Institutionalizing AI: Democratizing Artificial Intelligence Across the Enterprise

Institutionalizing AI: Democratizing Artificial Intelligence Across the Enterprise

Who is allowed to build AI agents at your company? For most organizations, this question remains unresolved, caught between the instinct to centralize control within IT and the recognition that the people closest to the work are the ones best positioned to identify where AI can create the most value. AGCO, the world's largest pure-play manufacturer of agriculture equipment and the company behind brands including Fendt, Massey Ferguson, Valtra, and PTX, has answered the question definitively. Everyone. Viren Shah, SVP and Chief Digital and Information Officer at AGCO, and Aryn Drawdy, Director of Strategic Platforms, present how the company is institutionalizing AI through a dual-track strategy that combines bottom-up citizen development with top-down enterprise-scale deployment, in a session on 1ArtificialIntelligence hosted by Sharon Chen, Head of Go-to-Market at Akila North America.

The session covers AGCO's three-pillar framework for responsible AI adoption, the structured process through which citizen developers move from personal sandboxes to enterprise deployment, the shift from task automation to decision automation in enterprise-scale AI, and the product innovation that produced the Talking Tractor, a multilingual voice interface that is transforming how operators learn to use agricultural machinery around the world.

Everyone Builds Agents: The Bottom-Up Strategy

Shah opens with a question that frames the entire session. Who is allowed to build agents at your company? The answer at AGCO is unambiguous. Every office worker has access to the company's preferred AI tools, and every team member is encouraged to become a practitioner. This is not about building a small AI team that develops solutions for the rest of the organization. It is about bringing the human along the whole AI revolution, as Shah describes it.

The strategy is deliberately bottom-up. Shah argues that the only way to scale AI at an enterprise level is to ensure that everybody participates in the journey. The alternative, where IT controls everything and delivers finished solutions to the business, fails because it cannot anticipate where agents will be most impactful. Drawdy reinforces this point directly. Nobody knows today where agents are going to pop up and be most impactful within the entire organization. By enabling citizens, the company discovers efficiencies, cost savings, and process improvements that a centralized team would never have identified.

Three Pillars: Value, Technology, and People

Shah outlines three pillars that structure AGCO's approach to intersecting citizen and enterprise development. The first pillar is value. Every AI initiative, whether citizen-developed or enterprise-built, must demonstrate value creation. But value creation alone is insufficient. It must be governed to ensure responsible and ethical delivery. The company explicitly recognizes that AI development carries risks, and the governance layer exists to ensure that value is delivered responsibly.

The second pillar is technology. Data infrastructure and AI platform architecture are critical to enabling both citizen and enterprise development. Shah describes a landscape that includes tokens, subscriptions, and products deployed across both processes and machines. The architecture must support observability, meaning the ability to see what is happening across all AI activity as it scales.

The third pillar, which Shah identifies as the most important, is people. This encompasses two dimensions. The first is change management, ensuring that as the way people work changes, the affected team members understand what is changing and are brought along the journey. The second is communication, ensuring that the work being done with AI is visible across the organization and that engagement is broad and sustained.

Citizen Development: From Sandbox to Enterprise

Drawdy leads the detailed walkthrough of AGCO's citizen development process. The starting point is executive alignment. Every member of the leadership team must understand what the company is doing and why, because the initiative extends across every function, from finance to marketing to supply chain.

Governance must be defined and, in Drawdy's words, rock solid before citizen development begins. This involves security, legal, and governance teams working together to establish the frameworks within which agents can be built. On the question of data readiness, Drawdy takes a position she acknowledges is slightly controversial. The data does not have to be perfect. The company has made a deliberate decision to meet the data where it is today, rather than spending excessive time perfecting data before starting. A sample set of data that can show quick results and enable the organization to start envisioning agentic work is more valuable than delayed perfection.

Once the foundational governance is in place, AGCO's citizen development scales through cohorts. As employees across different departments build agents for similar use cases, the company organizes them into functional cohorts, such as a marketing cohort with five or six members building related agents across different brands. Champions emerge naturally from these cohorts, individuals who volunteer to help others in their department adopt and build with AI. This organic emergence of champions is a deliberate feature of the design, not an afterthought.

Structured training runs across the entire spectrum, from employees who do not know what an agent is to experienced builders. The training is maintained as a living repository that is continuously updated. Deployment follows a controlled model that Drawdy describes as ZOE, with defined zones. In the green zone, employees have a safe sandbox where they can build agents, pull real data, and evaluate outcomes without risk. When an agent is ready for broader deployment, it enters a governance review process that evaluates data sensitivity, compliance, and security before clearing it for wider use.

Enterprise-Scale AI: From Tasks to Decisions

Shah describes the enterprise side of the equation as a fundamentally different undertaking from citizen development. Enterprise-scale AI focuses on cross-functional processes, applying what Shah calls the good old lean approach of mapping current state and future state, but with a critical new dimension. In the future state, the team must determine not only who is performing what tasks but which decisions are going to be made by the system.

