Adam Rubenstein, CEO of TRAQ, makes one central argument: your sales conversation data is valuable, but it’s trapped inside an application most of your company never opens.

Here’s the core problem he’s solving:

TRAQ already analyzes sales calls well. But the insights sit behind a login. Sales leaders check it. Marketing, Product, and Finance largely don’t.

The fix is an MCP (Model Context Protocol), an open standard that lets AI assistants like ChatGPT or Claude pull directly from TRAQ’s analyzed conversation data, without anyone opening TRAQ at all.

The key distinction he draws: most MCP announcements you’ll see this year will just expose raw transcripts, essentially giving AI a searchable pile of text. TRAQ’s MCP exposes analyzed intelligence, already scored against your sales methodology, your definition of good, your required questions.

The practical payoff: Anyone in your company can ask a natural question like“Which deals are missing an economic buyer?”or“What are we losing on this quarter?”and get an answer built from what buyers actually said, not from CRM dropdowns or salesperson memory.

The bottom line he lands on: every other system tells AI what happened. TRAQ tells it why.

“Data is a precious thing and will last longer than the systems themselves.”
— Tim Berners-Lee

I keep coming back to that line.

Almost every piece of software I bought in my first twenty years of selling is gone now. The CRMs changed. The dialers changed. The reporting tools changed. What our customers told our salespeople never changed. It was true when they said it, and most of it is still true today.

The trouble isn’t the information. The trouble is where we keep putting it.

Recently I did something I suspect a lot of sales leaders have tried. I opened an AI assistant and asked it a real question. Not a research question. A Monday morning question.

“Why are we losing to our biggest competitor this quarter?”

It gave me a thoughtful, well organized, completely useless answer. It talked about pricing pressure and differentiation and discovery. Reasonable advice. Generic advice. The kind of advice you could give any company selling anything to anyone.

That wasn’t the AI’s fault.

The answer to my question was sitting inside hundreds of sales conversations the assistant had never been allowed to hear.

That’s what I want to write about. Not whether AI can understand a conversation. It can. We built a company on that. The question now is where the intelligence is allowed to go once it exists.

I keep coming back to that line.

Almost every piece of software I bought in my first twenty years of selling is gone now. The CRMs changed. The dialers changed. The reporting tools changed. What our customers told our salespeople never changed. It was true when they said it, and most of it is still true today.

The trouble isn’t the information. The trouble is where we keep putting it.

Recently I did something I suspect a lot of sales leaders have tried. I opened an AI assistant and asked it a real question. Not a research question. A Monday morning question.

“Why are we losing to our biggest competitor this quarter?”

It gave me a thoughtful, well organized, completely useless answer. It talked about pricing pressure and differentiation and discovery. Reasonable advice. Generic advice. The kind of advice you could give any company selling anything to anyone.

That wasn’t the AI’s fault.

The answer to my question was sitting inside hundreds of sales conversations the assistant had never been allowed to hear.

That’s what I want to write about. Not whether AI can understand a conversation. It can. We built a company on that. The question now is where the intelligence is allowed to go once it exists.

THE REAL QUESTION

It’s no longer whether we can understand a conversation. It’s whether anyone outside the application can use what we learned.

Business software has always rested on one quiet assumption. The person comes to the software.

Log in. Find the dashboard. Pick the date range. Run the report. And before any of that, remember the thing exists at all.

That assumption worked fine when software was the destination. It works a little worse every month.

Let me be honest about our own product for a minute. TRAQ analyzes conversations very well. It finds the discovery question that never got asked, the objection that never got resolved, the customer who is quietly unhappy. Then it puts all of that on a screen, and waits for somebody to come look.

Sales leaders come look. Good managers come look.

Marketing doesn’t. Product doesn’t. Finance doesn’t. The CEO sees a slide about it once a quarter, in a deck that took somebody two days to build.

So the intelligence exists, and most of the company never touches it.

I’ve managed enough salespeople to know how this usually goes. Every new application you roll out gets the same reception. Another login. Another tab. Another thing somebody has to remember to check. You can practically hear the groan from across the sales floor.

