For twenty years, the shared services and BPO model was built to scale people. Move transactional work to lower-cost locations, standardize the process, and add headcount as volume grows. Now AI can handle much of that same work: keying in data, matching it, routing it, answering supplier questions. So what happens to the model?
That paradox was the focus of the latest episode of Extropy On AIr. Ravit Gutman, Founder and CEO of Extropy Advisors, welcomed back Bo Senu-Hodo, who leads global invoice-to-pay transformation at a multinational pharmaceutical company. Bo’s AP operations have been outsourced to a BPO partner in Hyderabad for six years, and she has just returned from speaking at SSON in San Diego. She lives this question every day.
A mindset shift on the conference floor
AI was the buzzword everywhere at SSON, but Bo noticed something more important in the attendees themselves. They’re looking past the next tool. In her session, the first lab series in SSON’s history, the group started with the process rather than the technology: who the stakeholders are, what the objective is, and what key results define success.
“Everyone is like, yeah, AI. But what is AI gonna do for me? What’s gonna change my world beyond just FTE reduction, beyond cost?”
Your BPO is an extension of you
Bo’s view of the outsourcing relationship is direct. A BPO doesn’t understand your processes the way you do, so it can’t be the one driving your transformation. Too many programs have been a lift and shift built to hit headcount and cost commitments, with no streamlining at all.
“If I did a lift and shift and you’re still doing the status quo, all because labor is cheaper in your area, there’s been no streamlining, no synergy, no change. It’s pointless.”
Instead, she meets with her BPO’s transformation team to align on strategy, and she pushes them to explain what technology they’re using and why. What works for one client may not fit another at a different maturity level.
Standardize before you automate
The theme Bo returned to most often was foundation. AI learns from history, so if the process is broken, the machine learns the broken process. You still end up with exceptions, variations, and manual intervention, and the cycle time looks the same as before.
She also pushed back on the idea that every country needs its own process. Tax and statutory requirements like e-invoicing mandates, VAT, and GST do vary, but in her estimate they’re roughly twenty percent of the picture. The core of invoice processing (goods receipt, matching, bill-to and ship-to, payment terms) should look the same everywhere. As Ravit put it, without that foundation, AI just makes a bigger mess faster.
Agents execute, humans govern
Bo doesn’t buy the fear that AI is coming for everyone’s job. She sees it redefining the human role instead, with people moving into governance, accountability, and exception judgment. Her team’s AI-enabled OCR rollout saved more than 15,000 hours in three months, and the goal was never just headcount. It was hours returned to the business, more compliant vendor behavior, and more time for the team to focus on the exceptions that need a human eye.
“You still need to own that driver’s seat, irrespective of whether you’re using agents.”
Measuring outcomes, not green boxes
Ravit pointed to SSON and EY research showing only eleven percent of shared services organizations are truly recognizing value creation, while most still measure cost and service levels. Bo’s approach looks different. She runs the data herself and compares it against what her BPO reports. She tracks manual steps removed, touchless PO invoices, and whether her ServiceNow inquiry volume is climbing, which signals a process problem even when every invoice was processed within SLA.
She also built a playbook her BPO team uses so recurring invoices don’t trigger a question to the business every time. And on her ServiceNow virtual agent, Tetsu, she gets notified whenever the same question keeps coming in without a good answer, so she can write the knowledge article that closes the gap.
“You need to manage the BPO, and not the BPO managing you.”
What an AI-ready partnership looks like
When Bo evaluates AI vendors, she goes beyond cost. She wants to know what data and test cases the tool was built on, what accuracy looked like in UAT, and whether it can handle exceptions autonomously within human-defined guardrails. Her test is simple: is this true autonomy, or traditional rule-based automation with a new label? She also uses free, scenario-based proofs of concept with her own data, and she recommends that vendors bring experienced practitioners onto their advisory boards so they stop building the same tool under different names.
On bringing work back in-house, her advice is to deploy and prove the AI inside the BPO first, then decide. Reverse that order and “it’s going to bite us.”
The takeaway
Bo closed with a book recommendation, Kaizen: The Key to Japan’s Competitive Success by Masaaki Imai, and a reminder that fits the whole conversation. Improvement should be continuous and incremental. Shared services wasn’t built in a day, and decades of work can’t be rebuilt in one day just because AI has arrived.
Her one piece of advice for finance leaders: own the relationship, own your processes, and move from tactical to strategic. The paradox is real, but as Ravit put it, it isn’t the end of the model. It’s the beginning of a better one, built on outcomes, clean data, strong partnerships, and people at the center.
Listen to the full episode of Extropy On AIr on Spotify, or wherever you get your podcasts, and share it with someone on your team.
Book a Discovery Call to talk through how your AP and shared services operations are set up for AI, and where your BPO partnership could deliver more.
