Why most enterprises stay stuck at the pilot stage
AI dominates the headlines, but most enterprises still struggle to move their initiatives from isolated pilots and proofs of concept to a production deployment aligned with business outcomes. The gap between the enthusiasm on display and the value actually delivered remains one of the most discussed friction points of AI transformation in 2026.
In the webinar Scaling AI: Turning Hype into Business Value, our speakers explore a common barrier: many teams take a technology-first approach to AI, ignoring the organizational complexity that underpins real scaling. Stacking models and tools isn't enough if the operating model, data quality and governance don't advance at the same pace.
To address this, the webinar presents the FOREST framework, a structured lens built on six dimensions that treats AI scaling as an organizational transformation as much as a technical one.
The FOREST framework at a glance
FOREST brings together six dimensions whose initials form the acronym: foundational technology architecture, operating model, ready data, human experience, strategic alignment, and trusted AI. Each answers a different question, but none can be addressed in isolation without weakening the others.
Three questions come up systematically in organizations struggling with scaling: where should AI be built and by whom, what data can you actually rely on, and how do you prove that a use case creates measurable value rather than a one-off demo. The following chapters break down each of the six dimensions, with the concrete levers shared by the webinar's speakers.
Technology architecture and operating model: the foundations
A modular architecture rather than a stack of tools
The framework's first dimension, foundational AI architecture strategy, consists of building a modular, scalable technology foundation rather than stacking tools as ad hoc needs arise. In practice, this means defining the target architecture, data platform, model orchestration layer, inference gateway, observability, before purchasing tools, not the other way around.
Centralized, federated, or a hybrid model
The second dimension, the operating model, consists of balancing centralized and federated structures to drive business-aligned execution. A purely centralized AI center of excellence standardizes practices but slows down local execution; a fully federated model speeds up business units but multiplies duplicated effort and compliance risk. The hybrid model, where a shared platform is made available to empowered product teams, now dominates the most mature organizations.
Ready data for AI: quality, semantics and observability
The framework's third dimension, ready data, directly determines the reliability of any AI system built on top of it. Without high-quality, contextualized, governed data, even the most capable model will produce results that aren't usable in production.
Shifting quality left in the pipeline
The speakers stress the value of shifting data quality left in the pipeline: validating and cleaning as close to the source as possible rather than fixing things downstream, in the data warehouse or, worse, in the prompt sent to the model. This approach cuts the cost of correction by a significant factor compared with late detection.
Knowledge graphs and semantic layers
Standardizing business semantics through a shared reference, often materialized as a knowledge graph, lets different AI systems share a consistent understanding of entities and their relationships, rather than reinventing divergent definitions project by project.
Data and model observability
A continuous observability layer, data drift, model performance drift, decision traceability, completes the picture. It turns data governance from an after-the-fact control into an early-warning signal that can be acted on before the incident reaches the end user.
Human experience and strategic alignment: connecting AI to the business
Designing for real adoption, not for the demo
The fourth dimension, experience for humans in AI, is a reminder that a technically flawless system poorly integrated into users' actual working habits ends up being bypassed. Designing interfaces focused on real needs and on adoption, rather than on feature completeness, remains the most underrated factor in scaling.
Connecting use cases to measurable business objectives
The fifth dimension, strategic alignment, requires linking every AI use case to a business metric tracked by leadership, before development even starts. It's this direct link between business value demonstrated at the use-case level and executive buy-in that secures the sponsorship needed to fund scaling beyond the budget of an isolated pilot.
Trusted AI: governance, compliance and the Zeiss case
The sixth dimension, trusted AI, establishes governance, compliance and ethical guardrails from day one rather than as a catch-up exercise after an incident or a regulatory audit.
Governance by design
Documenting the traceability of automated decisions, mapping risk by use case, and defining escalation thresholds to a human are prerequisites, not options, in a regulatory environment that keeps tightening, the European AI framework chief among them for organizations operating in Europe.
The Zeiss case: governing AI in a regulated industrial environment
The speakers illustrate these principles with the case of the Zeiss group, where AI governance has to work within traceability and quality requirements specific to precision manufacturing and optics. The approach combines federated teams, accountable for their own use cases, with a central governance framework that sets the non-negotiable guardrails.
Toward agentic AI: laying the foundations for tomorrow's capabilities
What autonomy changes
The webinar closes with a forward-looking warning: agentic AI, where autonomous agents execute tasks and make decisions across entire workflows without human validation at every step, changes the very nature of the risk to be governed. The six dimensions of the FOREST framework don't disappear with agentic AI, they harden: the architecture must anticipate multi-agent orchestration, and governance must cover decisions made in cascades.
Federated teams and clear accountability
Establishing federated teams with clear accountability by agent domain becomes a condition of safety as much as of operational efficiency: without an identified owner for every agent in production, the traceability of autonomous decisions quickly becomes untraceable.
Conclusion: from pilot to production
Laying solid foundations today across the six dimensions of the FOREST framework is the best investment for harnessing agentic AI's capabilities tomorrow. The organizations that succeed at scaling aren't the ones that adopt the newest model first, but the ones that align architecture, operating model, data, human adoption, strategy and governance at the same pace.
The Scaling AI: Turning Hype into Business Value webinar brought together Danilo Sato, VP Global AI, Adservio; Tiankai Feng, Director, Data and AI Strategy, Adservio; Amy Raygada, Principal AI and Data Strategist, Adservio; and Norbert Gergeli, Head of Data Architecture and Lifecycle, Zeiss Group.
Disclaimer: The statements and opinions expressed in this article are those of the authors and do not necessarily reflect the positions of Adservio.
STAY POSTED
Get our next analyses and field notes straight to your inbox.
