AI for sales teams: the tasks worth automating first

If your reps feel busy but the pipeline stays thin, the problem is usually where their hours go. Salespeople spend most of the week not actually selling, and that admin is exactly what AI is good at. Automate the research, the notes, the CRM updates and the follow-up drafts, and keep the discovery, the relationship and the close firmly human.

The temptation is to point AI at the selling itself: let it run the call, write the pitch, close the deal. That is the wrong end. The parts of sales that win business are the parts a person does well, and the parts that drain the week are the parts a machine does well. Get that split right and a team sells more without working longer.

This post is about which sales tasks to hand to AI first, which to keep human, and a short audit you can run this week to find your own team's wasted hours.

Where the time actually goes

The headline number is stark. In Salesforce's State of Sales report, seventh edition, 2026, a survey of 4,050 sales professionals, the average seller spends only about 40% of their time actually selling. The other 60% goes on research, data entry, updating the CRM (the customer record system where deals are tracked), writing routine emails and switching between tools. That is more than half the week spent on work that does not directly move a deal forward.

This is not a niche complaint. In the same research, 87% of sales organisations already use some form of AI for tasks like prospecting, forecasting or drafting emails. And the reps who use it are clear on why: in HubSpot's 2025 State of Sales Report, a survey of 1,000 global sales professionals, 84% said AI saves them time and improves their process, and only 8% reported not using AI at all.

So the question for most teams is no longer whether to use AI in sales. It is where to point it, and where to keep it well away.

What to automate first

The reliable wins share one pattern: AI does the drafting and the legwork, a person keeps the judgement and the relationship. Five stand out, in the order most teams should tackle them.

1. Prospect research. Before a first call, someone has to read the company's site, recent news, the buyer's role and history. AI can pull that into a short, structured brief in seconds, so a rep walks in prepared instead of skim-reading in the car park. Salesforce found sellers expect AI to cut prospect research time by about 34% once it is properly set up.
2. Meeting notes and call summaries. Note-taking during a call splits attention, and writing it up afterwards eats the gap before the next one. Let AI transcribe the call and produce the summary and the action list, so the rep can actually listen while they are in the room.
3. CRM updates and data hygiene. The CRM only helps if it is current, and reps hate keeping it current. AI can draft the update straight from the call notes, log the next step and flag records that look stale or duplicated, so the pipeline reflects reality without an hour of admin each evening.
4. Follow-up drafts. The chase email, the recap, the "here is what we discussed" note: these are templated work a person should sign but not write from scratch. AI drafts them from the call summary, the rep edits the specifics and sends. Salesforce found sellers expect AI to cut email drafting time by about 36%.
5. Pipeline summaries and reporting. The Friday roll-up and the pre-forecast tidy are pure overhead. AI can summarise where each deal stands and surface the ones that have gone quiet, turning an hour of report-building into a five-minute read.

Notice what these have in common. Every one produces a draft a person checks, not a decision a machine makes. That is the line.

The selling-time audit

Before you buy anything, find your own 60%. The selling-time audit is a table each rep fills in for one typical week: every recurring task, the hours it takes, whether AI could draft it, and whether it must stay human. Add the "AI could draft it" hours across the team and you have the size of the prize, in hours, in your own numbers.

Here it is filled in for a fictional four-rep team, one rep's week shown (illustrative figures, a 40-hour week):

For this rep, 24 of the 40 hours are draftable admin, and 16 are the human selling that wins the deal. Across four reps that is roughly 96 hours a week that AI can turn from blank-page work into quick review-and-send. It does not delete those hours, it shrinks them: the rep checks a draft instead of writing one. Even halving them hands each rep the better part of a day back for the calls that actually close.

To run it, give every rep the blank version, ask them to log one honest week, then total the "yes" rows. That total, not a vendor's promise, is your business case.

What to keep human

The flip side matters as much. Three parts of sales should never be handed to a machine, because they are the parts that create trust.

1. Discovery. Working out what a buyer actually needs, and what they are not saying, is judgement and listening. AI can prepare you for the conversation, but it cannot have it for you.
2. The relationship. People buy from people they trust. A perfectly worded AI message that the buyer can smell was automated does more harm than a plainer one a human clearly wrote. Use AI to draft, then make it sound like you.
3. The close and the negotiation. Reading the room, handling the objection, knowing when to hold and when to give: this is where deals are won and lost, and it is entirely human. Let AI tee up the numbers, then let the person do the deal.

A simple rule keeps the split clean: AI prepares and drafts, the person decides and speaks.

Data and accuracy guardrails

Two things go wrong when teams move fast. The first is bad drafts sent unchecked. AI writes confident, fluent text that can still get a name, a price or a promise wrong, so nothing it drafts reaches a customer without a person reading it. The recap that misstates what you agreed is worse than no recap at all.

The second is data. Prospect and customer records are sensitive, so use AI tools that are under contract not to train on your data, keep customer information inside the systems you already trust, and write a short note of which tools are approved for what. This is the same discipline that keeps AI useful in customer service without trapping the customer in a bot: speed with a person still accountable for the result.

It is also worth being honest about what "time saved" is for. Saving a rep six hours a week only pays off if those hours go into selling, not into more admin invented to fill them. We have written before about the gap between looking more productive and creating real business value, and sales is where that gap shows up fastest in the numbers.

Where to start on Monday

Pick one task from the audit that scored high on hours and clearly "yes" for AI, most teams start with follow-up drafts or meeting notes. Set it up in the sales tool you already use, run it for two weeks, and watch one number: hours that task now takes, and whether the quality held. If it saved real time and nothing slipped, add the next task. If it did not, you have lost two weeks, not a quarter.

The point is not to automate the most tasks. It is to take the drudgery off good salespeople so they spend their week on the calls that need a human. Choosing that first task well is most of the battle, and it is worth getting the use case right before you spend a penny on tools.

If you want help finding your team's 60% and setting up the first automation safely, book a session and we will map it to your sales week.