Adoption Is Not Transformation: A Conversation with Jonathan Kvarfordt

Jonathan Kvarfordt has spent over fifteen years deep inside the revenue engine, building enablement, sales acceleration, and go-to-market programs for Fortune 500 companies and high-growth startups. Today he is VP of go-to-market at Momentum, a conversational data platform, and the founder of GTM AI Academy, where more than ten thousand people have learned how to turn AI from a buzzword into real operational impact. He launched the academy in December 2022, when plenty of his peers were still calling AI a fad on the way out, like NFTs.

What makes Jonathan worth listening to is that he did not start in AI. He came up through revenue operations and enablement, which is exactly why his take is grounded rather than hyped. This conversation is about what AI is actually changing inside go-to-market teams, why most companies still cannot operationalize it, and what separates the people running real experiments from the people posting theoretical wins with no receipts.

AI is not an easy button

The biggest misunderstanding Jonathan sees is that AI is an easy button you drop in and suddenly get 50 percent more pipeline and 20 percent more revenue. It does not come that easily. What AI actually does is force you back to the fundamentals of your business, your customer experience, and your own job, so you can figure out what the real process is, what genuine difference a human makes, and what can actually be automated.

He points at a structural shift underneath all of it. Google released a web protocol to help AI agents traverse the internet, because the web was built for humans, not agents. People in banking say the same thing about financial systems. The theme keeps repeating: the world was not built for agents, it was built for humans, and now you are increasingly selling to, buying from, and building for two kinds of entities at once. How that plays out depends entirely on your industry and how advanced your ideal customer is.

Adoption is not transformation

Jonathan is blunt that the adoption statistics are misleading. McKinsey and Gartner will tell you most organizations are 80 or 90 percent adopted on AI, but that number is companies throwing a copilot license at someone who has no idea how to prompt it. Throwing a license at a person is not transformation. It is just another unused seat.

His own company ran the analysis on thousands of their sales conversations to see how many organizations had actually operationalized AI rather than just touched it. In the first half of 2025, only about 7 percent were even beginning to do something operationally real. By the second half, that had grown to 24 percent, across everything from 20,000-person enterprises down to startups. The 7 percent more than tripled in under six months, which is a good direction, but it means the vast majority are nowhere near operationalized. Getting there requires a willingness to tear apart what you do, break it, and rebuild it, and when you do that you can accelerate far past where mere optimization would take you.

Optimize, or actually reinvent

When Jonathan started three or four years ago, his pitch was that he would help teams do what they already do, but better, faster, easier, and cheaper. About a year in, he realized that was the wrong frame. The real opportunity is not doing the old way better. It is doing things totally differently, which demands you think in a genuinely different way. For enterprises with robust processes, data silos, security, and governance, that gets complicated fast, which is why you cannot just throw AI at it and hope.

His example is Sendoso, a mid-market company whose 15-person BDR team was producing only 15 percent of pipeline while marketing drove 85 percent. Because that team was a small share of pipeline, they could take the risk of blowing it up. They let the team go, kept the one person leaning into AI, and rebuilt the entire process around a single question: what should AI do, and what should the human do? By month three they were doing that same 15 percent of pipeline with two people instead of fifteen, and by month six they had scaled further. The lesson is not "fire your team." It is that the reinvention required a painful two months of rebuilding from the customer experience they wanted, backward. He is careful to add that if BDRs had been driving 90 percent of pipeline, he would have eased in by chipping away rather than blowing it up.

Work backwards from one metric

Jonathan's antidote to "what AI tool should I buy" is that it is a good question asked at the wrong time. You cannot know where to put AI until you know your revenue, your growth, and your actual problems. So he forces a single priority. Of revenue, upsells, profitability, and churn, pick one, even as an exercise. Then work backward.

Say it is revenue. You identify the metrics that drive it, like sales velocity, which is four things at once: average contract value, sales cycle, number of opportunities, and win ratio. Those have to move for revenue to move. So you ask what behaviors, tasks, processes, skills, and systems drive each of them, find where the most friction and lost conversion is, and only then point AI at that specific spot. He cites a customer, Kyle Norton, whose team used AI to statistically analyze hundreds of sales calls and discovered the biggest driver of win rate was not discovery, as everyone assumed, but the closing motions like setting up clear next steps. They operationalized that insight with humans and AI together and moved their win rate from 38 percent to 52 percent. As Jonathan puts it, this is the non-LinkedIn-worthy work of knowing your process and your data, and nobody wants to talk about it, but it is the whole game.

AI unveils bad processes

One of the sharpest ideas in the conversation is that AI is exposing weakness in the best possible way. Sometimes you do not know a process is bad until you try to speed it up and it breaks, because it was never really examined. AI is surfacing the weak spots people did not know they had. A lot of bad processes, he notes, are the ones made in a silo by someone who left long ago, that calcify into "this is just how we do it" without anyone remembering why.

His reframe is uncomfortable and useful: a good process in the old world that you are unwilling to let go of is a bad process, even if it works. The real danger is not the AI. It is whether you are willing to look honestly at what you do and ask if it is still the best way, or whether you are protecting it out of habit.

Where AI wins, and where it doesn't

Jonathan is precise about the boundary. AI is already outperforming humans at convergent thinking and analysis, pulling patterns from many data points. Having it analyze hundreds of calls and tell you why you are winning and losing in minutes is, in his word, insane. He would rather have AI surface the right slide from hundreds of decks than make a rep scroll through 150 of them. There are even AI systems that can carry a buyer through the early sales stages using a real methodology.