Shah frames this as the biggest change in enterprise AI. Moving from task-oriented automation to decision-oriented automation requires a different kind of conversation. When he asks leaders in his organization which decisions they are willing to delegate to AI, the response is often existential. What am I doing tomorrow, then? The resolution comes through recognizing that many decisions are repetitive and rule-based, and that delegating those decisions to an agent frees the human to focus on judgment-intensive work, while remaining in the loop to see which decisions are being made.

The second critical dimension of enterprise AI is reusability. AGCO is deliberately designing its agent architecture to prevent proliferation, ensuring that similar agents across different functions are consolidated rather than duplicated. The goal is to avoid a landscape where many agents are making similar decisions in slightly different ways, which would compromise the integrity and coherence of enterprise AI.

AGCO partners with Microsoft and Google to support the enterprise AI process. The structured methodology involves identifying which agents to build, how they will interact with each other, and which integration agents will pull data from underlying systems such as SAP. Shah emphasizes that this takes longer than citizen development because of the complexity of cross-functional orchestration, but the citizen development track accelerates the enterprise track by building organizational fluency with AI concepts and practices.

Guardrails with Speed: The Playground Model

Chen asks Shah to reconcile what appears to be a contradiction: guardrails and speed. Shah's answer is a metaphor that captures the entire philosophy. Think of it as a playground. In every playground, children are free to jump, swing, and do all kinds of things. But there are safety measures, like mats below the equipment, that protect them if they fall. AGCO creates zones where people can play and practice freely. The moment they say they want to share an agent with others, the governance checkpoints activate to ensure that policies, practices, and data sensitivity requirements are met.

Shah describes three zones in the playground. The first is the personal zone, where an employee builds and uses an agent within their own M365 license, a completely safe place to learn and experiment. The second zone is triggered when the builder wants to share the agent with a team or department, which initiates the governance review. The third zone is full enterprise deployment, which passes through the complete stage-gate process. The key insight is that creativity is unrestricted at the point of origin, and governance intensifies only as the blast radius of the agent expands.

The Infinity Loop: Where Citizen Meets Enterprise

Drawdy describes the relationship between citizen development and enterprise AI as an infinity loop. Citizen-developed agents that prove valuable within a single department are naturally candidates for elevation to enterprise-scale deployment. An agent that worked for one brand's marketing team can be tested across other brands and, if validated, becomes a super agent that serves the enterprise. Conversely, the experience of working with enterprise agents informs how citizen developers think about their own work, creating a continuous cycle of learning and scaling.

Shah codifies this with a pyramid model for organizational AI capability. The base is AI-aware, meaning every employee understands what AI can do. The middle tier is AI practitioners, a large percentage of the organization who actively build and use agents. The top tier is AI experts, a smaller group with deep technical capability. The only way to scale AI at an enterprise level, Shah argues, is to ensure the base of the pyramid is as broad as possible.

Change Management That Is Not Traditional

Drawdy, who owns change management as part of her responsibilities, is direct about the nature of the challenge. AI change management is not like an SAP project. In a traditional enterprise software deployment, you know how you are going to start and you know what the finish line looks like. In an AI project, the outcome may look very different from what was originally planned, hopefully for the better, but the unpredictability demands a different approach.

AGCO's approach is to embed change management from the very beginning of every project, including enterprise-scale AI efforts. Drawdy shares a lesson learned from a warranty-related project where the team initially planned to bring change management in later, then quickly realized that things were changing so fast, with new stakeholders being pulled in at every workshop and demo, that the change management function needed to be present from the start. The underlying principle is that if the change management team is not in lockstep throughout the process, the organization will be surprised at the end in ways that undermine adoption.

The Talking Tractor: AI in the Product

Shah closes the session by turning from process AI to product AI. AGCO recently debuted the Talking Tractor at Agritechnica, one of the world's largest agricultural machinery exhibitions. The product addresses a fundamental challenge for farmers globally. The difficulty of onboarding new operators onto complex agricultural machinery across multiple geographies and languages. AGCO's machines require significant learning before an operator can use them effectively, and the global customer base spans dozens of languages.

The Talking Tractor is an AI-powered voice interface that allows operators to communicate with the tractor in their language of choice, enabling faster onboarding and more intuitive interaction with the machine. It is a direct illustration of how AGCO's AI strategy extends beyond internal processes into the products themselves, delivering value directly to the farmer in the field.

The Future: Everyone Becomes a Technologist

Shah frames the broader vision with an analogy that resonates across industries. There was a time when only a few people in the world were typists. Then the keyboard became ubiquitous, and the whole world became typists. The same transformation is underway with AI. We are all going to be technologists, Shah says. The future, in his view, is about leveraging human creativity with AI capability to address not only corporate problems but social ones as well.

Drawdy adds that the cultural shift is already visible inside AGCO. Collaboration across departments and teams that previously operated in silos is generating positive change, and the culture of the organization is evolving alongside the technology. The excitement is not about AI as an abstraction. It is about what happens when an entire organization is empowered to build, experiment, and improve together.