Enterprise AI adoption is not a matter of if but how.
The largest companies in the world are making significant investments across tech, people and resources, angling for an advantage that could pay dividends for those that establish leadership positions.
Pilots and experiments are easy. The real challenge is embedding AI so deeply into an organization’s ways of working that it becomes second nature. That requires more than commitment to adoption of new tools or processes. It means cultural transformation defined by visible leadership, sustained reinforcement and a willingness to push people past the comfort of familiarity.
When I was tapped to develop an AI-first culture within Rakuten Symphony, I drew on lessons from earlier company transformation programs to make AI not just present, but unavoidable. I already had experience leading large-scale organizational change and understood Rakuten’s commitments to such changes. Years earlier, Rakuten undertook “Englishnization,” a company-wide mandate that accepted nothing less than reaching your target proficiency.
This is my perspective on Rakuten Symphony’s journey toward an AI-first culture. Read on for the moments that stood out, the lessons they revealed and why they may resonate beyond our walls.
The tl;dr is full-scale adoption does not happen by accident. This was never going to be just another technology rollout. The mission was cultural transformation, with AI embedded in the day-to-day work of every single person in the company.
Expansive scope made AI impossible to ignore
The scope of Rakuten Symphony’s “AI-nization” project covered all Rakuten Symphony staff, across all roles and regions. Our initial self-assessments revealed only about 30% adoption, and even that figure was inflated by theory. People could talk about AI and experiment with the occasional tool, but few were integrating it into their workflows in a meaningful way. It was immediately clear that adoption would not happen organically. We needed a top-down push that kept AI front and center until it became second nature.
We built AI into the company’s rhythm.
I personally contacted employees. Our group Asakais (i.e., regular company-wide meetings) dedicated half the time to this project with every other topic that previously consumed agendas pushed into the background.
The conversation did not fade after launch. Progress against undertaken initiatives was visible every week with leadership communicating clear participation expectations.
As a result, response rates to AI adoption surveys were not optional. If you were not engaging, you might not have a place here. This was the reality of a deliberate, sustained push, modeled in part on the visibility and accountability that had made Englishnization successful.
It worked. Over time, adoption climbed from 30% to 100%.
Measuring actual impact
We anchored the program around a “triple 20” goal: 20% efficiency gains in marketing, operations and client-facing work.
AI adoption surveys tracked usage, but we also measured tangible results. Optical character recognition transformed RF planning document reviews. Retrieval-augmented generation pulled from our knowledge base to build faster, sharper network operations and customer documentation. These were embedded into the way we worked, not set up as isolated experiments.
As the results came in, we recorded 37% reduction in critical task times, more than 32% code suggestion acceptance rates from developer AI assistants and considerable OpEx savings that are anticipated to continue, with potential for further increase.
One standout story comes from our OSS business unit.
This team placed AI at its core, developing both AI-enhanced products for clients and internal platforms that fundamentally changed how we work. They pioneered an internal AI agent studio, which lets employees create and provision their own AI agents for workflows, and introduced AIOps agents that monitor core cloud platforms and production systems, automatically triggering incident responses and running root cause analysis to keep services running reliably with proven performance within the Rakuten Mobile network.
Other notable use cases born out of our cloud and RAN business units featured advanced Kubernetes-specific AI Agents with domain and product knowledge, giving us network awareness at scale, and creation of RAN software development assistants capable of code refactoring, bug identification and feature design support.
These tools were built in-house using our own implementation knowledge and binaries, meaning they were tailor-made for our company’s needs.
The result has been dramatic: not just productivity gains, but higher accuracy, shorter release cycles and faster GTM execution.
What we learned from what didn’t work
Not every initiative hit the mark.
Early on, we experimented with commercially available AI tools that performed impressively for general use cases but ultimately did not meet our requirements for security, domain knowledge or access to proprietary implementations.
Rather than seeing this as a failure, it became a turning point.
Our experience taught us that Rakuten Symphony’s AI transformation needed to be built from within, using our own expertise and datasets. That lesson led directly to the development of our in-house agentic platforms and AI toolkits, which now serve as a natural extension of our teams.
Looking ahead
I see the future of our organization as high-output human contributors overseeing teams of AI agents, with initially up to 20 agents per individual and scaling from there. It is a way of pushing adoption into new territory, making AI a partner rather than just a tool.
Ultimately, AI adoption is not a moment in time. It is a steady march, renewed constantly, with the goal of making AI as normal in our workflows as email or spreadsheets.
If you are wondering how to get your people to truly change and adopt AI, my advice is simple: make it visible, make it a priority and don’t give up early.
As for what not to do? My humble advice is to not treat AI as an optional experiment or simply an external buzzword. Our journey from 0% to full adoption of our initial goals took persistence, pressure and proof. Hitting those milestones was only the beginning. We continue to raise the bar each year, deepening adoption and expanding benefits.”
Mention Ahmad Farid in the comments to ask questions about his journey or share your own experiences to start a conversation.
.svg)
