Why your AI investments aren't driving rep productivity (and how to fix it)

I'm going to share something that might make you uncomfortable.

When I asked a room full of revenue operations leaders how many had implemented AI for their go-to-market teams, about 60% raised their hands. But when I asked how many had seen real productivity gains? That number dropped to maybe 10% of those who'd implemented AI.

That's a 90% failure rate.

As someone who's spent years in AI research, working on the first GPT models at Stanford before anyone knew what they were and now runs a 60-person AI startup focused on revenue teams, I see this gap everywhere. And I think I know why it's happening.

The productivity paradox in revenue operations

Here's what's fascinating: while your engineering teams are seeing massive productivity gains from tools like Cursor and Claude Code, your revenue teams are stuck in the same inefficient workflows they've always had. The engineers on our tenth-floor ship feature faster than ever. Meanwhile, our go-to-market teams on the ninth floor are drowning in the same sea of tools and manual processes.

Why? Because we're approaching AI for revenue teams entirely wrong.

Think about how your reps actually work. They're jumping between Salesforce, conversation intelligence platforms, sales engagement tools, data enrichment vendors, Slack channels, and countless other systems. Each of these probably has some AI feature tacked on. But none of them talk to each other. None of them have the full context. And most importantly, none of them are actually working while your reps sleep.

That last point matters more than you might think.

The reason engineering AI tools work so well is they're constantly running in the background, doing work on your behalf. They're not waiting for you to open a chat window and ask them something. They're proactively solving problems and surfacing insights.

The illusion of progress

Most revenue organizations today are stuck in what I call the "illusion of progress."

You've implemented AI tools. You can check that box. But your reps are still spending hours every morning on account research. They're still manually updating CRM fields. They're still missing opportunities because they don't have the full context on their accounts.

I see this pattern everywhere. Companies implement one AI tool for call recording analysis, another for email writing and,maybe a chatbot for quick data lookups. Each tool promises to transform your revenue operations. But in practice? Your reps are spending just as much time stitching together information from different systems as they were before.

The problem is fragmentation. When your AI only has access to one slice of your revenue data, it can only give you one slice of insight. It's like trying to understand a customer by only looking at their email history.

You're missing the calls, the product usage data, the support tickets, the billing information—all the context that actually matters.

Three principles for AI that actually drives productivity

After working with nearly 100 enterprise revenue organizations, we've identified three principles that separate the 10% who see real gains from the 90% who don't.

1. Context is everything

Let me walk you through what it actually takes to get context on a single account. Say I'm trying to work a deal with Uber. Where do I need to look?

First, I check Salesforce for the basic account information and history. Then I dive into our conversation intelligence platform to review past calls. I scan through our sales engagement tool for email threads. I check our data enrichment vendor for recent leadership changes or company news. I search Slack for internal discussions about the account. Maybe I have notes from in-person meetings stored somewhere else.

That's six different systems just to understand one account. And that's assuming all the data is up to date and accurate (spoiler: it never is).

For AI to actually help your reps, it needs access to all of this context. Not just one system. Not just two. All of it. Because if your AI is missing even one piece of critical information, it's going to give bad recommendations. And once reps get a few bad recommendations, they stop trusting the system entirely.

2. AI must be goal-directed

Here's the beautiful thing about revenue operations: it's a conveyor belt. Every function has a clear goal tied to account progression. SDRs and marketing convert accounts to pipeline. AEs and SEs close deals. CSMs and AMs expand them. RevOps keeps everything running smoothly.

This clarity of purpose is actually a huge advantage for AI implementation. Unlike trying to build AI for, say, product management, where success metrics can be fuzzy, revenue teams have crystal-clear objectives. Every action either moves an account forward or it doesn't.

But most AI tools ignore this. They're built as general-purpose assistants rather than goal-directed agents. They can answer questions, but they can't tell you what you should be doing next to move your deals forward.

The AI that actually drives productivity doesn't wait for you to ask it questions. It knows your goals, understands your sales methodology, and proactively surfaces the next best actions to achieve those goals.

3. Learning loops are non-negotiable

AI makes mistakes. So do humans. The difference is that humans learn from their mistakes—at least, the good ones do.

Most AI implementations today are static. They give the same recommendations based on the same logic, regardless of whether those recommendations actually work in your specific context. There's no feedback loop. No learning. No improvement.

If you have 100 reps taking thousands of actions every day, you're generating massive amounts of data about what works and what doesn't. But if your AI isn't learning from that data, you're leaving enormous value on the table.

The most successful implementations we see have built-in learning loops. When a rep ignores an AI recommendation, the system learns. When a particular approach works well for closing deals in your industry, the system adapts. Over time, the AI gets smarter and more aligned with your specific business.