Where AI is overrated is divergent thinking, judgment, and messiness. Humans are good at holding two contradictory ideas, going down different paths, and being comfortable with optionality. Generative AI is naturally convergent, and it does not do well with the messiness of real life, balancing everything going on at home and at work while setting up something new. It needs direction and depth. The consequence is that the human superpower becomes communication, connection, genius, and passion, and AI becomes the force multiplier, not the replacement.

Your genius is the moat

If everyone has access to the same AI, Jonathan argues, then the AI itself is not where the value is. Give two people the same task and the same tools, and the output differs because the people differ. The future, he says, will be governed by those who create, and even self-generated AI will be a shadow of its creator, the way Ultron was still a shadow of Tony Stark. He points to the US Copyright Office ruling that a human has to author or meaningfully influence a work for it to be protected, as evidence that human IP is what carries the value. His line lands: AI cannot replace me, because I am me.

The system-versus-star point is the practical version. He would take a mediocre rep in a great system over a brilliant rep in a bad one, and reaches for Michael Jordan. Jordan's first coach built a system that depended on him and never reached the finals. Phil Jackson built a system that did not depend on Jordan but amplified him, and they won six championships. Consistent salespeople, he adds, usually have an internal system they follow. So his bet on the skill that matters most is a combination of metacognition, thinking about how you think and being willing to tear it apart, and critical, creative, systems thinking, layered on top of the discipline to build things rather than just talk about them.

Build skills, not hype

Jonathan's practical advice is to go learn how to build AI skills, because he believes the next era runs on the kind of skill architecture Claude has set up, the more flexible successor to custom GPTs. Go to YouTube, use the free resources, pay twenty dollars for a month, and actually build your own skills to see what is possible. It costs less than a pizza and a six-pack.

His frustration is with the theoretical noise. Plenty of people claim they generated ten million in pipeline with some tactic and have no receipts, no walkthrough, nothing to back it up, and it sits in the same noise bracket as the people doing genuinely great work, so nobody can tell what is real. He would rather people be transparent about where they actually are, including saying "I don't know," than sell the hype. And he is honest that everyone, himself included, feels behind, because the moment you catch up there is a whole new infrastructure to learn. That honesty, more than any tactic, is the throughline.

Key takeaways

A few things worth keeping.

Adoption is not transformation. A copilot license handed to someone who cannot prompt is not progress. Real operationalization means rebuilding the process around what AI and humans each do best.

Reinvent, do not just optimize. The big gains come from doing things totally differently, starting from the customer experience you want and working backward, not from making the old way marginally faster.

Work backwards from one metric. Pick a single priority, trace the metrics and behaviors that drive it, find the friction, and only then point AI at that specific spot. The tool question comes last.

AI exposes bad process. Speeding something up reveals whether it was ever good. A process you are unwilling to let go of, even one that works, is a liability.

Your genius is the moat. When everyone has the same AI, the differentiator is the human using it. Lean into passion and judgment, and use AI to amplify them.

Frameworks worth stealing

Work backwards from one metric

Force a single priority, whether revenue, upsells, profitability, or churn. Identify the compound metric that drives it, such as sales velocity, break it into its components, and map the behaviors, skills, and systems behind each. Find where the friction and lost conversion live, and aim AI precisely there instead of buying a tool and hoping.

Rebuild, do not optimize

For a function ripe for change, do not tune the old process. Start from the brand experience you want the customer to have, decide what only a human should touch, and automate everything else. Expect a painful rebuild period, and ease in rather than blowing it up when the function drives most of your pipeline.

The AI-versus-human split

Assign the work by strength. Give AI the convergent jobs: analyzing large volumes of calls and data, surfacing patterns, and retrieving the right asset instantly. Keep humans on divergent judgment, ambiguity, connection, and the brand-defining moments. Use AI as the force multiplier, not the replacement.

Build skills, cheaply

Do not wait for a course or a vendor. Spend twenty dollars, use free YouTube resources, and build your own AI skills hands-on to learn what is actually possible. The skill architecture you learn now is likely the foundation of the organizational agents to come.

Quotes worth keeping

The lines I wrote down.

Throwing a copilot license at someone is not transformation.

It's not about doing the old way better, faster, cheaper. It's about doing things totally differently.

AI is unveiling a lot of weaknesses in the best way possible.

A good process in the old world that you're not willing to let go of is a bad process, even if it works.

And the one that explains where value goes in an AI world.

AI can't replace me, because I'm me.

Rapid fire round

Same questions every guest. No prep, no warning. Here is how Jonathan handled it.

Advice you ignored or wish you had listened to? AI can do things fast, but humans take time, and that is okay. Be patient in the process. Slow is smooth, and smooth is fast.

What would you tell your younger self? Be patient, it is coming. He is a late bloomer who only found his calling in the last few years with AI, and all the failures and hard times are what made the current work possible. You will be grateful for what you are going through, even when it feels stupid in the moment.

Ongoing challenge that keeps you up at night? Less worry about himself, more about the people who are not willing to do the work to adjust fast enough, and what happens to them as the ground keeps shifting.

Favorite spot? Red Iguana, the Mexican place in Utah, and its sibling Blue Iguana. A Cali-style burrito on another level, plus a good barbecue spot he rates.

Tool you can't live without? His computer, and lately Claude and its cowork setup, which he uses to build his agent teams and skill infrastructure.

---

Jonathan Kvarfordt is VP of go-to-market at Momentum and the founder of GTM AI Academy, where more than ten thousand people have learned to operationalize AI across sales, marketing, rev ops, and customer success. He hosts the GTM AI Podcast. Find him on LinkedIn, at gtmaiacademy.com, or on the GTM AI Podcast.