Everyone wants to talk about AI agents. Few are ready to do what it takes to make them actually work.

At Ada , we’ve written before about the difference between reactive bots and truly agentic AI : systems that don’t just respond, but resolve. Automation that doesn't just follow instructions, but can learn, reason, and act with autonomy. The kind of AI that doesn’t wait for permission to fix the problem, it just fixes it.

But turning AI into a high-performing team member doesn’t start with technology—it starts with the operating model. Agentic AI only thrives when supported by a customer service model built for it: one grounded in clear roles, performance metrics, and continuous improvement.

That’s why we’re pulling back the curtain on one of the most sophisticated AI-first customer service transformations we’ve seen to date: Blackhawk Network (BHN) .

Over the past year, BHN didn’t just deploy an AI agent—they restructured their operations to help it perform. Their AI agent, Timmy Tango, now takes the front lines, resolving issues at scale and improving continuously alongside the human team. By establishing the right roles, workflows, and systems, they’ve embedded AI into the core of their customer service.

The result? A leaner, smarter, more strategic team that’s setting the bar for what it really means to operationalize agentic AI.

So what does it actually take to bring agentic AI to life inside a complex customer service organization? BHN’s transformation offers a glimpse at what’s possible when AI is treated not as a tool, but as a teammate.

From ticket chasers to AI collaborators

Before BHN acquired Tango Card, the team followed a familiar model: agents fielded high volumes of emails and calls, supervisors handled people management and escalations, and knowledge lived mostly in people’s heads.

It worked for a time, but it wasn’t scalable. There was no bandwidth for strategic planning, no space for proactive improvements, and no infrastructure to support automation, let alone autonomy.

Early automation helped relieve pressure by handling routine inquiries. That breathing room opened the door to bigger changes. When generative capabilities were introduced, the shift accelerated. Agents were no longer expected to know everything. Instead, they were expected to become domain experts, and use that expertise to guide and coach the AI.

BHN began assigning agents to specific product areas and training them to review, critique, and improve AI conversations. To support this evolution, BHN also introduced new roles and specializations:

• Product Specialists who analyze contact drivers and advise on content improvements.
• Content and QA Specialists who maintain the knowledge base and audit AI + human conversations.
• And soon, AI Certified Agents, a role focused on testing AI outputs, validating self-service flows, and reinforcing continuous learning within the team.

What BHN proved is this: agentic AI doesn’t replace the team, it elevates them . It frees people from repetitive tickets and invites them into higher-order work like training, coaching, and strategy.

Agentic AI needs structure, not scripts

Autonomy doesn’t mean absence. Just like your top-performing employee, agentic AI only performs when it’s supported by the right systems: structured knowledge, clear feedback loops, and ownership across technical and operational teams.

To support that structure, BHN used a maturity model to guide their transformation, mapping out what it takes to operationalize AI. That framework helped define clear roles and workflows across operations, content, and systems.

• AI Managers focused on performance tuning, reviewing transcripts, and tracking resolution metrics.
• Knowledge Managers acted as content strategists, structuring answers that were optimized for both accuracy and autonomy.
• Technical leads built and maintained the integrations that let the AI agent take meaningful action—not just provide information.
• And a dedicated program owner aligned it all to broader goals, driving collaboration across support, systems, and product.

They even invested in formal upskilling, rolling out training in analytics, agile methodologies, and systems thinking to turn frontline staff into AI collaborators.

This wasn’t just a customer service team learning to use a new tool. It was an operational shift—one that gave agentic AI the foundation to deliver real business outcomes.

From structure to ownership: what AI needs to do its job

Structure alone isn’t enough. To unlock real impact, BHN gave their AI agent the ownership, access, and oversight it needed to act with confidence.

This looked like:

• System access: Integrations that let the AI take action, not just provide information. Things like resending rewards, looking up order data, and routing cases to the right owner.
• Shared context: A knowledge base structured for both humans and AI, so everyone operates from the same source of truth.
• Ongoing coaching: A consistent feedback loop—transcript reviews, performance audits, and optimization cycles—to help the AI improve like any high-performing team member.

This is what distinguishes agentic AI from traditional automation. It doesn’t just answer or follow instructions, it learns, adapts, and acts—effectively taking ownership of customer service interactions.

At BHN, that ownership is what makes the AI agent a true part of the team.

Lessons from BHN: what agentic AI looks like when it works

These aren’t theoretical takeaways. They’re backed by BHN’s real results after a year of operational maturity and scaling across brands, channels, and use cases. Here’s what their story shows us about doing agentic AI right:

1. Don’t staff around tickets. Staff around outcomes. Roles like AI Manager, Knowledge Strategist, and Technical Owner are critical. Without them, even the best AI hits a ceiling.
2. AI agents need coaching, just like humans. BHN built coaching workflows into their week. They audit transcripts, analyze resolution gaps, and tune the AI constantly (not just quarterly).
3. Integrations = intelligence. The more systems your AI can talk to, the more useful it becomes. BHN’s team prioritized reusable API actions to reduce overhead and scale efficiently.
4. Success starts with structure. BHN didn’t just launch an AI agent. They restructured the team to support a new model of customer service—one where AI takes the front line, and humans steer the ship.
5. Autonomy is earned. Agentic AI doesn’t happen by accident. It’s built through clear ownership, intentional design, and a willingness to rethink the way service gets delivered.

This wasn’t a textbook implementation. It was an intentional reinvention of what customer service can be when AI isn’t just layered on, but fully embedded into how the team operates, collaborates, and delivers impact.

BHN didn’t just adopt new technology—they built the structure, systems, and roles to make it perform. With AI now integrated across channels and business lines, they’ve created a model for what the modern support team looks like: leaner, smarter, and strategically aligned.

And BHN’s team is just getting started. As they scale Ada across more brands, channels, and use cases, they’re not just meeting expectations. They’re setting them.