TFT: You're Better Than AI At These 5 Things
The five moves in a sale that stay yours, no matter how much you automate.
By David Roy ·
I let AI write the comment. I skimmed it. I posted it. It was a reply to somebody else's article. It failed miserably. I was trying to do too many replies at once and automate too much. It's the same problem I watch founders run into as they build out their sales process.
The tool had blended a few drafts of notes to different articles and confidently pushed them all into one paragraph that had nothing to do with the piece I was actually reading. My name on it. That one was on me. Every time you schedule an agent you open yourself up to the same thing. AI is better today than it was a year ago, and there's still real risk anytime you automate a human touch point.
That comment was a door. Someone whose work I respect had published something, I had a real reason to be in that thread, and it was the same kind of opening a good cold email gives you. I spent it on nonsense.
I still use AI every day. The reason I put it in the loop was to buy back time, so I could spend that time being human with customers. Somewhere in there it flipped, and the machine ended up standing in front of the person. Buyers can feel that. You've opened a message and known inside of one line that nobody was home — that this was just generic AI-generated outreach.
So I got specific about what I hand off and what stays mine. Reading what wasn't said. Adjusting mid-sentence. Naming the problem in the customer's words. Earning trust before pitching. Owning the outcome. Five things. AI still cannot do any of them.
1. You read what wasn't said
You'll notice when somebody stops taking notes and know you're losing them — a full minute before anything in their voice changes. Buyers almost never tell you that you missed. They keep nodding and saying "interesting." They ask a polite pricing question they have no intention of acting on. On the transcript it looks like a good call. The room was telling you something else, and you had about four seconds to catch it.
That part stays human. AI can write a strong version of your pitch. It cannot see the guy who folded his arms the second you said "implementation," or hear the half-beat pause before "yeah, that could work."
AI reads the words. You read the person.
The move feels awkward the first few times. Stop the deck. Say what you noticed out loud: "I think I lost you on that last piece." Then be quiet and let it sit. Half the time they tell you exactly what's wrong. No transcript was ever going to hand you that.
Where AI fits
Let it summarize the call afterward, then hold that summary against what you felt in the room. The gap between the two is your read — worth more than the summary. Prompt: "Here are my raw notes and the transcript from my last call with [client]. Summarize what was said and list the action items you can support. Do not tell me how the client felt or where they were interested. I'll fill that part in myself."
2. You adjust mid-sentence
There's a founder running his company almost entirely on AI agents. One of them, his HR manager, called a candidate for a job interview — a day early. The candidate was caught off guard: "Isn't this supposed to be tomorrow?" The agent agreed that it was. Then it asked its first screening question and ran the whole interview anyway.
The agent could hear the objection. It could even agree with it. What it couldn't do is the human thing: stop, recognize the plan is now wrong, and change the plan.
Your calls do this constantly. You open with the pricing conversation you agreed to last week and learn in the first ninety seconds that their budget got frozen on Friday. Or the champion who invited you is no longer at the company. The agenda you built is the wrong agenda now, and the rest of the call depends entirely on how fast you drop it. Run the script anyway and you become the vendor who wasn't listening. Nobody buys from that vendor twice.
Where AI fits
Have it build three potential directions for the call, before the call. When the ground moves, you're choosing a path instead of improvising one. Practicing different approaches prepares you for the unknown unknowns.
3. You name the problem in their words
Ask AI what problem you solve and you get a generic category: "Technical founders struggle to build a repeatable sales process." True, and almost useless in a specific room. Your customer's version of that problem has a name, and they're all different:
- The founder who closed four deals last quarter and can't tell you why three of them happened.
- The one who hired a sales rep who interviewed well and then never closed a deal.
- The one who has postponed a pricing conversation twice because they don't know how to show the value they've delivered.
- The one who grew 50% last year and has no idea how to do it again, because half the sales team turned over.
Say the problem back in the customer's terms and they start believing you can help solve it. That makes you worth their afternoon — and it comes entirely from handing them their own language back. Here's the test I use before every second call: could I say my version of their problem out loud to their co-founder, and would that person nod without needing me to explain it? If not, I have a category. I don't have a problem to solve yet.
Where AI fits
Paste your discovery notes and ask it to identify the problem your customer really wants to solve. Compare that to the problem you solve. If they're similar, you have a potential customer. If not, consider a longer-term follow-up or a second discovery call.
