In many companies, sales information lives in too many places. Conversations with prospects sit in each salesperson's personal WhatsApp, meeting notes are in a notebook, quotations are in email, and the CRM holds little more than names and phone numbers. When a salesperson goes on leave or moves on, the history of the customer relationship leaves with them.
AI for sales teams and CRM can help clean this up. Your salespeople still do the meeting and the persuading. AI takes over much of the admin work that keeps the CRM empty, and it helps managers see the pipeline more honestly. This article covers the four most realistic uses, their limits, and how to get started.
Lead and customer summaries
Before calling or visiting a prospect, a salesperson needs to know what has already been discussed. That history might be dozens of messages, several meeting notes and two versions of a quote. AI can read all of it and produce a short summary: what the customer needs, objections raised so far, the latest offer, and any promises not yet kept.
Take a baking supplies distributor serving hundreds of shops. A salesperson covering for a colleague does not need to reread the entire history. The summary in the CRM is enough to start the conversation without sounding unprepared.
A summary is only as good as the data behind it. If meeting notes never make it into the CRM, the AI cannot summarise them. That is why this use usually goes hand in hand with better note-taking habits.
Follow-up drafts that stay personal
Many deals are lost because follow-up comes late or not at all. AI can prepare a draft message based on the latest context, such as a reminder about an unanswered quote, a recap of a meeting, or confirmation of a demo time. The salesperson reads it, adjusts the tone and sends it. On WhatsApp, short drafts usually work better than long messages that read like a template.
Two things are worth protecting. First, the salesperson still sends the message, so a person reads it and stands behind it. Second, the tone has to fit the customer. A shop owner in a small town and a procurement manager at a large company usually expect different styles. Keep a few strong messages from your senior salespeople as style references for the AI.
AI pipeline insight in your CRM
Pipeline reports tend to be optimistic. Deals that have not moved in three months still sit in the negotiation stage, and expected close dates slip every week. AI can help managers by reading CRM data and flagging patterns worth attention, such as deals with no activity for a long time, large quotes with no follow-up, or long-standing customers whose orders are becoming less frequent.
Managers can also ask questions in plain language, for example which customers in Central Java have not reordered since last quarter. That used to mean exporting data to a spreadsheet and losing half a day. The AI reads the data, but deciding priorities remains the job of a manager who knows the customers and the market.
CRM data hygiene
A messy CRM makes every use above less valuable. The same customer is recorded three times with different spellings, phone numbers lack area codes, and company names are abbreviated in one place and written in full in another. AI is quite good at this kind of clean-up work.
- Finding likely duplicates, such as the same shop name with and without a CV prefix, and suggesting a merge.
- Standardising phone numbers, addresses and city names.
- Filling empty fields from existing notes, such as business type or city.
- Flagging leads with incomplete data before they enter the pipeline.
- Grouping free-text reasons for lost or cancelled deals into categories you can analyse.
Review proposed changes before applying them, especially merges. Two branches of the same retail chain can look like duplicates when they are in fact separate customers.
Limits and customer data
CRM data contains personal data, such as names, phone numbers and purchase history. The UU PDP rules on processing personal data still apply when that data is processed with AI. Use company accounts rather than salespeople's personal ones, and choose services whose data terms you have checked. Access rights matter too. Salespeople only need to see the customers they handle, and the AI should follow the same access rules.
Do not let AI send messages to customers automatically without review, at least in the early stages. One message sent to the wrong person or with the wrong price can damage a relationship built over years.
Start with a single use, such as lead summaries for the sales team in one branch. After two or three weeks, ask the team directly whether the summaries are accurate and actually used. Their feedback is worth more than any feature list.
CRM and AI on one set of data
All of this is easier when customer, finance and project data are not scattered across many apps. Trellis brings Harvest, its CRM, together with other business apps on one company record, and Sage lets you ask questions about your own company data, kept isolated per company. See the details on the Trellis page.
