From MIT Technology Review · · 4m
From pilots to scale: how enterprises build AI that works
As AI moves from isolated experiments into mission-critical workflows, companies must rethink data, processes, governance and people together.
In brief
As AI moves from isolated experiments into mission-critical workflows, companies must rethink data, processes, governance and people together. Enterprises moving AI from pilots to scaled operations must treat it as an operating model transformation, not a technology project, embedding governance and data foundations…
The pilot trap
Seventy-four percent of organizations are stuck in pilot mode with no plan to scale, according to a BCG report cited by Thorsten Leiduck, SVP and global head of SAP Business AI Platform Ecosystem.
“74% of the organizations are stuck in pilot mode with no plan to scale”
Data quality blocks scaling
Forty-two percent of organizations struggle to get access to data and quality needed to scale AI capabilities across the organization, an Anthropic survey found.
“42% are really struggling with getting access to the data and the quality”
The AI department trap

Only fourteen percent of global two thousand enterprises have successfully scaled AI, Infosys and HFS research found. Eighty-six percent remain in early exploration, proof of concept, or fragmented pilot phases.
“only 14% have successfully scaled AI. Eighty-six percent remain in early exploration”
Why early movers win
Only eight percent of companies qualify as AI native today, yet they achieve four times the revenue contribution from AI versus laggards, the Infosys Knowledge Institute AI Maturity Index shows.
“only 8% of the companies qualify as AI native today, yet they achieve four times the revenue contribution from the AI versus laggards”
Accuracy matters in mission-critical work
Eighty percent accuracy might pass in consumer settings, but in supply chains or financial close, that level is unacceptable, Leiduck says. Stakes are much higher in business-critical processes.
“80% might be good enough... but in areas such as supply chains or financial close, that level of accuracy is a very terrible result”
Three lenses for evaluating AI use cases
Satinder Pal Singh, Associate VP North America for SAP BAIP and SAP AI Strategy at Infosys, recommends evaluating potential use cases through impact, feasibility, and scalability.
Impact: where agents move the numbers
The first lens asks where deploying an agent moves metrics that matter: revenue, cost, working capital, risk. Infosys has built a library of over three hundred enterprise use cases across industries.
Feasibility: can your data support it
The second lens determines whether your data foundation, platform and talent can execute the high-impact use case. A high-impact use case on a broken data foundation is not viable.
“The high-impact use case on a broken data foundation is not viable”
Scalability: pilots that cannot scale are experiments
The third lens asks what architecture lets you replicate a proven use case with clear ROI metrics. Singh says a pilot that cannot scale is just an experiment, not a path to value.
“A pilot that can't scale is just an experiment”
Data is the foundation
AI is absolutely dependent on the data foundation, Leiduck says. Without data structures and the ability to access the right information at the right time, there will be no successful AI.
“AI is absolutely dependent on the data foundation”
Build a semantically rich data layer
Companies need to establish a semantically rich connected data layer that reflects their business, so agents are grounded in the company's reality and not just in generic patterns.
Governance prevents risk backlogs
Clear policies, access controls, monitoring and oversight ensure agents remain trustworthy and aligned with business requirements. The absence of governance does not speed up AI adoption; it creates a risk backlog.
“The absence of governance doesn't speed up AI adoption. It creates a risk backlog”
Nearly half deploy AI without governance
Forty-eight percent of companies deployed generative AI without a formal governance policy, the Infosys Knowledge Institute generative AI radar found. Nearly half ran AI with no documented framework or oversight.
“48% of companies had deployed generative AI without a formal governance policy”
SAP Business AI Platform three pillars
The platform consists of build, contextualize and reason, and governance. Build provides intent-based development tools. Contextualize and reason establishes a data foundation and aggregates over fifty language models.
SAP Agent Hub governs enterprise AI
SAP Agent Hub is a command center for enterprise-wide AI and agent governance. It manages the end-to-end life cycle of agents and features from hire to retire, including non-SAP agents integrated via A2A.
Month-end close transformed
A global diversified manufacturer deployed Topaz Fabric on Business AI Platform with a master agent coordinating five specialist child agents for month-end close.The result eliminated the entire L1.
Specification documents cut by ninety-two percent
A North American utility in an S4HANA transformation reduced complex SAP functional specification documents from eighty-six hours to six hours, a ninety-two percent reduction, using AI native development on Business AI Platform.
“cut complex SAP-specification documents from 86 person hours to 6, a 92% reduction”
Time to value drops dramatically
Combined effects of AI-assisted delivery and governance frameworks drop time to value from twelve to eighteen months down to four to six months, Singh says.
“time to value drops from 12 to 18 months, down to four to six months”
AI as decision-making partner
The longer-term opportunity is a shift from AI as a productivity tool to AI as a decision-making partner, helping leaders understand what is happening, anticipate what comes next, and determine how to respond.
“AI as a decision-making partner”
What matters most
Enterprises moving AI from pilots to scaled operations must treat it as an operating model transformation, not a technology project, embedding governance and data foundations from the start.
Read the TL;DR — this story on one card.