We believe it doesn’t have to be this way.

THE ISLAND PROBLEM

Intelligence that requires someone to remember to go look for it is intelligence most of your company will never use.

MCP stands for Model Context Protocol. It’s an open standard, and it has become the common way AI systems connect to business information.

Now let me explain what that actually means, because the name is terrible.

Think about the best sales operations analyst you ever worked with. The one who knew your numbers cold. You never had to tell that person which report to run or which field to filter on. You just asked. “How did the enterprise team do on renewals last quarter?” And they came back with a real answer, because they understood the question and they knew exactly where to look.

Think about what made that person valuable. It wasn’t that they had the data. Everybody had the data. It was that they knew what questions the data could answer, and they knew how to go get it.

That’s the job an MCP does for AI.

TRAQ uses it to introduce itself to an AI assistant. It says, in effect: here is what I know about your conversations, here is what you’re allowed to ask me, and here is how to ask it properly.

From that point forward, the assistant can answer questions about your customers the same way it answers anything else you type into it.

Two things it is not.

It is not a copy of your data. Nothing gets exported, duplicated, or shipped off to some new warehouse to sit. Your conversations stay in TRAQ. The AI knocks on the door and asks a question, and TRAQ answers it.

And it is not a search box. A search box requires you to already know what you’re looking for. Here you ask the question you actually have, in the words you’d use with a colleague, and the AI works out how to get it.

What That Looks Like

Four examples, and what happens behind the curtain.

You ask: “Which of my accounts brought up pricing more than once in the last ninety days?”

The assistant asks TRAQ for conversations at your accounts inside that window, narrows to the ones where pricing came up, and groups them by account. You don’t see any of that happen. You see a list. The accounts, how many times it came up, and what was actually said.

You ask: “Is Zach getting better at discovery?”

The assistant asks TRAQ for Zach’s discovery scores over time, then for the calls underneath the trend. You get a straight answer. Yes, he’s moved from a 6 to an 8 over two months, and here are the three calls where it turned.

You ask: “Draft a follow-up email to the Riverside account based on our last conversation.”

The assistant asks TRAQ what happened on that call. What the buyer said they wanted. What they objected to. What never got resolved. Then it writes the email from what was actually said, instead of from what you half remember on a Thursday afternoon.

You ask: “Run MEDDPICC on the Riverside opportunity.”

This is my favorite one, so let me slow down on it.

There are eight conversations attached to that opportunity. A discovery call in April. Two  technical deep dives. Three follow-ups with five different people. A demo. And a call last week that the regional manager sat in on.

Nobody at your company has heard all eight. Your salesperson lived them, but he lived them one at a time, over four months, and he remembers the good parts.

The assistant asks TRAQ for every conversation tied to that opportunity and evaluates all of them against MEDDPICC. Not the summaries. The conversations.

What comes back is a scorecard. More useful than the scorecard is the gap list.

Metrics. Strong. In the second call the VP of Operations said they’re losing roughly 40 hours a week to manual reporting. She put a number on it herself.

Economic Buyer. Missing. The CFO has been named twice and has never been in a conversation. Your salesperson has been selling to the people who feel the pain, not to the person who signs.

Decision Criteria. Partial. You know three things that matter to them, and you learned two of them from the champion. That means you’re hearing one person’s version of what matters.

Decision Process. Missing. In eight conversations, nobody has asked what happens after they choose a vendor.

Paper Process. Missing. Legal, procurement, and security review have never come up once. On a deal this size, that’s usually where the timeline goes to die.

Identify Pain. Strong.

Champion. Likely, but untested. The Director of Sales Operations has advocated for you internally on two occasions that she described on calls. Nobody has ever asked her to do something hard.

Competition. Weak. A competitor was named in April and never mentioned again. Nobody has asked where else they’re looking.

Now look at what you just got.

If you asked your salesperson how Riverside is going, he’d tell you it’s in great shape, and he wouldn’t be lying to you. Discovery was strong. The pain is real. The champion likes him.

