The Bottleneck Isn't the Algorithm

I recently joined Sana on the BizBlend podcast to talk about something the tech industry loves to ignore: most enterprise technology doesn’t fail because the tech was bad. It fails because nobody used it. Or nobody used it the way it was meant to be used. Or nobody used it long enough for it to matter.

The conversation stuck with me, so I want to expand on a few of the ideas we explored. After twelve years at Salesforce leading product adoption within customer success, and years of strategic AI work with major enterprises after that, I keep watching the same movie with different actors.

“If you build it, they will come” is not a strategy

Every hype cycle follows the same arc. Everyone gets excited about the technology, invests heavily, and rolls something out. Then, at the end of the project, comes the dreaded conversation: “Okay, now we need to talk about change management, because nobody’s using it.”

That conversation happens because the users were never brought into the fold. Nobody asked them what they needed, what they wanted, or what frustrated them about the tools they already had. And here’s the thing about asking: people will tell you. Ask anyone about their experience with the technology they use today and you’ll get a hundred frustrations. That’s not noise, that’s your roadmap. If your rollout plan doesn’t start there, you’re planning a soon-to-be failed launch, not an adoption.

The AI industry built for one user. There are millions of others.

Right now, AI’s biggest wins are concentrated on one persona: engineers. Developers using coding agents are getting extraordinary leverage. But look at everyone else. Sellers, customer success teams, ops, the business functions that actually run a company. A few tech-forward people are extracting real value. Most are just not there yet.

And it’s not because they’re behind. It’s because we keep asking them to become something they never signed up to be.

I spoke with a business leader recently who told me she has a tech guy who figures out all her AI tools for her, because she doesn’t want to learn AI. She just wants to use it. That’s not a confession of laziness. It’s one of the most common sentiments in business today, and it’s completely reasonable. A great seller is great at building relationships and talking to people. Asking them to first master the difference between predictive and generative AI, understand what an agent is, and stitch ten tools together, create a fun infographic for LinkedIn… isn’t enablement. It’s homework.

The swivel chair problem

Here’s what a broken experience actually looks like on a normal Tuesday. A salesperson gets leads from one tool, generates emails in another, logs everything in a CRM that’s a third, and exports spreadsheets to bridge the gaps. We’ve reached a strange inversion: the human is doing the work for the AI tools, instead of the tools doing the work for the human.

Every one of those swivel-chair moments between point solutions is where value leaks out. Data gets lost between funnel stages. Context evaporates. And the person who was hired to sell spends their day being middleware.

IT tries to solve this by consolidating point solutions and building more in-house, which brings its own trap. The “software is dead, just build it yourself” crowd is right that you can build a lot now. What they don’t plan for is everything after the rollout: keeping it functioning, secure, and compliant over years, with funding and focus that most organizations can’t sustain. New features are exciting. Maintenance is not. Initiatives fail on the unexciting part.

What good looks like

Flip the model. When a lead comes in, the research should already be done, generated from an email, a name, and a LinkedIn profile before a human ever touches it. Use that intelligence to route the lead to the rep with the best shot: the one who’s great at that use case, lives in the same city, roots for the same team. Have the first-touch email drafted and waiting, referencing the prospect’s recent promotion, ready for the rep to review and send.

The human’s job in that flow is the part humans are irreplaceably good at: the judgment call and the relationship. “Yes, that’s right. Send it.” Everything else, the exports and imports and tab-switching and nitty gritty, should simply happen.

And the interface matters more than we admit. A conversational agent works best when it’s paired with what people already know. If someone lives in a CRM, pair the agent with the CRM. If they live in dashboards, let the agent answer the questions the dashboard doesn’t, without filing a ticket for a new widget.

Adoption is a multi-year program, not a launch email

At Salesforce we planned adoption journeys that spanned years, and a few lessons held regardless of the technology.

Measure relentlessly. If you can’t see at a glance how people are using a tool and whether usage is growing, you can’t know where to intervene. Sometimes things take off and blow past every target. Sometimes they need work, and it’s almost never the technology that needs fixing. It’s the communication and the sponsorship around it.

Executive sponsorship changes everything. There is a massive difference between a rollout where the CEO says “this matters and we’re excited about it” and one championed from somewhere lower in the org. Where the champion sits determines how seriously people take it.

Communicate continuously, and personally. Adoption is not one launch email. It’s a sustained chain of communication, and the more personalized it is, the more people read it and feel it’s something they need to act on. Generic announcements get archived.

Intelligence isn’t enough

There’s a lot of fear in the air right now about AI taking jobs. My honest view: AI is a technology like every other wave before it. Computers, cloud, mobile. Each cycle drove things forward, and each one ultimately came down to the same question: how do people actually use it, and does it make their work more valuable?

That belief is why I founded youex.ai. The name is the point. Your experience. Intelligence isn’t enough; it’s the human experience that matters. We’re building for the seller who wants agents working for them, capturing leads, doing research, and drafting outreach, without needing to learn what an LLM is first.

If you’re a founder, a rev ops leader, or in customer success and any of this sounds like your Tuesday, I’d love to hear from you. You can listen to the full conversation with Sana here, find me at youex.ai, or just reply to this post and tell me: what’s the tool your team quietly stopped using, and why?