An AI sales assistant on the levelpartner.ai platform works for your salespeople rather than in place of them. It drafts replies for them to approve, summarises long conversation threads, prepares context before a call and keeps the pipeline up to date. The salesperson stays in every conversation; the assistant removes the typing, the searching and the admin around it. This guide explains how to set up that human-in-the-loop model and why it suits many firms better than a fully autonomous agent.
Sales AI assistant or autonomous agent
There are two ways to put AI into sales. An autonomous agent talks to leads by itself. A sales AI assistant sits beside the salesperson and suggests. The difference is who presses send.
The assistant model fits firms where relationships matter, deal sizes are meaningful and buyers expect a named person. Consultancies, mortgage brokers and agencies selling retained work usually sit here. The salesperson gets most of the speed benefit without handing the conversation to software.
What a virtual AI sales assistant does in a day
In practice the assistant handles four jobs:
- Drafting a reply to each inbound message, which the salesperson edits and sends from Conversations.
- Summarising a long thread so a salesperson picking up a lead knows the history in seconds.
- Drafting follow-up emails after meetings from notes the salesperson adds.
- Prompting pipeline updates, so stages and values reflect what was said.
Before you start
- Salespeople set up as users, each with their own calendar and assigned contacts.
- A pipeline in Opportunities with stages your team agrees on.
- A knowledge base with services, process, common objections and approved answers.
- A short tone guide so drafts sound like your salespeople, not like a brochure.
- Agreement from the sales lead that drafts are never sent unread.
How to set it up
- Go to Conversations > Conversations.
- Open a conversation with an AI Suggested Reply.
- Review the reply card and the Suggested actions.
- Turn off any action you do not want to run.
- Click Edit Reply to change the message if needed.
- Click Send & run.
Suggested replies only appear when the agent's mode is set to Suggestive, which you choose when you save the bot. Before rolling this out, sit with each salesperson for ten minutes and show them where the reply card sits in a thread, how to edit it and what each suggested action will do. After two weeks, pull the drafts your team rewrote most heavily and fix the knowledge base material behind them, because a weak draft nearly always traces back to a missing or vague answer.
Where the time goes back
A salesperson at a firm of 50 people often spends more of the day writing and searching than selling. Replies to routine questions, recap emails, finding the last thing a lead said. Each task is small. Together they eat the hours that should go to calls and proposals.
Say each salesperson handles about 30 active conversations and sends 40 written replies a day. If drafting cut the average reply from five minutes to two, that would be 120 minutes a day each, or roughly six hours a day across the three of them.
Weekly running cost in time
Set-up is half a day plus a short training session for the team. After that, the sales lead spends about 30 minutes a week on the knowledge base, fed by the drafts people rewrote. The assistant improves only if someone reads what people change.
Salespeople start sending drafts unread once they are usually right. That is how a wrong price or an outdated policy reaches a client. Keep the rule simple: read every draft in full before sending, without exception.
When an assistant is not enough
If your team cannot keep up with inbound volume at all, suggestions will not help; the queue is the problem. That is when an autonomous AI sales agent handling first response is the better fit. If your issue is after-hours enquiries, a suggestion-only assistant does nothing overnight. Look at AI follow up for chasing quiet leads automatically.
Troubleshooting
Drafts are generic. The knowledge base lacks specifics. Add real examples of good replies and the details your team usually includes.
Salespeople ignore the suggestions. They are slower to edit than to write. Shorten the target length in the instructions and ask the team which reply types to drop.
Thread summaries miss key facts. Important details live in calls that are not logged. Ask salespeople to add a short note after each call so the summary has something to work with.
Pipeline stages still lag. The assistant prompts but nobody acts. Make stage review part of the weekly sales meeting and use the pipeline report as the agenda.
Frequently asked questions
Is a virtual AI sales assistant the same as a chatbot?
No. A chatbot talks to leads. A sales assistant drafts for your salespeople, who stay in control of every message. The conversational AI guide explains the difference in more depth.
Can it write proposals?
It can draft proposal cover emails and summaries. Full proposals depend on scope and pricing decisions that a person should make. Use the AI email writer for the covering note.
Will salespeople accept it?
They accept it when it saves them typing and stays out of their way. They resist it when it is presented as monitoring. Introduce it as a drafting tool and let the team shape its tone.
Can we switch to autonomous replies later?
Yes. Many firms run the agent in Suggestive mode for a few weeks, then let the agent reply on its own for one channel once drafts consistently need no edits.