By Sumit Jha, Co-Founder & Director, CXBERRIES
Everyone seems to have an AI success story today. But very few have a scaling story.
The rush to invest in AI has taken over boardrooms. In my experience, investments are driven by efficiency, cost, and competitive pressure. Across industries, teams are building pilots and demonstrating early wins, but most don’t go much further.
Research from MIT Sloan Management Review and Boston Consulting Group suggests that nearly 70% of AI and digital transformation projects never get out of the “stuck” phase; they struggle to scale or deliver meaningful business value. Only a small percentage, often cited at under 10%, actually see enterprise-wide impact.
While these statistics highlight the scale of the problem, they also point to a much deeper issue – Enterprises are trying to bolt AI onto old operating models that just aren’t built for it.
Where things start to break
Most organisations are not failing at the pilot stage. Many pilots work exactly as intended. Models perform well, workflows get automated, and teams see immediate benefits.
In several engagements, this is exactly where momentum starts to fade – what worked in isolation becomes difficult to replicate across teams, systems, and processes.
A big reason is how these initiatives are set up. They usually sit within specific functions like Business, IT, Operations, or Support each solving a local problem. There is rarely a view of how they connect.
What’s missing is a unified strategy. So what we get are small pockets of efficiency. Not transformation.
And often I have seen that there’s a tendency to take existing processes as they are and layer AI on top. If workflows are fragmented, AI doesn’t fix them; it exposes the gaps. This aligns with broader industry observations, including findings from MIT Sloan, which highlight that while organisations actively experiment with AI, far fewer redesign their operating models to support it at scale.
What’s actually missing
A clear pattern emerges – organizations are trying to scale AI on operating models never designed for it. This shows up in a few ways:
First, data is still scattered across systems, with little consistency. Teams spend more time reconciling information than using it.
Second, processes haven’t really changed. Automation is introduced, but the workflow around it remains the same.
Third, organisational change is often underestimated, with resistance and lack of adoption becoming a critical barrier.
Finally, and most importantly, functions still operate independently. There is limited integration between business, IT, and operations teams.
Interestingly, broader industry research also points to this. Fewer than 20% of organisations have meaningfully reworked their processes and roles to support AI at scale. Which explains why progress stalls after the initial phase.
A shift that’s beginning to happen
I’ve worked with organisations where we have approached this differently. Instead of asking where AI can be applied, we have stepped back and looked at how work flows across the enterprise.
This changes the focus. It’s no longer about individual use cases. It’s about how data, workflows, and decisions connect.
In these environments, automation isn’t treated as a standalone initiative. It becomes part of how operations are run.
The impact is easier to see:
Better visibility across systems
More consistent decision-making
Fewer handoffs and delays
Clearer linkage to value and business outcomes
Clear visibility into human vs AI activity in the workflow
The focus shifts from isolated improvements to integrated automation.
The question of value
Another reason many initiatives don’t move forward is simple – value isn’t clear.
Pilot-level benefits are visible, but they rarely translate into enterprise metrics at scale. Without that connection, priorities shift. Budgets get reallocated. Momentum drops.
Organisations that are seeing better outcomes tend to be more deliberate here. They define what success looks like early on, track it continuously, and make sure both business and IT teams are accountable.
I’ve worked with organisations to identify and prioritize their use cases based on ‘Return on Value’, covering tangible and intangible aspects. As a result, enabling them to make value-driven investment decisions.
So, what needs to change
The conversation around AI is evolving. It’s moving away from “what can we automate?” to something more fundamental – “how should this actually work at scale?”
Because scaling AI is less about adding capability and more about aligning how the organization operates. And that takes time. It also requires a willingness to rethink existing structures, not just improve them.
In some of our work, we’ve used our structured automation framework covering golden data sets, data and cross integration, orchestration, AI observability & governance and value tracking.
Closing thoughts
Most AI initiatives don’t fail because the technology falls short. They struggle because the environment around them isn’t ready. Until these changes, pilots will succeed, but scaling will remain the exception. In my experience, organisations that successfully scale AI don’t treat it as a technology initiative, but as an operating model shift.