Skip to content
Giallo

AI sales agent: your rep walks into the call prepared, and the CRM tells the truth.

What does an AI sales agent do in a sales department?

An AI sales agent works on the steps sales reps put off: reading and qualifying incoming inquiries, preparing a brief before the call, writing the follow-up after the meeting, flagging stalled deals in the CRM. Giallo Studio builds these agents to measure with Claude, GPT and Gemini; prices, discounts and commitments to the customer stay with the salesperson.

Frequently asked questions, straight answers.

A chatbot answers; an agent moves a job forward. A lead generation chatbot chats on the website; AI sales agents read the inquiry, qualify it, update the CRM, alert the right salesperson and prepare the next step, even outside the chat.

No, and in the projects we handle we don't want it to. It can reply, prepare quotes and send reminders about deadlines, but prices, discounts and commitments are decided by the salesperson, because they are agreements the company must be able to keep.

No, the one you already use works if it offers APIs, such as HubSpot, Salesforce, Pipedrive or Microsoft Dynamics. If your CRM today is a shared spreadsheet, we start there, and we tell you frankly whether and when it's worth switching.

No. It removes the side work, such as looking up information, summarizing meetings and updating the CRM, and leaves more room for the customer relationship, which stays with the salesperson.

No. We build agents that work on requests received and on your existing customers, not on mass mailings to strangers.

One per task, as you can see by following an inquiry that came in from the website after a campaign. Claude Haiku sorts it along with the others in seconds, Claude Sonnet writes the first reply and the follow-ups, GPT builds the quote from the price list. If calendar and email live in Google Workspace, Gemini steps in for the pre-call briefs.

The difference is who sells. Here we're talking about an in-house sales department with a company CRM; for multi-firm agents who visit customers on behalf of several companies, we've designed different agents, dedicated to orders, price lists and commissions.

The AI agents we design, one by one.

Inbound inquiry agent

Reads, qualifies and assigns every new inquiry, with a first reply already written.

As soon as an inquiry comes in from the website, the chat, an email or a campaign.

Connects to: Website forms and chat · Sales inbox · Company CRM · Salespeople's calendars

  1. Reads the inquiry and works out what is needed, by when and with what constraints.
  2. Looks up public information about the company, such as industry and location, without collecting unnecessary personal data.
  3. Checks the inquiry against the qualification criteria written by the sales director.
  4. Assigns it to the right salesperson or territory and prepares a draft first reply.
  5. If the website chat is active, it asks the visitor the few missing questions before handing over to a person.
Claude Haiku
Classifies each inquiry in seconds, even when many arrive at once after a campaign, at a cost that holds up at volume.
Claude Sonnet
Writes the first reply and runs the website chat, where you need relevant questions and no promises on prices or delivery times.

Who approves. The sales director approves the qualification criteria, assignment rules and the questions the chat may ask. In chat, the agent asks only qualifying questions and gives no prices or timelines; every email reply, beyond the acknowledgment of receipt, goes out after the salesperson's review.

Call prep agent

Puts everything the salesperson needs to know before walking into the meeting into a short brief.

Before every scheduled appointment with a customer or prospect.

Connects to: Google Calendar or Outlook · Company CRM · Quote archive · Customer's website

  1. Reads the calendar appointment and identifies the company and people involved.
  2. Gathers past purchases, open quotes, support requests and recent emails from the CRM.
  3. Adds relevant public news from the customer's website.
  4. Writes the brief: situation, open points, questions to ask, risks to watch.
  5. Sends it to the salesperson on the channel they use most, email or the CRM app.
Gemini
When the company works in Google Workspace, it reads calendar, Gmail and Drive documents together and lines up many different sources without intermediate steps.
Claude Opus
For customers with a long history and complex deals, it reasons over many emails and quotes and points out what is really still pending.

Who approves. The brief is an internal tool for the salesperson: it is never sent to the customer and contains no judgments about people.

Follow-up and quote agent

Turns the meeting into a recap email, a draft quote and an updated CRM.

Right after a meeting, when the salesperson dictates a memo or uploads notes.

Connects to: CRM app or voice memo · Company CRM · Price list and terms of sale · Email

  1. Transcribes the voice memo and extracts decisions, customer requests and next steps.
  2. Writes the recap email in the salesperson's tone.
  3. If the customer asked for a quote, it prepares the draft from the price list and standard terms and highlights every off-list item.
  4. Updates the deal in the CRM with stage, next contact and notes.
  5. Holds the email and quote pending the salesperson's approval.
Claude Sonnet
Writes lean, friendly recaps and doesn't add discounts, timelines or commitments the salesperson didn't make.
GPT
Builds the draft quote from the price list and terms in structured fields, so every price can be checked line by line.

Who approves. Discounts, special terms and delivery dates are decided by the salesperson within the limits set by management. No quote goes out without their confirmation.

