Practical AI applications for sales and marketing teams

Research: before a call, an assistant that reads the prospect’s website, recent news and LinkedIn and produces a half-page brief — what they make, who runs it, what changed recently, a likely angle — for the salesperson to check in two minutes. Drafting: the first version of the follow-up message from the call notes, the proposal section from the discovery facts, the LinkedIn post from the founder’s voice note, the answer page from the expert’s twenty-minute interview — always edited by a person, and better the more of your own material it is given. Qualification: reading incoming enquiries against your written standard and sorting them, with the borderline ones flagged for a person, so the coordinator’s morning starts with a sorted queue. Analysis: a month of call transcripts for recurring objections, lost-deal reasons for patterns, customer feedback for themes, the CRM for accounts going quiet — questions a person could answer with a week and rarely does. Assistance: a salesperson asking, in plain words, what the last three conversations with a customer were, what was quoted, what is due — answered from the CRM they never had time to read.

Start with two: research and drafting, because they are the safest, the easiest to review, and the quickest to show hours saved.

Human review is not optional. Every AI output that reaches a customer — a message, a proposal, a page — is read and edited by a person who is responsible for it; every research brief is checked against the source before it is trusted in a call; every qualification decision that rejects an enquiry is sampled. AI is confident when it is wrong, and a customer who receives an invented fact about their own company does not come back.

Data protection: decide what may be put into which tool — customer names and personal details into an enterprise tool with a data agreement, never into a free consumer chat; your pricing, contracts and customer lists treated as confidential; India’s data-protection law applied to personal data, which means purpose, consent and the right to be forgotten still hold when a machine is doing the reading. Quality controls: a written rule for each use — what goes in, what comes out, who reviews, what is never sent unedited — and a monthly look at a sample of outputs by the owner of the AI use.

Choose one use case, one team member, one month, and one number: hours spent on pre-call research per week; time from call to follow-up sent; enquiries sorted per hour and the error rate on the sample; days from expert interview to published page. Measure the number before, run the pilot with the rules, measure after, and read the outputs. Keep what saved time without costing quality; drop what did not; write down what was learned. Then the next use case.

The pilots that work in SMEs are unglamorous: a salesperson who spends twenty minutes a day on research now spends five; a coordinator who sorted enquiries by hand now reviews a sorted list; a founder who never wrote the content now records a voice note and edits a draft. The pilots that fail are the ones that automated the customer conversation.

The mistakes: automating the customer conversation; sending drafts unedited; pasting customer data into consumer tools; buying an “AI CRM” before the CRM has data; ten use cases at once; no owner; and judging AI on impressiveness rather than on hours saved and errors caught. A safeguard: for a month, keep every AI draft beside the version that was actually sent — the difference is the review the process needs.

Practical AI in a small team looks like two or three well-defined uses, each with a rule and a reviewer, saving several hours a person a week and making the CRM finally worth its data. This is the AI-readiness work we do with SMEs — the task inventory, the two starting uses, the data rules, the tool choice with a data agreement, the pilots with numbers, and the monthly quality check — and the free audit starts with the hours list.