How We Use AI to Know Our Customers Better

Here's an uncomfortable truth about customer success: the more customers you have, the less you know about each one.

At Pelles, we build AI tools for construction professionals. But what we don't talk about often is how we use AI internally — not to replace customer relationships, but to make them deeper.

The Moment We Knew Something Had to Change

Picture this: A CS manager is preparing for a quarterly business review. The customer is a 50-person electrical contractor in Phoenix. They've been with us for eight months.

The prep takes 45 minutes. Digging through Mixpanel. Cross-referencing Salesforce notes. Scanning Slack for mentions of their company name. Trying to remember what the last conversation was about.

And still, the call opens with: "So... how's everything going with the platform?"

That question is a white flag. It signals we don't actually know. The customer knows it too.

We were asking variations of the same questions before every call:

• Who's actually using this?
• What are they using it for?
• Is anyone struggling?
• What should we be talking about?

The answers existed — scattered across a dozen systems. But assembling them took longer than the conversations themselves.

What We Built (And What We Didn't)

Let's be clear about what this isn't: We didn't build a chatbot. We didn't automate conversations. We didn't replace anyone.

We built three things: a research assistant that does the homework, a query tool that answers questions in plain English, and an action-item tracker that turns meeting notes into Linear issues. Everything so our team can do their actual job — building relationships.

The Weekly Brief

Every Monday morning, each CS manager gets an account brief for their portfolio. Two minutes to read. Here's what it includes:

Activity snapshot: Who logged in, how often, what they did. Not vanity metrics — actual workflow completions.

Change detection: What's different from last week? New users onboarded? Power user went quiet? Sudden spike in a specific feature?

Conversation starters: Based on what we see, here are three things worth asking about.

That's not something you'd catch in a dashboard glance. But it's exactly the kind of signal that turns a generic check-in into a valuable conversation.

From Meeting Notes to Action Items in Seconds

Every customer call generates action items. Most of them end up in a notebook, a Slack message, or the void between intention and execution.

Our AI gets to the meeting summary, extracts the commitments, and creates Linear issues before the call ends.

Customer mentioned they need help with API integration? That's a ticket — assigned, prioritized, and tracked.

Someone asked about a feature that doesn't exist? That's product feedback — routed to the right team with full context.

The follow-up email promised by Friday? That's a task — with a due date and the original request attached.

No more "I'll get back to you on that" turning into "I forgot to get back to you on that." Every commitment is captured. Every follow-up is tracked. Every customer feels heard — because they actually are.

Plain-English Questions, Instant Answers

"Which enterprise customers haven't logged in this month?"

In most companies, that question goes to an analyst. Maybe you get an answer in a day. Maybe a week.

Our CS team just asks. The system translates the question into a query, runs it, and returns the answer in seconds.

This changes behavior. When data is easy to access, people get curious. They explore. They spot patterns they wouldn't have looked for.

The Principles Behind It

Three ideas guide everything we built:

AI prepares. Humans connect. The system surfaces insights, answers questions, and captures commitments. But every single customer interaction goes through a human who knows the account, knows the relationship, knows the context that no algorithm can capture.

Nothing falls through the cracks. When action items are captured automatically and tracked in Linear, promises become deliverables. Customers notice when you follow through — and they notice when you don't.

Trade research time for relationship time. Every hour saved on data-gathering is an hour reinvested in conversations, faster responses, proactive support. The ROI isn't efficiency — it's depth.

What Changed

Our CS team manages more accounts now. That's the efficiency story, and it's real.

But here's what actually matters:

Customers tell us they feel understood. They say our check-ins are relevant. They mention specific conversations that helped them — not because we solved a support ticket, but because we noticed something they hadn't noticed themselves.

That's what AI should do. Not replace the conversation — make it worth having.

---

At Pelles, we build AI tools that help construction teams work smarter. If you're curious how we can help your team, let's talk.