Spotlight.ai in USA News: Qualification Theater
Spotlight.ai on why a full MEDDPICC record says almost nothing about whether a deal is qualified, and why the AI filling it in makes the problem harder to see.
We're excited to share that USA News just published a piece by Spotlight.ai's Lolita Trachtengerts, VP GTM Ops & Growth. It takes on the most popular AI build in RevOps right now, transcripts piped through ChatGPT into MEDDPICC fields, and explains why a pipeline full of green boxes keeps producing deals that close lost.
The article is called "You Don't Have a Forecast Problem. You Have Qualification Theater." And it makes a claim most sales leaders suspect but rarely put in writing: filling in a qualification field and proving it are two different jobs, and most teams have only automated the first one.
Every Box Was Green on the Deal You Lost
Here's a test to run before your next pipeline review. Pull a deal that closed lost last quarter. Go back sixty days, to before anyone knew it was dying, and read the qualification fields.
Economic buyer, identified. Pain, identified. Champion, confirmed. Decision criteria, captured. Every box green, right up until the deal went dark.
That opportunity was never qualified. It was populated. The fields were full because somebody, or something, filled them, and nothing on the record says whether any of it was ever true.
The Build Everyone Is Running
The setup comes up on almost every call right now. The team runs Gong, hands the transcript to ChatGPT, has it work out the MEDDPICC answers, and pushes the result into Salesforce. It's cheap, it's fast, and it's live this quarter with no procurement cycle.
It's also never trusted. Every version of this build keeps a human between the model and the CRM. One prospect described the reason in three words: garbage in, garbage out. When experienced operators build an automation and then refuse to let it run on its own, they're telling you something about the output.
What a Populated Field Actually Costs
Evidence and opinion end up in the same field. The buyer said their renewal cycle starts in March. The rep believes the buyer has budget. One is a fact and one is a hope, and once both sit in the same picklist they look identical. The forecast is built on a column that can't tell what was heard from what was assumed.
One call becomes the whole truth. A model reading a single transcript has no idea whether the champion said the same thing three weeks ago, or whether the economic buyer has shown up to a meeting since. One call is a claim. A claim that holds up across meetings, emails and who actually attends is evidence.
Nothing re-checks the answer. "Economic buyer: identified" gets written in March and sits there through Q2, through a reorg on the buyer's side, through the champion going quiet. It was true once. Nobody asks again.
"We spent a decade getting reps to fill in the boxes. Then we automated the filling in. Nobody automated the proving." Lolita Trachtengerts, VP GTM Ops & Growth, Spotlight.ai
The Real Problem Is the Question
Ask a general-purpose model which MEDDPICC box a conversation belongs in and it will always give you an answer. That's what it's built for. It has no way to return nothing, so when the evidence is thin you get a plausible sentence where a gap should be.
That's where the hallucinated field comes from. It also drives the cost: every follow-up question means reading the whole deal again and paying for every word written back. Fine for four deals in a pilot. A line item somebody questions across four hundred open opportunities.
Spotlight.ai is built the other way around. Every question the engine asks has a closed set of possible answers, defined before a model is ever called. Is this true, yes or no, with a probability. Which of these contacts is the champion, chosen from a known list. The model can't answer outside that set, so there's nowhere for an invention to go. Qualification status, deal score, gaps, risks and the forecast are then assembled from those answers in code. Computed, not generated.
What Changes When the Question Is Closed
The answer stops moving. Ask the same thing about the same deal on Monday and on Friday and you get the same answer, because it's computed from stored evidence. When the score changes, something in the deal changed.
Missing data shows up as missing. A question the evidence can't answer comes back with low confidence and lands on the deal as a visible gap. Knowing what you don't know is worth more to a forecast than a confident guess.
And the bill stops growing with curiosity. Nothing is being written at the decision layer, so a team can ask the pipeline a thousand questions this quarter without anyone doing the math on whether it was worth it.
Before You Scale Your Own Build
If you're in RevOps, look at any qualification field in your CRM and ask whether you can tell if it came from the buyer, the rep or a model. If all three look the same in your schema, your data quality work is measuring the wrong thing.
If you're a sales leader, ask whether you can find the moment a slipped deal stopped being qualified, or whether the record just stays green until the day it closes lost.
If you've been asked to build this in-house, you can, and it will demo well on one deal. Before you scale it, ask what happens on the four hundredth deal, who writes the questions, and what it does when it doesn't know.
See What Evidence-Based Qualification Looks Like
The Spotlight.ai MEDDICC Evidence qualifies every deal from what buyers actually said across calls, emails and meetings, keeps evidence and opinion apart, and puts the gaps on the deal where your managers can see them.
No re-prompting and no copy-paste. A qualification record you can trace back to the conversation it came from.
FAQs - Qualification Theater
Q: What do you mean by qualification theater?
A pipeline where every qualification field is filled in and none of it is proven. The boxes are green because a rep typed something or a model read one call, and there's nothing on the record that tells you which answers came from the buyer and which were assumed. It looks like rigor in the QBR. It behaves like a guess in the forecast.
Q: What's wrong with running Gong transcripts through ChatGPT to fill in MEDDPICC?
Nothing, as a starting point. It's clever and it's cheap. The problem shows up at scale. It reads one call at a time, it stores what the buyer said and what the rep hoped in the same field, and it never goes back to check an answer once it's written. Every team I've seen run it keeps a human in the loop, and that tells you how much they trust it.
Q: Why do AI models hallucinate qualification fields?
Because we ask them open questions. A general-purpose model is built to produce a readable answer, so it always produces one. When the evidence is thin, you don't get a blank. You get a confident sentence. Close the question, give it a defined set of possible answers, and there's nowhere for the invention to go.
Q: How is Spotlight.ai different from an in-house build?
We never ask an open question. Every question has a closed answer set defined in advance, and the conclusions, like qualification status, deal score and the forecast, are computed from those answers in code rather than written by a model. So the same deal gives the same answer twice, gaps show up as gaps, and every answer traces back to the conversation it came from.
Q: What's the first thing a RevOps leader should check?
Pick five closed-lost deals and read their qualification fields as of sixty days before close. If they were all green, your qualification data isn't telling you anything about risk. Then check your schema: can you tell a field the buyer confirmed from one a rep assumed? If not, that's the fix to start with.
Q: Does this replace Gong?
No. Spotlight.ai works with the recordings you already have, from Gong, Chorus or our own recorder. What it adds is the layer on top: qualification that holds up across every interaction in the deal, not a summary of the last call.
Comments