AI for Sales: Where It Actually Moves the Number

AI for sales means using AI inside the sales process itself: researching accounts, preparing for calls, capturing what was said, keeping the CRM accurate, drafting follow-ups and proposals, coaching sellers and forecasting the pipeline. Its job is to give sellers back the hours that currently go on everything except selling, and to make the hours that remain better informed.

That is a narrower claim than most vendors make, and it is the one the evidence supports. AI moves the number in sales where it removes friction around the conversation. It tends to make things worse where it is pointed at the conversation itself.

The strategic decisions that sit above this, which customers to pursue, how they prefer to buy and which motion to run, are covered in the go-to-market hub for this silo. This briefing is about what happens once those are settled.

Where a seller’s week actually goes

Salesforce’s seventh State of Sales report, a survey of 4,050 sales professionals conducted in August and September 2025, found sellers spend 40% of their working week actually selling and 60% on everything else, from data entry and quoting to internal coordination. Almost a full day a week goes on prospecting, and 48% say they still lack the bandwidth for adequate outreach.

Salesforce sells AI to sales teams, so read its enthusiasm with that in mind. But the time split is consistent with what any sales manager sees, and it defines the opportunity precisely. The biggest gains are not in making sellers more persuasive. They are in giving back some of the 60%.

The four places AI moves the number

Research and call preparation. Before a first conversation, a good seller wants to know the company’s situation, recent changes, the likely priorities of the people attending and what was said last time. Assembling that by hand takes long enough that it is often skipped. AI assembles it in minutes. Sellers in the Salesforce survey expect AI agents to cut prospect research time by 34%, which is expectation rather than measurement, but the direction matches practice.

Capture and CRM hygiene. Conversation capture that produces a summary, the agreed next steps and the updated record, reviewed by the seller rather than typed by them, removes the task sellers most reliably neglect. That matters beyond the seller’s time, because forecasting, handovers and management attention all run on the record. A CRM filled in by exhaustion at six on a Friday is a CRM nobody trusts.

Follow-up and first drafts. Follow-up is where deals quietly die, because it depends on the one person who could send it being free to send it. AI drafts the recap, the tailored resource and the proposal outline while the conversation is fresh, and the seller edits rather than starts from nothing. Salesforce’s respondents expect a 36% cut in time spent drafting. The editing step is not optional, for reasons covered below.

Coaching and deal review. Most sellers get little feedback on their actual conversations: in the same survey, 46% of the youngest sellers said they rarely get feedback on their calls. Recorded and summarised conversations let a manager review five calls in the time one used to take, and let a seller rehearse a difficult conversation before having it. This is the least glamorous use and often the most valuable, because it improves the part of selling AI does not do.

Where it breaks

The failure modes are as predictable as the gains.

Automated volume. The easiest thing to automate is outreach, and it is the most damaging. Gartner’s survey of B2B buyers found 73% actively avoid suppliers who send irrelevant outreach. More messages at lower cost per message can raise the cost per customer, because each irrelevant contact spends reputation with exactly the accounts you want. Cold email still works, but only where the research behind it is real.

Inconsistent claims. Gartner also found 69% of B2B buyers report inconsistencies between a company’s website and what its sellers tell them. AI-drafted emails and proposals that nobody checks against the company’s actual positioning, pricing rules and commitments widen that gap. A generated promise is still a promise.

Bad data in, confident forecast out. AI forecasting is only as good as the pipeline record underneath it. A model trained on stage data that sellers update optimistically produces an optimistic forecast with a precise-looking number attached, which is worse than a rough one everyone knows is rough.

Tool sprawl. Salesforce found the average sales team runs eight standalone tools, and 51% of sales leaders say disconnected systems are holding back their AI efforts. Each new assistant adds a login, a data silo and a place where the record diverges. Wiring AI into a workflow that already exists beats adding another window.

What stays with the seller

Gartner’s research on buyer preferences draws the line clearly. Buyers are happy to self-serve for general information, but they prefer a seller’s input for the tasks that need contextual judgment, above all working out whether an offer genuinely fits their situation.

That is the job that remains. Diagnosing whether this buyer’s problem is the one you solve. Hearing what was not said. Knowing when to walk away from a deal that would close but go badly. Negotiating price and scope with someone who is weighing their own career against the decision. AI can prepare a seller for every one of those moments and summarise them afterwards. It cannot be in the room for them, and buyers can tell when it has been asked to try.

For founder-led firms there is a second point. If the founder is still the only person who can close, AI does not fix that on its own. It makes the founder faster, which is useful, but the lasting fix is a pipeline that does not depend on one person, with AI carrying the preparation and follow-up that used to require them.

How to tell whether it worked

Measure the outcome, not the activity. Emails sent, summaries generated and seats licensed all rise automatically once the tools are switched on, which is exactly why they prove nothing.

The measures that matter are the share of seller time spent in conversations with buyers, time from enquiry to first meaningful response, win rate on qualified opportunities, sales cycle length, and forecast accuracy against what actually closed. Take a baseline before rollout. If the share of time spent selling has not moved after a quarter, the tools have been added to the job rather than taken away from it.

Where to start

Start with capture and follow-up, not outreach. They are the lowest-risk uses, they touch every deal, and they fix the record that everything else depends on. Once the CRM reflects reality, add research and call preparation. Leave automated prospecting until last, and only once there is a clear definition of who is worth contacting and why.

Throughout, keep one rule: nothing generated reaches a buyer without a seller reading it first. That single habit prevents most of the ways AI in sales goes wrong.

If your sellers still lose their week to everything except selling, tell Daniel how a deal moves through your business today, what you have already automated, and where the number needs to go. Book 30 minutes with Daniel.

Frequently Asked Questions

How is AI used in sales?

Mainly around the sales conversation rather than in it: researching accounts, preparing for calls, capturing and summarising conversations, updating the CRM, drafting follow-ups and proposals, coaching sellers and forecasting. The aim is to give back time spent on non-selling work, which Salesforce’s 2026 State of Sales report puts at 60% of a seller’s week.

Will AI replace salespeople?

Not for sales that depend on judgment about fit. Gartner found B2B buyers prefer self-service for general information but still want a seller’s input when deciding whether an offer suits their situation. AI takes over preparation, record-keeping and first drafts, and leaves diagnosis, negotiation and the relationship with the seller.

What is the best first use of AI for a sales team?

Conversation capture and follow-up. They are low-risk, apply to every deal, and fix the CRM record that forecasting and management depend on. Research and call preparation come next. Automated prospecting should come last, once the team has a precise definition of who is worth contacting and why.

Does AI-generated sales outreach work?

Only when it is genuinely relevant. Gartner found 73% of B2B buyers actively avoid suppliers who send irrelevant outreach, so cheap volume can damage your standing with the accounts you most want. AI is valuable for the research that makes outreach relevant, and risky when used simply to send more.

How do you measure the impact of AI on sales?

Measure outcomes rather than activity: the share of seller time spent with buyers, time to first meaningful response, win rate on qualified opportunities, cycle length and forecast accuracy. Take a baseline before rollout. Counts of emails sent or summaries generated rise automatically and prove little.

What are the risks of using AI in sales?

The main ones are damaging your reputation with automated irrelevant outreach, sending buyers inconsistent claims in unchecked AI drafts, trusting forecasts built on poor pipeline data, and adding disconnected tools. Salesforce found sales teams already average eight standalone tools. Requiring a seller to read everything before it reaches a buyer prevents most of these problems.