Pipeline cleanup agent

Flags what in the CRM no longer matches reality and suggests corrections to whoever owns the deal.

Friday afternoon, before the sales meeting at the start of the week.

Connects to: Company CRM · Pipeline report · Email or internal chat

  1. Scans open deals and finds those with no recent activity or a past close date.
  2. Recognizes duplicate contacts and companies and suggests how to merge them.
  3. Prepares for each salesperson a short list of deals to update, close or revive.
  4. Writes a summary for the sales director that separates live deals from stalled ones.
Gemini Flash
Going through every deal and contact each week is a high-volume check with simple rules: it needs speed at a contained cost.
Codex
It's used to write and maintain the connector to the CRM, including the matching rules for duplicates; we review the code ourselves before every release.

Who approves. Contact merges, closures and stage changes are confirmed by the salesperson who owns the deal. The agent proposes and deletes nothing.

Tell us how your team sells

Tell us how inquiries come in, which CRM you use and how a quote is put together: we'll propose the first agent and the salesperson to test it with.

Prefer the full form? Go to the contact page.

No spam or sales pitches. Response within 1 business day.

We use this data to contact you and, if we work together, to manage the project. We store it in Giallo's internal management system, on servers in the European Union, where only our team can see it; by email we only receive a notice, without your data. If you do not become a client, we delete it 3 months after our last contact. Details and your rights: privacy policy.

Do you prefer to write or call us?

Where time gets lost today.

These are the points we hear about most often. Each one is a possible first agent.

Website inquiries wait for days

Forms, emails and chats end up in a shared inbox. Whoever has time reads them, whoever is traveling doesn't; small and important inquiries sit in the same queue, and the reply arrives when the prospect has already talked to a competitor.

The agent reads every inquiry as soon as it arrives, enriches it with the company's public data, checks it against the sales director's criteria and assigns it to the right salesperson with a draft first reply.

Before a call, everything gets looked up from scratch

History in the CRM, scattered emails, old quotes in a folder, news on the customer's website. The rep pieces the picture together in the car, or walks into the meeting without it.

Before every scheduled appointment, a short brief arrives: who the customer is, what they've bought, what's still open and which questions are worth asking.

The post-meeting follow-up never goes out

You get back to the office with half-finished notes and new urgent matters on the desk. The recap email slips, the CRM stays stuck at the previous stage and the customer perceives less interest than there actually was.

From notes or a voice memo, the agent writes the recap email with next steps and updates the deal; the salesperson reviews and sends.

The CRM shows a pipeline that doesn't exist

Deals stalled for months stay open, close dates aren't updated, contacts get duplicated. Forecasts are made by gut feeling, in the meeting.

Every week the agent flags deals with no activity, overdue dates and duplicates, and suggests to each salesperson what to update. The salesperson confirms the changes.

Which AI model we use, and for which task.

We don't tie you to a single provider: we work with Claude by Anthropic, GPT and Codex by OpenAI, and Gemini by Google, and we choose the version during testing, on your own cases.

Claude Haiku

  • Inbound inquiry agent. Classifies each inquiry in seconds, even when many arrive at once after a campaign, at a cost that holds up at volume.

Claude Sonnet

  • Inbound inquiry agent. Writes the first reply and runs the website chat, where you need relevant questions and no promises on prices or delivery times.
  • Follow-up and quote agent. Writes lean, friendly recaps and doesn't add discounts, timelines or commitments the salesperson didn't make.

Gemini

  • Call prep agent. When the company works in Google Workspace, it reads calendar, Gmail and Drive documents together and lines up many different sources without intermediate steps.

Claude Opus

  • Call prep agent. For customers with a long history and complex deals, it reasons over many emails and quotes and points out what is really still pending.

GPT

  • Follow-up and quote agent. Builds the draft quote from the price list and terms in structured fields, so every price can be checked line by line.

Gemini Flash

  • Pipeline cleanup agent. Going through every deal and contact each week is a high-volume check with simple rules: it needs speed at a contained cost.

Codex

  • Pipeline cleanup agent. It's used to write and maintain the connector to the CRM, including the matching rules for duplicates; we review the code ourselves before every release.

How we choose between models: the best AI model for agents.

How we work with you.

  1. Won and lost deals, retraced with the salespeople

    With the team, we walk through a few recent deals to see where they stalled and what information was missing at the decisive moment.

  2. Qualification criteria in writing

    We turn what the sales director knows by heart into rules: which inquiries deserve a call right away, which an email, which should be routed elsewhere.

  3. The CRM as the single source of truth

    We connect the agents to the CRM via API, with separate read and write permissions, and verify that every change is attributed to a person.

  4. A pilot salesperson

    We start with one rep willing to try it, collect their corrections and only then extend the agents to the rest of the team.

AI agents for sales teams

At which point in a deal do you lose the most time?

Tell us how inquiries come in, which CRM you use and how a quote is put together: we'll propose the first agent and the salesperson to test it with.