Scaling BPO success with agentic AI: Real-world client wins

How BPOs like Humach are winning clients over and over with AI automation

June 25, 2025
12
Min Read

While BPOs are getting serious about AI, the Fortune 500s they work with are stuck in the same pilot. We recently sat down with Bruce Sharpe, Chief Product Officer at Humach, to talk about agentic AI, evaluation criteria, and what’s actually working in the real world.

Bruce has been in customer experience for over 30 years. He's seen the BPO industry evolve from basic IVR systems to today's AI-powered digital workers. And he sees through the hype.

"There's a lot of companies putting stuff out there, but they're not putting processes in place to manage it," Bruce said during our recent webinar. "That's where we come into play.”

"Our name, Humach, comes from humans and machines," Bruce explained. "AI requires care and feeding. You need good customer experience design, testing, deployment, and training."

Read a summary of our chat below, and watch the recording at any time.

The AI agent for music talent managers

Unexpectedly, one of Humach’s biggest AI successes came from the music industry.

Bruce worked with Gate52, a talent management company that books over 2,500 shows per year. Their buyers were spending hours researching venues, checking where artists had played before, and analyzing revenue data across multiple systems.

"They realized that people were the limitation," Bruce explained. "They wanted to take the 2,500 plus shows they're doing a year and quadruple that with the same number of people."

Using Pinkfish, Humach built Caitlin: an AI agent designed for music talent managers that works through SMS. Talent managers can text questions like "What venues has Metallica played?" or "What were Taylor Swift's revenues over the past three years?" Pinkfish scrapes multiple databases, some with APIs and others without, then returns formatted answers in seconds.

"It's really complex," Bruce noted. "There are several systems that don't have APIs. We do screen scraping. There's authentication. One system has 600 attributes for a single show."

Gate52 can now research shows in minutes instead of hours and let buyers focus on actual bookings.

Start fast, see feedback

Despite these successes, some Fortune 500s are still cautious about AI.

"Some are afraid to get started,” Bruce observed. “They want proof, ROI data, and case studies. Others are afraid of failures because they've seen so much in the public eye.”

But their biggest barrier is good old fashioned analysis paralysis.

"They haven't deployed anything themselves, but they're putting roadblocks in place because of fear," Bruce said.

Bruce’s advice: start small. “Look for a use case that's simple, something you can get out fast and see positive feedback. Don't go with your most complex process."

10 AI agent platform evaluation criteria

When Humach needed to choose an agentic AI platform, Bruce didn't want to get stuck in analysis paralysis. He opted for a more practical approach.

"We didn't sit down and try to boil the ocean," he said. Instead, he used AI to build a comparison matrix of potential vendors, then narrowed it down based on 10 key criteria:

  1. Cloud-based environment
  2. AI at the core with multi-LLM
  3. Low or no-code ready
  4. Prompt-driven design
  5. Custom connector, API integrations, and scraping
  6. Multi-org-based client management
  7. Multi-channel input and output
  8. Strong security: HIPAA, SOC2, GDPR, etc.

"You want to make sure you're lean and mean, and move really fast," Bruce advised. "Otherwise you're going to take many, many months, even years, and by that time things are going to pass you by."

Building AI agents in hours

Humach partnered with UT Dallas students to build AI agents using Pinkfish. Out of 166 total projects and 600 students, their AI executive assistant was the most requested project by students.

"AI is the number one requested type of project at universities," Bruce noted. With Pinkfish, students built functioning AI agents by working only 10 hours per week.

"These students were blown away by how fast they could implement things because of the automation," Bruce said.

“AI has gotten much better at generating code and building software, which is what Pinkfish does,” he continued. “We talk about AI agents, but in the background, the way you build any automation with Pinkfish is describing it in English.”

“It writes software code in the backend and builds it for you. It's amazing to see the students here embracing that and winning awards.”

The opportunity for BPOs

Bruce is excited about agentic AI's potential but realistic about the challenges. Mostly, he’s excited about automating "low-hanging fruit"—repetitive tasks you can handle quickly without humans intervening.

His only concern is deploying AI without proper monitoring and feedback loops.

"We're putting together a process where we can evaluate AI responses to determine if they were right, biased, or inappropriate," Bruce explained. "They can give thumbs up or down, suggest better answers, and the system learns from that."

For BPOs specifically, Bruce sees huge opportunity. While AI companies are popping up everywhere, most don't understand enterprise customer experience like BPOs do.

"A number of those companies don't understand customer experience. They just want to throw AI out there to solve a problem they think they know about."

BPOs understand enterprise operations, have existing client relationships, and can bridge the gap between AI hype and real solutions.

“AI-enabled BPOs have the unique opportunity to be a trusted partner for customers and bring AI into production for them,” Bruce concluded.

Ready to build AI agents for your clients? See how Pinkfish helps BPOs deploy enterprise AI automations faster and safer than traditional solutions.

Get started with automation today