AI assistant trial for sales: what to connect, what to test and how to judge results in 14 days

BOTrento Team13 October 20269 min read

Before you start: pick one process and write down your baseline

Most trials fail because they try to test everything at once. Pick one process, usually incoming enquiry → quote → follow-up, and leave invoicing, support and marketing for later. If you're not sure which tasks suit an assistant, our overview of which sales tasks to hand over first and which to keep human helps you draw the line.

Then spend half an hour noting where you stand today. Use your own figures from the last month, not estimates:

• How many enquiries came in, and through which channels (email, web form, phone).
• Your typical time from enquiry to first reply. Check a handful of real threads in Outlook or Gmail.
• How many quotes went out, and how many of those got no follow-up at all.
• Who on the team handles enquiries, and who approves prices.

This baseline is what you'll compare against on day 14. Without it, "it felt quicker" is the only verdict you'll get.

What to connect, and what to leave out

Connect

• Your sales inbox. This is the shared enquiries address or the mailbox of the person who handles most enquiries. Mailboxes that support IMAP can usually be connected directly. With Gmail, Google Workspace or Microsoft 365, forwarding or importing is often the practical first step.
• Open deals. An export from Pipedrive, HubSpot or your quote spreadsheet as CSV or XLSX, with client, subject, quote date, value, status and next step.
• Your standard answers. Your price list structure, typical lead times and terms, so the drafts match the way you actually work.

Leave out

• Accounting data from Xero or QuickBooks. You don't need it to test sales routine.
• Personal mailboxes that have nothing to do with sales. Limit the trial data to what the task needs, which also makes the GDPR side simpler. Our GDPR checklist for AI sales tools lists what to ask any vendor.

The 14-day trial plan, day by day

The example below follows an equipment rental company with five people. The owner approves prices and two salespeople handle enquiries. Change the roles to fit your team.

1. Preparation (before day 1): the owner writes down the baseline, picks the mailbox and exports open deals. One salesperson is named trial lead.
2. Day 1, connect: the trial lead connects the inbox or uploads the export and checks that recent enquiries appear correctly. Fix wrong client names and merged threads now, not halfway through.
3. Days 2–3, review what's open: the team goes through the open deals the assistant has picked out. Which quotes have had no reply? Which promises ("I'll send the revised price by Friday") are still waiting? Each person marks what's accurate and what isn't.
4. Days 3–7, drafts on live enquiries: salespeople reply to every new enquiry using the AI draft as a starting point. They score every draft on the scorecard in the next section. The owner still approves prices.
5. Day 7, mid-point check (a short meeting): look at what's working, which drafts keep needing the same edits, and what context the assistant is missing (lead times, minimum hire periods, delivery fees). Add that information.
6. Days 8–12, follow-ups and daily priorities: the team starts each working day from the assistant's priority list and approves follow-up drafts for quotes that have gone quiet. Record every follow-up you wouldn't have sent otherwise.
7. The day before the decision, measure: the trial lead compares reply times and follow-up counts with the baseline and adds up the scorecard.
8. Day 14, decide: hold a short meeting and go through the criteria below. Then you buy, take the conversation further, or stop and export your data.

A short kick-off message keeps everyone on the same page:

Hi team, from Monday we're trialling an AI assistant on our enquiries for 14 days. It only prepares drafts, so nothing goes out until one of us approves it. Please use a draft for every new enquiry, score it on the shared sheet (as is / minor edits / rewritten / unusable) and note any follow-up it reminded you about. Prices still come to me. We'll review the results together on [date].

How to test AI email drafts properly

Testing drafts on made-up enquiries tells you very little. Use the messages that actually land in your inbox, including the messy ones: vague requests, price objections and last-minute changes.

A simple scorecard

What to watch for

• Accuracy: does the draft answer what the client actually asked, in their language?
• Invented details: drafts must never confirm availability, dates or prices you didn't supply. Your approval step is there to catch exactly this.
• Attachments: if key details sit only in a PDF brief, check how the tool handles that before you rely on it.

For the reply process itself, see five steps to replying to sales enquiries faster.

How to judge the results on day 14

Set your criteria before the trial starts so nobody argues about the decision afterwards. Three questions are usually enough:

1. Reply time: is the time from enquiry to first reply shorter than your baseline, and more consistent across the team?
2. Drafts accepted: do most drafts end up as "as is" or "minor edits"? If many are rewritten, work out whether the assistant is missing context you can add or whether it's a real limitation.
3. Follow-ups caught: how many quotes got a follow-up, or promises were kept, that would otherwise have slipped? Our guide on how to stop forgetting quote follow-ups explains why this is often where the value shows first.

Then do the maths honestly. Put the subscription and your team's time on one side, and the hours saved and the deals you kept from going quiet on the other. Our article on what an AI assistant really costs a business breaks the cost into its parts. Don't expect the trial to prove revenue growth. Two weeks is too short for that, and no honest vendor will promise it.

How a BOTrento trial runs in practice

BOTrento Chief of Staff is a managed process for enquiries, quotes and follow-ups, not a chat window. Here's how the plan above works with it:

• Setup with a person: after you sign up, someone from our team contacts you within 24 hours on working days and helps you connect your mailbox. You get 7 days to prepare. The 14 days only start once your first data has been uploaded successfully, through an import or a mail connection.
• What you upload: an IMAP mailbox, forwarded or imported email from Gmail or Microsoft 365, and your deals from Pipedrive, HubSpot or a spreadsheet as CSV/XLSX.
• Early insight: at the start you get a "Lost money" report showing enquiries and quotes that went unanswered. It only looks back in time for mailboxes connected via IMAP.
• Every working morning: a "Today 1+3" overview, in the app and by email, shows the most important thing to handle today plus three more.
• What you approve: drafts of replies, quotes and follow-ups in English, German, French or Czech, plus reminders for promises it picked up from your emails. Nothing is sent until a person approves it. BOTrento does not confirm dates or bookings, and it doesn't replace your CRM.
• Cost during the trial: AI usage is included. You don't need a payment card or your own API key. If you later want to use your own Anthropic key, the API keys guide explains how.

When we ran BOTrento on our own processes in an internal pilot, it handled 35 enquiries, 18 emails, 24 quotes and 15 actions in 7 days. That was our own test, not a client result. Your data will give you your own numbers. If you rent out equipment, see how this fits equipment rental enquiries. You can also walk through a sample process, check the plans and what they include, or start your 14-day trial on your own data.