4. You earn trust by digging first
Every AI tool I've used has the same reflex: it wants to close fast. It leads with value props, then jumps to a meeting request or a pitch far too early. Which makes sense — it was trained on a million pieces of sales content written by people trying to close, so it does the average thing, confidently.
The problem is that assumes trust is already there. In reality trust wins more sales than any tactic. You ask a fourth question instead of taking the opening. You tell somebody their timeline is unrealistic. You say "I don't think we're the right fit for that piece" out loud, and you watch what it does to the room. All of that potentially costs you the sale in the moment — which is exactly why the tool won't do it. It doesn't know you're playing for a second and third conversation.
The founder who tells you the real number, the real blocker, the real reason the last vendor got fired is telling you because you spent a call earning trust. That usually won't happen on the first call, which is why you have to lead that conversation.
Where AI fits
Ask it for the ten questions it would ask before pitching. Throw out the seven you already know the answers to. The three left over are what you start the call with. Plan for that first call to just answer those questions and confirm the problem the customer is trying to solve.
5. You own the outcome
AI built the chart, but you gave it the numbers. AI builds the pitch deck, but you told it what you're pitching. That's the whole arrangement. It formats, it drafts, it cleans up the deck at eleven at night. The thinking about what your customer should do next quarter comes from you, because you're the one who's been inside their business.
Hand that part over and you can watch the output go generic in real time — recommendations that would fit any company in their industry, a roadmap of best practices nobody asked for. It reads fine and might even sound good, but it's worth nothing to the person paying you, because they could have generated the same thing in ninety seconds and they know it.
And when something goes sideways — the number is wrong, the timeline slipped, the integration took three weeks longer than promised — nobody in that room cares which tool made the slide. They're looking at you.
Ownership has no automation path. It's a position somebody has to be standing in.
This is the Improvement driver of the Revenue Flywheel, and it's the one most founders skip. You delivered what you sold. The question that decides your next twelve months is whether your customer is measurably better than they were two quarters ago, and whether you can point at exactly what did it: Outcome delivered → customer gets measurably better → you earn the right to solve the second problem → expansion. Nobody hands you the second problem because your follow-up cadence was tight. They hand it over because their business got better with you in it.
Where AI fits
Have it summarize the before-and-after record on every account. When a renewal is upcoming, you can share the value you've brought. Note: track real numbers from your customer — don't let AI make up the value.
Audit your own stack
Most founders can't tell you where AI sits in their sales process. They can tell you which tools they pay for. Those are very different questions, and the second one is the one that costs you deals. Paste this into whichever model you use, and answer honestly — which means listing the parts you're slightly embarrassed about.
You are auditing where I have handed human judgment over to AI in my sales process. Here is every place AI touches a customer or a customer-facing decision in my business: [list your tools and what each one actually does — drafting, research, outreach, comments and replies, call notes, proposals, reporting, follow-up]. Score each one against these five human edges: (1) reading what wasn't said; (2) adjusting mid-conversation; (3) naming the problem in the customer's own language; (4) earning trust by digging before pitching; (5) owning the thinking behind the outcome. For each task tell me: which of the five it touches, if any; whether AI is supporting me or standing in front of my customer; what my customer would notice if they knew AI did this part. Rank everything by the cost of getting it wrong. Start with the one most likely to lose me a deal in the next ninety days. Be blunt. Do not tell me the whole stack is fine."
The output you want is short and uncomfortable — one or two tasks where the machine ended up in front of a human who was expecting you. Then your call is: keep it as is, keep it with a human check, or take it back. Want a tool to help? The Revenue Leak Finder can audit your current sales process in under 10 minutes and show you the three steps to improving it this month.
Action step
Run the audit before your next customer call this week. Take the top result off AI for the next thirty days and do it by hand. If you skip this, you'll keep sending confident output nobody's home behind, and you won't find out until somebody stops replying.
Then pick one account you delivered for this quarter and write one sentence on how they're measurably better than they were in June. If you can't write that sentence, you have a renewal conversation coming that you're not ready for. Not sure which driver is leaking momentum? The Revenue Flywheel Diagnostic will show you in under ten minutes.
I still let AI draft my comments. I read them now, every time, before anything goes out with my name on it. That's the trade: let it work in the background so you get more hours in front of people, and keep the five things that were never the machine's to begin with.
Sales is just problem solving with another human.
This article was originally published on the ENG Sales newsletter.
Free Weekly Newsletter
Get the next one in your inbox
One sales-system upgrade every week, for founders still closing deals themselves.