The evolution of AI in revenue operations

Looking across our customer base, we see organizations at four distinct levels of AI maturity:

Level 1: Bolted-on AI features - You've added AI capabilities to existing tools. Maybe your CRM has some predictive scoring, or your email tool has AI-powered templates. Limited impact because each tool only sees part of the picture.

Level 2: Task-specific agents - You're using tools like Claude or ChatGPT to handle specific tasks. Meeting prep, email writing, data analysis. Helpful, but requires manual effort and lacks context across tasks.

Level 3: Prescriptive account intelligence - This is where the magic happens. AI has full context, works continuously in the background, and proactively prescribes next best actions. This is where you start seeing real productivity gains.

Level 4: Self-improving revenue systems - The holy grail. Not only is AI prescribing actions, but it's learning from the outcomes and continuously improving its recommendations. Every deal makes the system smarter.

Most organizations are stuck at levels 1 and 2. The jump to level 3 is hard—it requires rethinking your entire approach to AI. But it's also where you finally start seeing the 30%+ productivity gains that make all the investment worthwhile.

Building AI infrastructure, not just applications

Here's the mindset shift that matters: stop thinking about AI applications and start thinking about AI infrastructure.

When you bolt AI onto individual applications, you get individual point solutions. When you build AI infrastructure that connects across your entire revenue stack, you get compound value that grows over time.

Think of it this way: imagine if you could assign one dedicated seller to every single account in your TAM. They'd maintain perfect context on that account, constantly monitor for opportunities, and always know the next best action to take. Obviously, that's economically impossible with humans. But it's precisely what AI infrastructure can provide.

This is the concept we've built our entire company around—per-account agents that live with each account throughout its entire lifecycle. These agents maintain memory of every interaction, have access to all available context, and continuously learn from what works and what doesn't.

The key is that these aren't task-specific agents that wake up, do one thing, then go back to sleep. They're long-running agents that maintain state and context over time. Just like a great AE who's been working an account for months knows exactly where things stand and what needs to happen next.

What this looks like in practice

Let me paint a picture of how this actually works when implemented correctly.

Your SDR wakes up in the morning. Instead of spending three hours researching which accounts to prioritize, they have a prioritized list waiting for them. Not just a list—but detailed reasoning for why each account is prioritized, what messaging will resonate, and what actions to take.

Your AE preparing for a meeting doesn't need to dig through six different systems to understand the account context. Everything is synthesized and available, with specific recommendations for how to advance the deal based on your sales methodology and what's worked in similar situations.

Your RevOps team running forecast calls has an objective AI perspective on every deal. Not replacing rep judgment, but augmenting it.

The AI flags risks the rep might have missed, identifies deals that could be pulled forward, and provides reasoning based on the full context of the account.

And critically—none of this requires your team to learn new interfaces or change their workflows. The AI shows up where your reps already work. In Salesforce. In Slack. In their email. Even in tools like Claude's new Cowork interface.

The path forward

The gap between the 10% of organizations seeing real AI-driven productivity gains and the 90% who aren't comes down to approach.

The successful ones aren't just adding AI features to their stack. They're building AI infrastructure that works across their entire revenue operation.

This requires three things:

First, you need to centralize context. All the data, from all your systems, is accessible to your AI. This is non-negotiable. Partial context leads to partial insights, which leads to reps who don't trust the system.
Second, you need goal-directed agents that understand your business, your methodology, and your objectives. Generic AI assistants won't cut it. You need AI that knows what success looks like in your specific context.
Third, you need learning loops that make the system smarter over time. Every action your team takes, every deal that closes or doesn't, should feed back into the system to improve future recommendations.

If you're in the 90% not seeing real productivity gains from AI yet, you have two choices. You can keep adding point solutions and hoping something sticks. Or you can take a step back and think about AI infrastructure for your entire revenue operation.

Start by auditing where your context lives today. How many systems does a rep need to touch to get full context on an account? That's your baseline.
Then look at your current AI investments. Are they connected? Do they share context? Do they learn from outcomes? If not, you're building on sand.
Finally, think about what success looks like. Not just "we use AI" but specific, measurable productivity gains. What would it mean for your business if every rep was 30% more productive? That's the bar you should be aiming for.

The engineering teams have already figured this out. They're shipping faster than ever thanks to AI that has full context, works continuously in the background, and gets smarter over time. It's time for revenue teams to catch up.

The question isn't whether AI will transform revenue operations. It's whether your organization will be in the 10% that figures out how to make it work,or the 90% still wondering why their AI investments aren't paying off.