Field notes
Concrete questions about AI agents and systems for operations, with a short answer up top and the technical detail below.
Every note comes from a published case or from a project in flight on this site, with its exact status, and uses only the figures that case already publishes.
How do you build assisted quoting without losing control?
AI takes the customer's request (an email, a chat message, or a PDF), cleans the data, and fills in the quote; a person reviews it before it goes out. In the portal of an industrial equipment distributor that works this way, a real request became a review-ready quote in 42 seconds, work that used to take days.
Read the note →What does a WhatsApp agent do, and when does it hand off to a person?
It answers inbound messages, asks 3 qualifying questions, places the lead in a funnel stage, and schedules the follow-up. When the conversation needs a person, it hands it off with a written summary: what the customer needs, by when, and the suggested next step.
Read the note →What changes when a microfactory runs on its own system?
The whole chain, from demand to the shop floor, is instrumented in one system instead of a generic ERP template: per-piece costing, deterministic quoting, and a record of runs and incidents. IteraShop², our microfactory, works this way with 13 fabrication machines and 5 part families made to order.
Read the note →What happens to a request that arrives out of catalog?
The system flags the line item as out of catalog or without a price, and the request stops in the inbox until a person resolves it. The numbered quote is generated only when every line item is ready. That is how the portal of an industrial equipment distributor works today, in production with the client team.
Read the note →How do you reactivate customers from one to three years ago without bothering them?
With a relevant message, not a mass campaign. The agent writes to customers who bought one to three years ago, handles the reply with the same three qualifying questions, and, when the conversation matters, hands it to a person with a written summary. The first deployment, for a sign and graphics company in Houston, is built and in pre-launch.
Read the note →What gets mapped in a discovery before anything is built?
Three things: each workflow with how it works today and how much it matters, the pending items waiting for a reply, and where a person approves. They are seen in a prototype with fictitious data before a single line of the system is written. That is where the project with a commercial glazing contractor in Houston stands today: in discovery.
Read the note →
What would your first agent be?
Tell us which process costs you the most. The guided diagnostic is five short steps and shows no prices.
Email: ventas@iterasystems.mx