When it comes to AI customer service , a good integration is worth its weight in tickets. Because here’s the thing: even the most advanced AI agent can’t live up to its potential if it’s cut off from the systems your business runs on.

We tackled this challenge head on in the latest installment of the ACX Framework webinar series. I sat down with Jed Thorn, Customer Support Manager at AllTrails, and Dennis Evans, Director of Client Services at NinjaTrader, to talk integrations—how they did it, what they learned, and what’s next.

The result? A goldmine of insights for any support leader trying to evolve from “this is fine” to fully integrated, AI customer service excellence.

the invisible difference between helpful and hero

An FAQ bot is nice. An AI agent that can do something? That’s a game changer.

Today’s customers expect self-service to be more than a search box. They want immediate, personalized, and effective support—without waiting in a queue. That’s only possible if your AI agent is connected to the systems that house customer data and power business logic.

At AllTrails, their AI agent is hooked into Zendesk, subscription management platforms, and payment processing systems—enabling it to understand who a customer is and act on their behalf. For NinjaTrader, Salesforce acts as the nerve center, powering ticketing, handoffs, knowledge, and a data lake integration for compliance reviews.

In my experience, this is what makes the magic happen. When AI is fully integrated into your systems, it’s a game-changer. That’s when your agent stops repeating static answers and starts resolving real issues—like updating payment info, issuing refunds, or escalating time-sensitive requests. It’s also when your AI starts paying dividends: higher containment, faster resolution time, and CSAT that makes your execs take notice.

But getting there isn’t just a technical challenge. It's an operational shift. And for teams like AllTrails and NinjaTrader, the path from basic automation to full-on AI impact wasn’t always straightforward. So how’d they do it? Let’s break it down.

the real blockers: not technical, human

You might assume integration challenges are technical. But most often? It’s people.

Jed and Dennis both highlighted that the biggest hurdles weren’t API configurations—it was resource allocation. Engineering teams are busy. Product roadmaps are full. And AI is still new enough that customer service doesn’t always get prioritized.

That’s why internal buy-in and a strong business case are everything.

Dennis shared that, for NinjaTrader, showcasing early ROI made all the difference: “It was easier to say, ‘you know, this will scale faster than bringing on more CSRs.’”

Both leaders emphasized the importance of early wins. Start small. Prove value. Use the momentum to unlock more.

This pattern mirrors what we’ve seen across dozens of Ada customers: one successful integration can kick off a domino effect of investment, support, and momentum. Especially if you use metrics like automated resolution and CSAT to show measurable improvement.

compliance isn’t a blocker—it’s a blueprint

For regulated industries like banking or finance, the stakes are higher, and the risks are real. But that doesn’t mean AI is off the table. It just means you need a smarter approach.

Instead of viewing this as a limitation, Dennis used it as a strategic starting point. He walked through how NinjaTrader approached integrations with compliance in mind: start small, start safe. “Let’s pass some low-risk information first. Let’s measure the results of that success. And then let’s see how we can push the envelope.”

By proving value in low-stakes contexts, the team built trust with their security stakeholders, eventually earning the green light to expand. This methodical rollout wasn’t just about reducing risk—it created a repeatable playbook for future integrations.

And let’s be honest: AI adoption is no longer optional—it’s rapidly becoming a standard across industries, with 78% of organizations reported using AI in at least one business function. Those who figure out compliance early will be the ones reaping the rewards later.

By starting with non-PII data and involving their security team early, NinjaTrader proved the safety and effectiveness of integrations—creating trust and appetite for more.

Let unresolved tickets write your integration roadmap

AI isn’t magic. If your AI agent is consistently hitting dead ends, it’s not working for your customers—or your bottom line.

Jed shared a brilliant tactic: use unresolved conversations to guide your next integration. “Looking at unresolved conversations, looking at AR, and looking at what’s not being resolved... that to me is almost like a roadmap for integrations.” Instead of overthinking where to begin, just follow the data . Look at where your agent is failing, and build from there.

For AllTrails, that meant zeroing in on subscription-related queries—their highest volume driver. By integrating their bot with subscription and billing platforms, they were able to automate the types of inquiries that were historically low-effort but high-volume for their agents.

NinjaTrader took a similar approach. They started by identifying one-touch tickets that were being escalated simply because the AI didn’t have the data it needed. With each small win, the team built a blueprint for scaling automation that actually worked. Dennis reported 98% containment on one specific action due to a small integration.

No big replatforming. No six-month dev sprints. Just one smart integration at a time—driven by real support data.

AI agents need teams, too

If your AI agent is shouldering a growing share of your customer conversations, it deserves more than ad hoc attention. It deserves a team.

But here’s the reality: most companies don’t start with a dedicated AI manager. AI responsibilities often get crammed into training, QA, or content roles—and while that can work for a while, it’s not scalable. The workload and complexity are growing fast.

Dennis is already working to carve out a dedicated role at NinjaTrader. “It’s honestly a full-time job,” he said, especially in a 24/7 environment like financial trading.

And it’s not just a tech role. According to Jed, “If I couldn't find someone who had deep AI bot management experience, I’d look for QA specialists, trainers, instructional designers… people who understand coaching and training.”

Here’s the lesson: you don't need people to come from technical backgrounds to fill these roles. In fact, skills in coaching, instructional design, or content optimization can be even more valuable. The future of customer service will be built by people who know how to coach both humans and machines.

Both leaders shared that their AI programs are starting to influence human agents, too. At NinjaTrader, Dennis said, "We're starting to learn some things from Ada that we might turn around and use with our human CSRs.” That’s when you know your AI agent isn’t just answering questions—it’s raising the bar.

from support tool to business engine

The most exciting part? Integrations are just the beginning.

Both companies are now exploring how their AI agents can drive more than customer service outcomes—whether that’s proactive messaging, in-product education, or even conversion.

Dennis described a vision where Ada could cancel fund transactions before critical cutoffs, proactively deliver product updates, or route users toward relevant services in real-time. “It’s a great customer service tool, but there are so many other use cases… it also becomes a product marketing tool.”

AllTrails is focused on continuing to peel off more subscription-related interactions—one API at a time. For NinjaTrader, the goal is to move from read-only actions to full-on transactional ones. “We can scale our business without scaling our headcount. That’s where the true cost savings is," Dennis added.

The bottom line: when AI is deeply embedded in your systems, it can do far more than answer questions. It can become a central player in how you engage, serve, and grow your customer base.

That’s where AI in customer service is headed: not just reactive support, but proactive engagement.

integrations are the key to unlocking your AI strategy

Even without integrations, your AI agent can deliver value—but the real magic happens when it connects to the systems that power your business.

AI doesn’t live in a vacuum. It needs to be connected to the tools your team already uses. When it's properly integrated, it can personalize responses, pull real-time data, and even automate complex tasks.

The best part? You don’t have to do it all at once. Start with low-risk use cases. Prove the impact. Build momentum. And watch your agent evolve from helpful support to indispensable teammate.

Here’s the playbook:

1. Start small and low-risk
2. Use real metrics (like AR% CSAT and containment) to prove ROI
3. Build internal momentum through early wins
4. Let unresolved conversations define your roadmap
5. Invest in the right people to manage and scale your AI agent

This isn’t about connecting a bot to a knowledge base. This is about transforming AI from a helpful add-on into an integrated, high-performing member of your team.

Because your AI agent can do more than answer tickets—with the right integrations, it can become your MVP.