But nobody has met the economic buyer. Nobody knows how the decision actually gets made. Nobody knows what the paper process looks like. And nobody has asked about the competition since April.

That’s a forecast problem you found in about ten seconds, instead of in the last week of the quarter.

Here’s the part I care about most. That answer didn’t come from asking your salesperson how the deal is going. Salespeople are optimistic. It’s a job requirement, otherwise they’d be defeated by all the no’s they hear. It came from what the buyer actually said, out loud, across six conversations.

And it doesn’t have to be MEDDPICC. Run MEDDIC, BANT, SPICED, Sandler, Challenger, or the homegrown methodology your VP of Sales wrote himself and swears by. The point isn’t the framework. The point is that your framework finally gets applied to every deal, by something that actually listened to all of it.

Now scale that question up. Instead of one opportunity, run it across every deal in the forecast.

Which deals are missing an economic buyer? Which ones have never had a paper process conversation? Which ones have a champion nobody has tested?

That’s not a pipeline review anymore. That’s a pipeline inspection, and it’s built on evidence instead of opinion.

That’s the one I’d pay for.

IN PLAIN ENGLISH

You ask a question about your customers the same way you’d ask ChatGPT or Claude anything else. The AI figures out where to look inside TRAQ, gets the answer, and hands it back to you. You never open TRAQ at all.

One more thing worth saying, because some of you are going to get asked about it. This is not a science project. MCP has a published specification, a regular release cadence, and support across the major AI platforms. Building on it is a bet on a standard, not a bet on a vendor.

Your AI assistant can probably already reach a lot of your company’s information. Email. Documents. Calendars. Support tickets. Your CRM.

Look at what all of that has in common.

It is the record of what happened. Stages and amounts and dates and statuses. Notes that somebody typed, or, more likely, didn’t.

The why has always lived somewhere else. It lived in the conversation. And the conversation has always been the one room nobody could get into.

I’ve said before that not everything in a conversation is important, but almost everything important can be found within the conversation. That isn’t a marketing line. It’s the reason our customers buy from us.

So when you connect conversation intelligence to the AI layer, you are not adding one more source to the pile. You are adding the only source that explains all the others.

Here is the difference in practice.

A CRM can tell AI that the deal was lost on March 12, and that somebody picked “price” from a dropdown menu.

TRAQ can tell it that the buyer raised a concern about implementation in the third meeting, that the salesperson gave a weak answer, that the subject never came up again, and that the same thing happened in nine other deals you lost last quarter.

One of those is a record. The other is an explanation.

WHAT MAKES IT DIFFERENT

Every other system tells your AI what happened. TRAQ is how it learns why.

I’d rather show you than describe it. These are the kinds of questions that stop being research projects and start being conversations.

If you run the revenue organization. “What is the biggest risk to our number this quarter, and what’s driving it?” Board preparation that used to take two weeks and three analysts becomes something you do on a Tuesday morning.

If you manage a team. “What should I coach each of my people on before one-on-ones this week?” The answer shows up where you already work, based on what your salespeople actually did, not on the two calls you had time to review.

If you run product. “What are prospects asking for that we don’t build, ranked by how often it shows up in the deals we lost?” The roadmap meeting stops being a contest between anecdotes.

If you run marketing. “What words do our buyers use when they describe this problem?” Not the words we use. Theirs.

If you own retention. “Which customers are showing increasing signs of dissatisfaction across their last three conversations?”

Let me stay on that last one, because it’s the problem that started all of this for me.

Picture an account at Company Z. In February, an operations manager mentions on a call that onboarding took longer than she expected. In April, a different contact says his team hasn’t really adopted the reporting. In June, the executive sponsor says, almost in passing, that they are “taking a look at the market.”

Three warning signs. Three conversations. Three different people at your company heard them, and each one dropped a note into a different record.

In an ideal world, somebody connects those dots. In the real world nobody does, and the first time your organization finds out is when the renewal doesn’t come back.

That is one problem, and it got treated like three unrelated events.

