When I moved from running marketing to also owning sales, I expected AI to help most with outreach. It turned out to help least there. The real gains came before and after the conversation, in the preparation, the follow-through and, most of all, in understanding the customers we already had.
Here's how I think about AI in a B2B sales motion with long cycles, several stakeholders and serious deal sizes.
Where AI genuinely helps
1. Account research that starts the conversation at level two
Before a first meeting, an agent gathers what we know about the company, its recent news, its likely priorities and the people in the room, then turns it into a one-page brief. A good seller used to spend an hour on this. Now they spend ten minutes reviewing it, and the hour goes into thinking about the customer's actual problem.
2. Demo preparation
Enterprise demos fail when they're generic. AI helps map a prospect's stated challenges to the parts of the product that solve them, so the demo follows their story rather than our feature list.
3. Follow-up that never slips
Most deals don't die from a "no". They die from a quiet week that becomes a quiet month. AI drafts the meeting summary, the agreed next steps and the follow-up email within minutes of a call. A person reviews and sends it.
4. Proposal and business-case drafts
The first draft of a proposal is mostly assembly. AI handles the assembly; the seller handles the judgement on scope, pricing and risk.
Where AI quietly hurts
Fully automated outreach is the obvious trap. Buyers in industrial and enterprise markets receive a lot of mass-personalised messages, and they can tell. A message that mentions their company name but misunderstands their business does more damage than no message at all.
The subtler risk is losing your own understanding. If an agent always writes the brief, sellers can stop really learning their accounts. We treat AI output as preparation, not knowledge. The seller still has to be able to explain the account without the brief in front of them.
AI should make you more prepared for a conversation, never less present in it.
The opportunity most teams are sitting on: cross-sell
Here's the part I wish I'd done years earlier. Most B2B companies sell more than one product, and most customers buy only some of them. The data showing who uses what usually exists somewhere: in contracts, invoices, support tickets and usage logs. Almost nobody looks at it systematically.
We built a simple workflow with n8n and Claude that reads our customer base and answers one question for every account: which of our products would logically help this customer next, and why? The first run surfaced more than thirty expansion opportunities. None of them needed new marketing spend; they needed a conversation with someone who already trusted us.
A simple cross-sell check you can run this month
- List every customer and the products or modules they use.
- For each product, write down which customer problem it solves and which other product it naturally leads to.
- Ask AI to flag accounts that use the first product but not the second, with a one-line reason.
- Have the account owner review the list and pick the five strongest. Then call them.
How the role of the seller changes
The seller's job is shifting from collecting and assembling information to judgement and trust: reading the room, understanding what a plant head is really worried about, knowing when to push and when to wait. Those skills were always the valuable part. AI just removes the excuse of not having time for them.
If you lead a sales team, the question to ask isn't "how many tasks can AI take over?" It's "what would my best seller do with an extra day every week?" Then build towards that.