Agentic AI
Global Business Services
Back-Office Automation in Practice: Exploring the Reality Behind the GBS AI Hype
Setting the record straight. How far has the GBS industry come with Agentic AI and how can we bridge the gap from ambition to reality.
.png)
Earlier this year, SSON Research & Analytics’ Managing Director Naomi Secor and Global Editor, Barbara Hodge were invited as expert guests onto Hypatos' podcast GBS Rewired. The idea was to set the record straight. Beyond the Agentic AI headlines, how far has the GBS industry actually come with Agentic AI and how can we bridge the gap to Agentic GBS. See below the key takeaways from their discussion with Hypatos’ Head of Thought Leadership, Sally Fletcher.
It's worth naming the gap up front, because it shapes everything else in this piece: what Agentic AI can theoretically do and what most GBS functions are actually running are two different things. SSON Research & Analytics data puts only 16% of organizations live in production or scaling agentic AI, and no organization has reached full end-to-end autonomy i.e. level five of five on SSON R&A's maturity scale. Just 42% of the market describes itself as familiar with the concept at all. That's not a reason to slow down; three-quarters of the industry still expects Agentic AI to significantly change the service delivery model. But it does mean the practical lessons below matter more than the possibility-space headlines. The organizations making real progress aren't the ones with the boldest vision; they're the ones doing the unglamorous groundwork of process readiness, governance, and skills-building that turns possibility into production.
The single biggest hurdle isn't the technology; it's readiness. Before any agentic deployment, teams need the business case in order and a clear-eyed view of which processes are genuinely suited to it. SSON R&A's research shows the market expects transactional processing and analytics/reporting to be hit hardest first, which tracks with what's actually reaching production today. Almost none of the market is running autonomous, lights-out processes; most live deployments are still doing intelligent automation work within levels one to three of agent maturity.
A practical detail that's easy to miss: Agentic AI is only as good as the work instructions behind it, and those need to be written by the people who actually own the process, i.e., typically the global process owner, rather than IT. That's a meaningful shift in accountability, and it's part of why GBS functions are increasingly driving agentic initiatives directly rather than just supporting them, with governance, security certification, and standardized operating procedures built in from the start.
CFOs funding agentic AI projects typically expect ROI within 12 months — a tough target if a team is starting from scratch on prompting, agent development, security sign-off, and upskilling all at once. Pre-built, pre-prompted agents for common processes like procure-to-pay and order-to-cash, which can be lightly customized rather than built from the ground up, are one of the more reliable ways teams are compressing that timeline. It's also a hedge against "agent-washing,” providers claiming agentic capability without being able to support real end-to-end exceptions.
Some of the clearest lessons come from organizations that built strong shared services capabilities well before the current AI wave. Naomi pointed to long-time standouts like Procter & Gamble and Novartis, which integrated finance, HR, IT, and procurement under one pillar early and treated technology ownership as core to the model; a pattern now resurfacing as companies stand up Chief Agentic AI Officer roles inside GBS. Another recent example is from this year's World's Best GBS winner: Siemens for engineering services, showing how far the scope of "back office" has moved beyond its original footprint.
SSON R&A has built several resources specifically to help teams avoid reinventing the wheel. The Intelligent Automation Universe (IAU) now lists more than 80 vendors with apples-to-apples comparisons on licensing, geography, fees, and technical differentiators. This is incredibly useful for teams vetting automation partners without a lengthy independent search. The GBS certification program, run with Filippo Passerini and Tony Saldana for the past seven years, gives individuals and teams a credential built on the lessons of others who've been through the transformation already. And the GBS industry standards, developed with a founding committee of ten current GBS leaders and drawing on practices from companies like Tesco, Cardinal Health, and Allianz Services, offer a non-prescriptive guide to what good looks like.
Even with strong tooling, the skills gap remains the practical constraint. AI model training and governance came out as the top skill needed in a post-FTE delivery model, but two-thirds of organizations are only somewhat confident their workforce can make that shift, and over 70% think more than a quarter of their workforce needs reskilling. Some GBS leaders are addressing this directly by adding prompt engineering to core training curricula, treating the ability to direct an AI agent effectively as a foundational, teachable skill rather than a specialist one.
Back-office automation is working where teams treat it as an operational discipline, not a switch to flip: right-size the process, put ownership with the business, borrow proven agents and community benchmarks rather than building everything from scratch, and invest as much in the people and relationships as in the technology itself.
To listen to the full episode click here.
Further stories from our blog