An MCP is how those dots get connected by whoever thinks to ask, instead of by whoever happens to log in.

And then there’s the part I’m most excited about.

The real unlock isn’t TRAQ answering a question on its own. It’s TRAQ answering alongside everything else your assistant can already reach. Conversation intelligence, plus CRM history, plus support tickets, plus billing. One question. One answer. Drawn from all of it.

That is the moment conversation intelligence quietly turns into business intelligence.

And once an AI can ask a question, it can act on the answer. Draft the follow up that addresses the concern the buyer actually raised. Flag the account before the renewal is in trouble. Update the CRM field your salesperson was never going to update anyway.

Salespeople love selling. They don’t love documenting their sales. We’ve been saying that for years. This is how we finally stop asking them to.

I want to be careful here, because a lot of companies are going to announce an MCP this year.

Most of them will expose transcripts.

Understand what that actually gives you. It gives an AI a very large pile of text and the ability to search it. That’s better than nothing. It is not intelligence. Ask it why you’re losing and it will helpfully find the word “expensive” four hundred times.

What TRAQ exposes is different, because it has already been analyzed. Scored against what your company decided good looks like. Your methodology. Discovery quality. Objection handling. Sentiment. Risk signals. The five questions you require on every single call.

We’ve always said we want to deliver customized intelligence without custom software. That doesn’t change when the intelligence leaves the building. An AI asking TRAQ a question gets an answer shaped by your definition of good, not a generic one.

And it isn’t limited to one call. With 100 salespeople you might have thousands of conversations every month. No manager can listen to all of them. No analyst can connect all of them. Cross-Call Analysis can, and an MCP is how anybody in your company gets to ask about all of it at once.

RAW VERSUS READY

A transcript tells an AI what was said. TRAQ tells it what that meant, measured against what good looks like at your company.

An MCP doesn’t make bad data good. If your conversations aren’t being captured, there’s nothing to ask about. That’s why we’ve worked so hard at capturing the whole sales narrative, including the phone calls and the in-person meetings that never show up on anybody’s Zoom calendar.

It doesn’t replace judgment. It gets you to the right question much faster. It doesn’t decide what you should do about the answer.

It doesn’t replace the manager either. Coaching is still a person sitting down with a person.

And an MCP by itself is not a strategy. Plenty of companies are going to ship one this year because it’s the acronym of the moment. That isn’t why we’re doing it. We’re doing it because your conversation data should be usable, and it shouldn’t be locked inside an application. Including ours.

I’ve written before that a feature list tells you who has the most boxes checked this morning. It doesn’t tell you who will be ready in two years.

Evaluate the trajectory.

Nobody knows which AI assistant your company will standardize on three years from now. I don’t. Your CIO doesn’t. That is exactly the argument for building on an open standard instead of a private connection to whichever vendor happens to be winning today.

FUTURE-BOUND, IN PRACTICE

I can’t tell you which AI your company will be using in three years. I can make sure it knows what your customers have been telling you.

I asked an AI assistant why we were losing, and it couldn’t tell me, because the answer was sitting in a place it wasn’t allowed to reach.

That’s the whole thing. It isn’t a protocol or an acronym. It’s a pretty simple idea about ownership.

Your company has years of accumulated knowledge sitting inside calls and meetings. Your customers told you what they think. Your prospects told you what they want. Your best salespeople showed you exactly how they sell. Your market has been talking to you every single day.

Historically, we threw almost all of it away.

More data leads to better insights. Better insights lead to better decisions. Better decisions lead to better sales results. That chain only works if the intelligence can travel.

That’s why we’re building an MCP.

If you’d like to see what TRAQ can uncover inside your own conversations, schedule a demonstration and let us show you.

— Adam Rubenstein Cofounder & CEO, TRAQ

About the Author

Adam Rubenstein is the CEO of TRAQ, a conversation intelligence platform for sales and customer-facing teams. He works with sales leaders to turn real conversations into structured insights, repeatable coaching, and measurable improvement, helping teams execute consistently and scale what works. Connect with Adam on LinkedIn or learn more at traq.ai.