos start, and suppliers change how they price. When that happens, you update your pricing rules document. It reads that document just like your team does, and the next invoice runs the new way. You don't file a ticket or wait on a developer.
You don't have to hand over the whole job at once. Most teams start small and give it more as they come to trust it. At first, have it ask before every change. Once it gets the routine ones right, let those go through and keep reviewing the unusual ones.
That's how Gregory Schroeder's accounting team at MGT Insurance got started. Their first agentic operation was a simple form for entering a journal entry. An agent drafted the entry, Gregory approved it, and it went into NetSuite.
"The first run was just two journal entries," he says. "It didn't save time, but it proved the hypothesis."
From there, the team automated one step at a time and has since built over 50 agentic operations. Read their story.
Try Malleable