--- title: "Automating quotes with an AI agent" description: "What quoting by hand costs, what the quote that arrives third costs, and what an agent does at each step until the document goes out." url: https://studiochat.io/blog/cotizar-sin-demorar-tres-dias locale: en translations: { es: https://studiochat.io/es/blog/cotizar-sin-demorar-tres-dias.md, pt: https://studiochat.io/pt/blog/cotizar-sin-demorar-tres-dias.md } site: https://studiochat.io --- [ Back to blog](/blog) A quote request arrives at ten on a Tuesday. Answering it means reading it and checking whether it is complete, which it almost never is. It means writing to the customer for what is missing. It means waiting. It means identifying the account and finding which price list applies. Calculating quantities and discounts. Assembling the document, sending it, noting that it went out, and writing again two days later if there is no reply. Eight steps. Seven of them need no judgment at all. ## [The cost isn't assembling the document, it's arriving third](#the-cost-isnt-assembling-the-document-its-arriving-third) Our own work on a quote is bounded and almost everybody does it quickly. The delay is somewhere else, and it is an uncomfortable place: the follow up question. The request arrived without a detail, somebody writes to ask for it, and the clock stops until the customer answers. That stretch does not depend on us, usually lasts longer than everything else together, and starts late: not when the request arrived, but when somebody had time to open it and notice something was missing. And the cost of arriving late does not look like the other costs in this blog. It is not a correction or rework: it is a quote that never got discussed. Nobody rejected it for being expensive. It arrived once the conversation had already moved on. ## [The same quote, today and with an agent](#the-same-quote-today-and-with-an-agent) One quote, end to end * the agent does it * a person does it * always a person the steps of the quote * the agent does itRead the request and see what's missing Email and chat Reads the request the moment it arrives, at any hour, and detects what is blocking the quote before anybody opens it. * the agent does itAsk for the missing detail immediately Email and chat Writes in minute one instead of two hours later, which is the only stretch of the turnaround that can genuinely be shortened. * the agent does itIdentify the account and its list Customers Resolves which account it came from and which price list applies, with the rate currently in force. * the agent does itCalculate quantities and discounts Pricing Applies volume discount rules across the whole order rather than line by line, which is where the errors get made. * the agent does itAssemble the document Document Generates the quote in the usual format, with the right terms and the lead time that comes out of stock. * the agent does itSend it and log it ERP Sends it through the channel the customer wrote in and logs it, with a date, so the follow up exists at all. * the agent does itFollow up two days later Email and chat Writes again incorporating what happened, which is the step abandoned when new quotes are waiting. * always a personDecide the exception price Pricing When the price the rule produces is not the one worth offering, a person decides, with everything else already assembled. Quote requests per month: 300 Minutes of our own work per quote: 14.0 Hourly cost of whoever quotes: USD 12 Share arriving late or going out with an error: 4% What a quote that never got discussed is worth: USD 40 Today that is 70 hours a month and 12 quotes arriving late or going out wrong. Between the time and the ones never discussed, USD 1,320 a month. With the agent holding 7 of the 7 steps, 4 hours and USD 66 are left. The difference is USD 1,254 a month, USD 15,048 a year. And 36% of that difference is not time: it is the quotes that entered the conversation instead of arriving once a decision had been made. Of what a person still does, the heaviest step is: Decide the exception price. You set all five numbers. How the minutes split across steps is ours, taken from how it gets done by hand. The modelling assumption: the quote that arrives late or with an error falls in the same proportion as the mechanical steps stop waiting for somebody to have time. Leave only the second step on, asking for the missing detail immediately, and switch off everything else. It is the cheapest implementation there is and it moves the turnaround more than speeding up our own work, because it attacks the only stretch that really weighs. This object also has [a page of its own](/quotes-with-an-agent), in case you want to send just that. ## [What the agent does that a person cannot](#what-the-agent-does-that-a-person-cannot) **It reads the request at eleven at night.** The follow up question that goes out when the request arrives, rather than when somebody opens the inbox, takes half a day out of the turnaround with nobody working faster. **It follows up on the older quotes.** The two day follow up is exactly what gets abandoned when new quotes are waiting, and it is where half the closing lives. **It uses the current list and calculates the discount on the whole order.** Two common errors, both from rushing, both expensive: one is discovered when the customer compares, the other when the invoice arrives. ## [What stays with a person](#what-stays-with-a-person) One thing: the exception price. When the price the rule produces is not the price worth offering, that is commercial judgment and it is precisely the work the team exists for. And the scope when the request is ambiguous. "I need a quote for 500 units" does not say whether freight is included, whether partial delivery is acceptable, or whether production can take the lead time. An agent that picks an interpretation and quotes on it produces a fast wrong quote, which is worse than a slow one. ## [How it gets implemented](#how-it-gets-implemented) The first thing is separating two categories with the person who quotes today: what can be calculated with a written rule, and what cannot. That conversation takes an afternoon and usually reveals that the calculable share is higher than the team believes. Then the order: detecting what is missing and asking for it first, which is the cheapest and the biggest mover. Assembling the document second. Sending and following up third. Over time, what lived in somebody's head gets written down, because every exception that repeats turns into a rule. That is the most valuable side effect of the project and it is never in the plan. ## [What to measure](#what-to-measure) Turnaround, measured from the request arriving to the quote going out, and read by percentile rather than by average. The average hides the tail, and what gets lost is the slowest ten percent. What share of requests arrived complete, which is the number that explains the turnaround. And the reply rate to the follow up question within two hours, which is the only stretch that does not depend on you and the one most improved by asking at the right moment. Frequently asked questions ### Which part of a quote can an AI agent assemble? It can read the request and detect what is missing, ask for it immediately instead of two hours later, identify the account and its price list, calculate quantities and rule based discounts, assemble the document, send it, log the send and follow up two days later. Only one thing needs judgment: the exception price. ### Why are quotes with a good price lost? Because they arrive late. The first one to land sets the frame everything else is read against, and the third lands once the conversation has moved on without it. Price decides between the quotes that arrived in time, not between all the ones that were requested. ### Where is the delay actually? Almost never in our own work. It is in the follow up question: the request arrived missing something, somebody writes to ask, and the clock stops until the customer answers. That stretch does not depend on us and usually lasts longer than everything else together. Which is why the useful move is asking in the first minute, not two hours later. Share [ X](https://x.com/intent/post?text=Putting%20a%20quote%20together%20is%20eight%20steps%20and%20seven%20need%20no%20judgment&url=https%3A%2F%2Fstudiochat.io%2Fblog%2Fcotizar-sin-demorar-tres-dias) [ LinkedIn](https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fstudiochat.io%2Fblog%2Fcotizar-sin-demorar-tres-dias) [ WhatsApp](https://wa.me/?text=Putting%20a%20quote%20together%20is%20eight%20steps%20and%20seven%20need%20no%20judgment%20https%3A%2F%2Fstudiochat.io%2Fblog%2Fcotizar-sin-demorar-tres-dias)Copy link Cristian Pereyra Cristian Pereyra · Co-founder · Engineering, StudioChat Want agents like these working for your team? [Talk to us now](https://cal.com/studiochat)