Selling is a human skill. The tools are how you buy time to use it.
That sentence is the whole editorial position of this site, and it cuts against most of what gets written about AI and sales in both directions. The people claiming AI replaces the rep are wrong. So are the people insisting nothing has changed.
Two disclosures first
This site is not affiliated with Anthropic. It is independent and is not authored, sponsored, reviewed, or endorsed by them. Claude is their product and their trademark. We write about using it. If something here is wrong, the mistake is ours.
We have a commercial interest. This site is maintained by the team behind Salesgear, a sales platform, and some pages link there. You should factor that in the way you would with any publication that has a parent. What we will not do is pretend a tool solves a problem it does not, because the readers we want are the ones who would notice.
Who wrote this
Sales people, with decades between us. We have sold through several supposed revolutions and adopted most of them: dialler floors, marketing automation, sequencers, social selling, workflow automation, and now this. Long enough to have seen the pattern repeat, and long enough to be sceptical of the version of the story where a tool does the selling.
Every one of those waves produced a real edge, and it went to whoever adopted it before it was obvious. Direct mail rewarded the person with the better list. The fax rewarded whoever reached the desk first. Email rewarded sequencing and measurement. LinkedIn rewarded relevance over reach. Automation rewarded the person who stopped doing work that never needed a human. Each edge was real, and each one lasted right up until everybody had it.
Which is the part worth internalising about this era. The edge is not that you use Claude. Everyone will. The edge is the eighteen months before that is true, and what you build into a repeatable system while it lasts.
What has never changed
Across all of it, the hard part has stayed exactly where it was. A person with a budget, competing priorities, and no reason to trust you has to decide that this is worth their risk. That is a human judgement made about another human.
Which means these have never been automated, and are not being automated now:
- Asking the uncomfortable question, then staying quiet long enough to get the answer.
- Knowing when the stated objection is not the real one.
- Telling a prospect their timeline is unrealistic, and keeping the deal.
- Deciding this is not a fit and walking, when the quarter says otherwise.
- Being trusted enough that someone stakes their own credibility internally on you.
Sales is a forever problem, not a workflow with a solution pending. Every generation of tooling gets sold as the end of it, and every generation discovers it moved the work rather than removing it.
What actually changed, stated narrowly
One thing, and it is genuinely valuable: a model can read context no human had time to read. Six months of tickets. Forty closed-lost notes nobody has reopened. A 90 minute transcript. The email thread from March that would have changed how you opened the call.
That is why the pages here are organised by stage of the funnel rather than by feature. Time returned at the research and admin end is time available at the human end. That is the entire trade, and it is a good one. It is not a transformation.
Two things we will keep saying
The old playbook is mostly gone, but not entirely
The predictable pipeline model that ran B2B software for a decade assumed volume was a moat. It stopped being one when everyone got the same tools on the same day. What survived is the unglamorous half: write down who you sell to, measure the funnel by stage, and follow up past the point that feels comfortable. That part still works and most teams still skip it.
The most underrated tool in the AI era is a good rep
A large amount of money in this category is spent by people measured on having bought something rather than on what it changed. The tool gets approved, the seats get counted, and nobody checks in month four whether one person works differently.
The antidote has not changed either. One workflow, one team, one month, one number. If it does not move, stop. That is why every page here says what a prompt needs and where it fails rather than promising a percentage. An honest audit of where your team actually is beats another tool almost every time.
So we log it, weekly
The two halves of that argument, that fundamentals hold and that tools keep getting faster, only reconcile one way: keep testing, and write down what happened. An opinion about whether some new thing is real is worth very little. Running it against a live pipeline and recording the result is worth a lot.
The log is that record, updated weekly. What we tested, what the number came out at, what we have stopped recommending, and where we were wrong before. Every entry says how we know. Anything that proves durable gets folded back into the stage pages, so those stay current rather than becoming an archive with a fresh date on top.
Use any of it
Copy the prompts, put them in your enablement docs, rename them, take credit. The prompts were never the moat and neither is this site. If you are starting, start at prospecting and work forward. If you already use Claude daily, closing is where most people are leaving the most on the table, because it is the stage where an agreeable assistant is worse than none.