CCO Says AI is Changing CX Org Charts — and That’s Why She’s Transitioning to a “Service Pod” Model
Sofia Barbosa explains how AI is reshaping CX org charts through service pods that unite AI and human teams around end-to-end customer outcomes.
Sofia Barbosa
Chief Customer Officer at BMC Software
Future Design: AI is prompting CX leaders to replace siloed teams and handoffs with simpler, outcome-focused structures.
Service Pods: Service pods unite support, services, training, and customer success around shared end-to-end accountability.
Closed Loop: AI analyzes customer signals, drafts training, and delivers guidance when adoption gaps appear, with human review.
Human Judgment: People retain responsibility for trust, escalations, and strategic customer decisions while automation accelerates analysis and content production.
Production Scale: Successful pilots still fail without integrated workflows, quality controls, and scalable operating processes designed from the beginning.
Sofia Barbosa is the Chief Customer Officer at BMC Software, a B2B company serving customers with complex IT environments.
We caught up with Sofia to learn how she's deploying AI — and how it's changing her CX org chart. Here's what she said.
Leading the full post-sales customer lifecycle
I'm Sofia Barbosa, Chief Customer Officer at BMC Software. I lead the Customer Office, a unified organization bringing together Professional Services, Support, Customer Success Management, Education, and Voice of the Customer.
Continue Reading for Free
Create a free account to finish this article, plus get ongoing access to timely insights and practical resources.
Have an account? Log In
Together, these functions span the full post-sales customer lifecycle, from implementation and enablement through ongoing success, adoption, and advocacy.
We serve B2B customers operating some of the world's most complex IT environments, which means our teams work across a wide range of channels and engagement models, from high-touch strategic account relationships to scaled, self-service support.
Why AI is Changing CX Org Charts
Most CX org charts today reflect decades of accumulated complexity: specialized teams, narrow handoffs, and processes that assume humans do every step of the work. AI breaks that assumption, but many leaders are still trying to fit AI into the existing structure rather than asking what the structure should look like if you designed it fresh, with AI as a given.
My advice is to deliberately forget the "as is" and focus on the "to be." That means resisting the urge to ask, "How does AI fit into our current teams and processes?" Instead, ask, "If we built this org from scratch today, with AI capabilities available, what would it look like?"
In most cases, the answer is simpler than what exists today: fewer silos, fewer handoffs, and roles defined by outcomes rather than narrow functional tasks.
This is hard because it means letting go of structures, titles, and processes built over years, but the leaders who redesign around the "to be" rather than retrofitting the "as is" will create org structures both simpler and more capable than their original ones.
How AI Enables a "Service Pod" Model
Over the past year, we implemented a "service pod" model — an ongoing restructuring to break down silos between services, support, CSM, and training, and integrate AI agents into that same unified org chart instead of treating them as a separate, bolted-on tool.
The catalyst was AI agents taking on a growing share of routine interactions. We realized that if AI operated outside the existing team structure, it would create a new silo — another handoff gap instead of closing existing ones. So, we are working to place AI agents, support, CSM, services, and training under shared leadership with shared visibility into account history, open issues, and customer context.
As a result, in areas where the pod model is already taking shape, customers no longer have to re-explain their situation across channels. Context travels with them, regardless of whether AI or a human responds first.
AI agents also surface patterns, such as common questions and friction points, in real time — allowing us to update training and documentation proactively, rather than reactively.
Accountability is shifting from "which team owns this" to "which pod owns this customer's experience end-to-end," which gradually reduces finger-pointing and makes it easier to see where AI genuinely helps versus where gaps still need human follow-up.
An End-to-End AI Workflow for Training Users
We are still early in deploying the adoption agent, but our goal is to turn content creation from a one-way broadcast into a closed-loop system where AI informs both content creation and delivery.
One end-to-end AI-powered workflow we use today centers on content and training creation, and we're now extending it with an adoption agent.
The workflow starts with raw inputs: customer interactions, support tickets, product usage data, and case notes. AI synthesizes these to identify common questions, friction points, and feature adoption gaps across accounts.
From there, AI drafts the relevant content: training modules, documentation updates, success stories, or localized/translated materials, depending on the gaps the analysis surfaces. A human then reviews and refines this draft output before publishing it or rolling it into our training and enablement library.
The tooling for this consists of Salesforce Service Cloud, Microsoft Copilot Studio, Easygenerator, EasyCoach, and product telemetry tools.
We are also deploying a newer piece: an adoption agent built internally alongside Copilot Studio. It closes the loop so that instead of just creating training content and waiting for customers to find and use it, the agent proactively identifies where a customer or user isn't adopting a feature or workflow as expected, and surfaces the relevant training or guidance directly to them at the point of need.
The end-to-end workflow looks like this: data signals → AI-generated content/training → human review and refinement → publication → adoption agent matches the right content to the right customer at the right moment, based on their actual usage gaps.
We are still early in deploying the adoption agent, but our goal is to turn content creation from a one-way broadcast into a closed-loop system where AI informs both content creation and delivery.
