Six of the seven stages of selling are logistics

Sellers still prospect, prepare, approach, present, handle objections, close, and follow up. That seven-stage sequence has survived every technology wave since it was first written down. Six of those stages are logistics more than selling, and in 2026 AI handles logistics well.

Ask a sales team where deals die and you'll hear about brutal negotiations. Look at where leads drop out, though, and most deals die before anyone has talked at all. A lead comes in at 11pm and sits in a queue. By the time someone calls in the morning, it's cold.

Two infrastructure problems sit underneath all of this, and almost nobody has solved either one. The first is getting a call through spam filters. The second is follow-up that doesn't feel like spam. I'll get to both after the stages.

The seven stages of selling in 2026

Every sales motion still runs on these seven stages, from SaaS to auto dealerships to home services:

1. Prospecting: identifying the accounts and buyers who fit your ICP.
2. Preparation: researching the account, the buyer, and the trigger event.
3. Approach: opening the conversation quickly, with context that matters to the buyer.
4. Presentation: the live conversation where the buyer decides whether to trust you.
5. Handling objections: addressing pricing, fit, timing, and competitive concerns.
6. Closing: getting the commitment and handling the mechanics of the deal.
7. Follow-up and expansion: onboarding, retention, reviews, and re-engagement.

Buyers have changed around those stages. Gartner found that roughly three-quarters of B2B buyers would prefer to buy without a rep at all. HubSpot's State of Sales data consistently shows reps spending less than a third of their week selling.

Because each stage asks something different of the buyer, AI for sales ends up as six different jobs, run by one system on one shared contact record. All six sit around the one stage a person handles.

How AI handles each stage

This is the table we give new customers when they roll RevDesk out across the whole cycle.

The six AI stages share one contact record, which is where the gains build up over time. The lead you source at stage 1 today is the customer you re-engage at stage 7 next year, and everything on that record shapes the next touchpoint.

The presentation stays human

Stage 4 is the exception. AI can hold real conversations now. It qualifies leads and answers questions about price and fit in the prospect's language, within seconds. A buyer about to spend real money still wants to look a person in the eye, metaphorically or literally, and decide whether to trust them. AI can't stand in for that relationship. Deals are won in that conversation, and it's the one part of the cycle where you need a person.

A human sales touchpoint costs $50 to $500 once you count salary, tools, and overhead. The same touchpoint from AI costs pennies. If your reps spend their touchpoints on dialing, data entry, and scheduling back-and-forth, you're paying presentation-stage prices for logistics work. Making them faster at it doesn't change the price. Take the logistics away from them entirely.

Go back to the form that came in at 11pm. With RevDesk, the lead gets a call within seconds and answers three qualifying questions. They get an answer to their question about price, then book a slot on your closer's calendar. Your closer walks in with a transcript of everything the prospect said they wanted.

Your closer still has help on that call. The AI moves to the passenger seat and listens live. It surfaces the answer to the objection that just came up, a competitive fact, or the number the prospect asked about three touchpoints ago. Your closer builds the trust, which only a person can do, and never gets caught without an answer. Even a rep in their second week can handle objections like a five-year veteran.

Most AI calls show up as Spam Likely

Every AI sales vendor demos how natural its voice sounds. None of them demo the phone screen on the other end, because that screen says Spam Likely. That frustrates me. The most human-sounding AI in the world books nothing when nobody picks up.

US carriers score every calling number through analytics engines. Hiya, TNS, and First Orion score reputation for the major carriers. Unregistered numbers that place outbound volume get judged on behavior alone. One busy week and your best local number picks up a label that tanks every campaign after it. Most AI calling platforms run on resold SIP routes several layers away from the carrier. They can't see the label, much less fix it.

Nobody had built a platform where the calls get through, so we did the unglamorous thing and went direct. Our calls run on tier-1 carrier partnerships instead of SIP resellers. Every business goes through KYC vetting and gets registered with the analytics engines. That way the carrier knows exactly who is calling before behavior scoring makes something up. We monitor every number continuously. Numbers that were flagged before we got there go through remediation, where we petition those same engines to clear the labels.

None of this is exciting. It's paperwork, vetting, and carrier relationships, but it decides whether your AI has conversations or talks to voicemail. At Apple I watched the least glamorous engineering decide whether the magic landed. Deliverability is that layer for sales. Everyone else skipped it, which is what makes it a moat.

Follow-up that reads like a text from a person

The other broken moment is the follow-up text that reads "Hi! It's [Brand]! Reply STOP to opt out. Msg&data rates may apply." It shows up as a green bubble with compliance boilerplate doing the talking, and it gets deleted before the second sentence. The industry settled on a norm where the loudest line in the message asks the customer to leave.

Apple obsesses over how a message feels: the bubble, the thread, the read receipt. Sales software treats messaging as a blast channel. Nobody was bringing iMessage to the application layer, so we did. RevDesk follow-up goes out over iMessage as a first-class channel, alongside voice, SMS, WhatsApp, and email. Your lead gets a blue bubble in a real thread, from an identity they can reply to. From there, the AI carries the conversation.

The legal machinery is all still there. Consent and opt-outs are enforced on every message, the same way DNC and TCPA checks run on every call. Compliance just stops being the personality of the message. Engagement on iMessage runs 3 to 4 times higher than SMS, because your follow-up lands in the app your customers answer and looks like something a person would send.

The first 30 days

Rolling AI across six stages changes your workflow one piece at a time. We walk new customers through it in this order over their first 30 days:

• Audit each stage and list what a person does by hand today: dialing new form fills, sending the recap, chasing the no-show, asking for the review. That list is your automation map. It's always longer than anyone expects.
• Wire the triggers before you touch scripts. Speed to lead only works if the AI hears about a lead the moment it exists, so connect your forms, ad platforms, and CRM first.
• Define the handoff line. Decide exactly what qualified means for you (budget confirmed, timeline stated, decision-maker on the calendar). The AI hands off at that line and never past it.
• Close the loop. Review every transcript, objection, and no-show reason, then tune the timing and the playbook from real data instead of gut feel.

Three constraints to plan around

• Consent comes first. The calls that work, and that stay within the law, are follow-ups to people who engaged or gave consent. Cold dials to strangers fail on both counts. Your system has to enforce TCPA windows, consent records, and DNC checks on every call.
• Number reputation needs constant upkeep. Registration and KYC get you clean numbers. Keeping them clean takes per-number volume limits and continuous monitoring. There's no one-time fix, which is why platforms that skipped the infrastructure can't bolt it on later.
• AI does not close deals. If you expect it to run the presentation too, you get demos nobody attends and deals that stall at verbal. Let it get the conversation onto your rep's calendar and stop there.

Where AI doesn't belong

Keep AI away from these four, because they're where the deal is won:

• The presentation itself. When the buyer shows up to decide whether to trust you, a person should show up too.
• Live negotiation. Pricing counters and executive escalations belong in a real-time conversation between people.
• Relationship repair. When something has gone wrong and the customer is angry, the fastest way to calm things down is a person who can own the problem.
• Any contact without consent. With no lawful basis to reach someone, you don't reach them, whatever the technology makes possible.

Make AI the default for six of the seven stages and protect the seventh. Everything we automate is there for that one conversation. The carrier partnerships and the iMessage thread make it more likely to happen, and the coach in your rep's ear makes it better when it does.