Early in my career at Salesforce, I watched two sales reps work the same territory with roughly the same product knowledge and the same amount of hustle. One consistently beat quota. The other spun his wheels. I watched one rep get more confident and spiral up, and the other get less confident and spiral down. The difference wasn’t talent or work ethic. It was how each of them used information. One was making decisions based on gut feel and habit. The other was paying close attention to what the data was actually telling him.

That observation stuck with me. And after nearly two decades in enterprise sales, across Salesforce, MuleSoft, and Arteria AI, it’s shaped how I think about building sales teams and strategies that actually scale.

Here’s what I’ve learned.

Start with the right questions, not the right tools

The most common mistake I see sales leaders make when going “data-driven” is leading with technology. They buy a new CRM, a new analytics platform, a new forecasting tool, and expect the insight to follow. It rarely does.

Data strategy has to start with questions, not software. What does your ideal customer actually look like? Which deals in your pipeline are genuinely likely to close, and which ones are wishful thinking? Where in your sales cycle are you consistently losing momentum? What does your best rep do differently in the first two discovery calls?

When you know what you’re trying to answer, the data you need becomes obvious. When you start with the tool, you end up with dashboards nobody looks at.

This is actually one area where AI agents are starting to earn their keep. Rather than a rep or manager manually pulling reports to answer these questions, a well-configured AI agent can surface the right signals proactively, flagging a deal that has gone quiet, or noting that a prospect’s engagement pattern looks similar to three other deals that stalled at the same stage. The question still has to come from a human. But the agent can do the leg work of finding the answer before you even think to ask.

Build your ICP from evidence, not instinct

One of the most valuable exercises I ran at Arteria AI early on was going through our best opportunities and asking a simple question: what did these customers have in common that we didn’t initially predict? The answers were humbling. Our assumed ICP was legal and our actual ICP was operations, turns out we were beginning with the wrong person and accidentally getting sent to the right one on occasion.

This matters enormously for how you allocate time and resources. If your reps are prospecting based on a persona that was defined in a whiteboard session three years ago, they’re working with a map that might not match the territory anymore.

A data-driven ICP is a living document. It should be revisited every quarter, stress-tested against actual win/loss patterns, and refined based on what your best customers have in common rather than what you hoped they’d have in common. Firmographics are a starting point. Behaviour, timing, and trigger events are where the real signal lives.

AI agents are beginning to change how quickly this analysis can happen. What used to take a revenue operations person a week to pull together, cross-referencing CRM data with external signals like hiring trends, funding announcements, or leadership changes, can now be done continuously and automatically. The ICP stops being a quarterly exercise and starts being something your team is always sharpening.

Pipeline quality beats pipeline quantity, every time

At MuleSoft, one of the first things I did as RVP was pull apart the Canadian pipeline and look at it honestly. What I found was a classic problem: a lot of volume, not enough velocity. Deals were sitting in mid-stage for months, reps were optimistic about accounts that hadn’t shown real buying signals, and the forecast was more fiction than fact.

The fix wasn’t to generate more pipeline. It was to get disciplined about what belonged in the pipeline in the first place.

Data gives you the tools to make that call objectively. Stage conversion rates tell you where deals go to die. Time-in-stage analysis surfaces the deals that are stalled versus the ones that are just slow-moving by nature. Win rate by segment, by competitor, by deal size, by rep tells you where you’re genuinely competitive and where you’re kidding yourself.

Today, AI agents can run this analysis in the background continuously, alerting managers when a deal has been sitting too long, when key contacts have gone dark, or when a competitor has entered a deal based on language picked up in email threads or call transcripts. The judgment call about what to do with that information still belongs to the human. But the time between a deal going sideways and a manager noticing is collapsing fast.

Make data part of the coaching conversation

The best use of sales data isn’t in the boardroom. It’s in the one-on-one.

When I was managing teams at Salesforce, the reps who grew the fastest weren’t necessarily the ones with the most natural ability. They were the ones who were willing to look at their own numbers honestly and have a real conversation about what they meant. Call-to-meeting conversion rate too low? Let’s listen to some calls together. Average deal size trailing the team? Let’s look at where you’re discounting and why.

Data makes coaching specific. Without it, feedback is subjective and easy to dismiss. With it, you’re having a conversation about evidence, not opinion. That shift changes the dynamic in the room completely.

AI is accelerating this too. Conversation intelligence tools can now analyse every call a rep takes, identifying patterns in talk time, question frequency, objection handling, and competitor mentions, and surface those insights automatically before a coaching session. A manager no longer has to listen to hours of recordings to know where a rep needs work. The agent does the scouting. The manager does the coaching.

The key is making sure that culture stays intact. Data should develop people, not surveil them. If reps feel like the AI exists to catch them doing something wrong, they’ll disengage. If they see it as something that helps them get better and earn more, the dynamic flips entirely.

Don’t let perfect be the enemy of useful

A word of caution for leaders who are earlier in the data-driven journey: you don’t need a perfect data infrastructure to start making better decisions.

I’ve seen companies spend eighteen months and significant budget on a data warehouse project before anyone in sales saw a single useful insight. Meanwhile, a spreadsheet with three months of closed-won data and a few honest questions would have told them most of what they needed to know.

Start with the data you have. Get disciplined about CRM hygiene so the data you’re generating is trustworthy. Build one or two metrics that your whole team understands and cares about. Then layer in sophistication over time, including AI tools, as the culture and capability develop together.

This sequencing matters especially when it comes to AI agents. They are only as good as the data they’re trained on and working with. A team with poor CRM hygiene that bolts an AI layer on top doesn’t get better insights. It gets worse ones, faster, and with more confidence. Get the foundation right first.

The rep who reads the numbers wins

I think back to those two reps in the same territory at Salesforce. The one who consistently won wasn’t just working harder. He was paying attention to patterns, adjusting his approach based on what was actually working, and making decisions that the data supported.

That instinct, to look at the evidence and let it change your behaviour, is learnable. It’s coachable. And as AI agents take over more of the work of finding and surfacing that evidence, the skill that matters most is knowing what to do with it. Interpretation, judgment, and the willingness to act on what the data is telling you, even when it’s uncomfortable, will separate the best sales leaders from the rest.

The teams that figure this out first will have a structural advantage that compounds over time. The ones that don’t will keep wondering why their best efforts aren’t translating into results.