How AI Creates Efficiency Gains in CX
As far as the results that we're seeing, AI has driven significant gains in scale and output. We've roughly tripled the number of success stories and seen similar step-changes in documentation automation, training/learning course creation, video content, and auto-translation.
Manual effort used to bottleneck work like writing customer stories, localizing content, or building training modules. AI now drafts this work in a fraction of the time, freeing teams to produce more content across more accounts and languages than was previously feasible.
The Split between AI Work and Human Work in CX
Overall, AI now informs several CX activities: backlog prioritization, case analysis, adoption planning, creating success stories, and realized-value reviews. In each of these areas, AI synthesizes data, surfaces patterns across cases, flags accounts at risk, and drafts initial outputs, while humans review, refine, and add judgment before anything reaches the customer.
Explicitly human activities require relationship context or strategic judgment: champion relationship management, handling escalations with key stakeholders, and value-realization conversations, where the "why" behind a recommendation matters as much as the recommendation itself. AI cannot replicate trust, tone, and lived account history in these areas.
The split is straightforward: AI compresses time spent on analysis and drafting, work that previously took hours of manual digging through cases, usage data, or notes. By offloading this to AI, we free up our teams' time to focus on higher-value activities.
Why AI Pilots Must Be Scalable From Day One
The challenges I've seen have been less about the AI itself and more about scaling from POC to production. Many of these use cases worked well in pilot or small-scale settings, but we found moving them to production-ready, repeatable processes — with the right quality controls, integration into existing workflows, and consistency across teams — harder than initial experimentation suggested.
The gap between "this works in a demo" and "this is reliably embedded in how we operate day-to-day" has been the main friction point.
When it comes to AI, experimentation is the right starting point, and moving forward without waiting for perfect certainty is key. But I'd tell teams starting out now to design for scale and production from the beginning, even while they're still experimenting.
That way, you won't get stuck in a POC model, running pilot after pilot that proves a concept works without building the underlying processes, integrations, and quality controls needed to move it into production. If you don't think about that path early, you end up with a graveyard of successful pilots that never became real operational capability.
So my advice is to keep the experimentation mindset, but pair it from day one with a clear eye on what production-readiness requires, so pilots have a real path forward instead of becoming dead ends.
Why AI Can't Replace Human Strategy in CX
The expectation that AI could scale strategic account coverage in the same way it scaled content production or documentation hasn't held up. High-touch, high-stakes relationships remain fundamentally human-driven, with AI playing a supporting rather than leading role.
Instead, AI has delivered efficiency within that process. It helps prepare for strategic conversations faster, surfaces relevant data and context ahead of meetings, and drafts materials that humans then shape and deliver. It's an accelerant to human-led strategy, not a substitute for it.
How AI has Shifted Generational Channel Preferences
One long-held assumption AI forced us to reconsider concerned generational channel preferences: The idea that younger customers default to chat/self-service while older customers default to phone or email, and that we should design experiences strictly around those generational lines.
We've found that AI has blurred those lines more than expected, but hasn't erased them entirely. Across generations, customers increasingly gravitate toward whatever channel resolves their issue fastest.
For example, a younger customer might pick up the phone for something complex, while an older customer might happily use an AI chat for something simple. In the past, we would have assumed the opposite. Speed and quality of resolution now often outweigh the channel itself.
That said, we've also seen that for certain customers — often, though not exclusively, in older generations — the need for human touch hasn't gone away. This is particularly true for complex issues, sensitive situations, or relationship-driven conversations where reassurance and trust matter as much as the answer itself. AI hasn't replaced that need. If anything, it's made the cases where human interaction truly matters more visible because AI now handles the more transactional volume.
So, rather than abandoning generational thinking altogether, we've shifted from "design by generation" to "design by need." We use AI to handle high-volume, fast-resolving interactions for everyone, while preserving and prioritizing human-led, high-touch paths for complex or emotionally significant moments, regardless of who's asking.
Why CX Leaders Must Create Differentiation with Proprietary Data
My advice to CX leaders is to focus on differentiation through your own data…Don’t compete on having AI. Compete on how well you ground your AI in the data and knowledge only you have.
My advice to CX leaders is to focus on differentiation through your own data. Generic AI capabilities are becoming table stakes; every vendor and competitor has access to similar underlying models and tools. Genuine differentiation comes from the proprietary data you've accumulated: customer interactions, usage patterns, support history, success outcomes, and institutional knowledge that's unique to your business and your customer relationships.
The leaders who will pull ahead are those who treat that data as a strategic asset and actively feed it into their AI initiatives, rather than relying on off-the-shelf AI that produces the same generic outputs others get. That means investing in data quality, accessibility, and integration now, so AI tools are working with the full context of your customer relationships rather than generic patterns.
Don't compete on having AI. Compete on how well you ground your AI in the data and knowledge only you have.
Follow along
You can follow Sofia Barbosa's work on LinkedIn.
More expert interviews to come on The CX Lead!
