How to Scale AI: The importance of modernizing data and integrating processes
Discover how to prepare your processes and data—getting them into a condition where AI can use them to deliver on its promise of radical transformation. In this thought provoking article, two AI experts from SAP explain what’s involved, where to get started, and which SAP solutions can help you really put AI to work. Dokument herunterladenHow to Scale AI: The importance
of modernizing data and
integrating processes
By Jürgen Butsmann, Product Marketing, SAP Cloud ERP Private
and Terry Penner, Product Marketing, SAP Business Technology Platform
of modernizing data and
integrating processes
By Jürgen Butsmann, Product Marketing, SAP Cloud ERP Private
and Terry Penner, Product Marketing, SAP Business Technology Platform
2 | 10
Expectations of the potential impact of AI have
never been higher. But so far, most projects have only
produced incremental improvements in isolated areas.
So, what do you have to do to get AI to deliver material
change across your business? The answer lies in
preparing your processes and data—getting them into
a condition where AI can use them to deliver on its
promise of radical transformation. In this thought-
provoking article, two AI experts from SAP explain
what’s involved, where to get started, and which SAP
solutions can help you really put AI to work.
How to Scale AI: The importance of modernizing data and integrating processes
Expectations of the potential impact of AI have
never been higher. But so far, most projects have only
produced incremental improvements in isolated areas.
So, what do you have to do to get AI to deliver material
change across your business? The answer lies in
preparing your processes and data—getting them into
a condition where AI can use them to deliver on its
promise of radical transformation. In this thought-
provoking article, two AI experts from SAP explain
what’s involved, where to get started, and which SAP
solutions can help you really put AI to work.
How to Scale AI: The importance of modernizing data and integrating processes
3 | 10How to Scale AI: The importance of modernizing data and integrating processes
The challenges in reaching AI’s potential
However, return on investment has so far been
disappointing and IT teams are finding it
challenging to scale AI beyond point solutions.
They’re also nervous for ethical and security
reasons about exposing their proprietary
business data to public large language
models (LLMs).
Observations from customers
The discussions we’ve been having with SAP
customers have revealed these common
challenges IT leaders face in realizing AI’s
potential.
• Insecure architecture. Using public LLMs
alone is not always reliable. AI really needs to
use the organization’s own business data. But
exposing corporate data to LLMs raises security
and ethical issues.
• Fragmented systems. Applying AI to a process
that is contained within a single system is
relatively easy. But most real business
processes span multiple departments and
systems. Often these applications and their
data are not integrated which prevents AI from
carrying out business processes end to end.
• No single semantic layer. Enterprise
applications today, whether SAP or non-SAP, all
have differences in their data structures. Data
transformation and harmonization are necessary
to allow the sharing of transactional data, and
for analytical processes and AI functions to
take place.
• One-way data transformation. Data
warehousing, as practiced by SAP and by other
vendors in the past, focused on extraction
and one-way transformation, storage, and
processing of data, decoupled from original
data structures and business process context.
Effective, enterprise-wide AI, however, requires
two-way transformation and re-integration of
data into business processes as they run.
What organizations need
AI agent networks need a solid data foundation
if they are to process queries across systems
effectively. The organization’s data structures—its
master data and system architecture—need to
be harmonized so AI can understand the different
systems in a consistent and accurate way.
When processes are integrated, consistent data
can be stored in the systems. Integrating systems
and maintaining consistent data will overcome
the problem of isolated, fragmented systems
and allow AI to execute end-to-end business
processes.
Business leaders believe AI has the potential to transform their organizations.
They see it accelerating cycle times, reducing errors, performing processes
automatically, and yielding insights. Their initial experiments and projects have
reinforced this expectation by delivering productivity improvements, albeit in
isolated areas.
1. “The GenAI Divide: State of AI in Business 2025”, Preliminary Findings from AI Implementation Research from Project Nanda, MIT, 2025.
The challenges in reaching AI’s potential
However, return on investment has so far been
disappointing and IT teams are finding it
challenging to scale AI beyond point solutions.
They’re also nervous for ethical and security
reasons about exposing their proprietary
business data to public large language
models (LLMs).
Observations from customers
The discussions we’ve been having with SAP
customers have revealed these common
challenges IT leaders face in realizing AI’s
potential.
• Insecure architecture. Using public LLMs
alone is not always reliable. AI really needs to
use the organization’s own business data. But
exposing corporate data to LLMs raises security
and ethical issues.
• Fragmented systems. Applying AI to a process
that is contained within a single system is
relatively easy. But most real business
processes span multiple departments and
systems. Often these applications and their
data are not integrated which prevents AI from
carrying out business processes end to end.
• No single semantic layer. Enterprise
applications today, whether SAP or non-SAP, all
have differences in their data structures. Data
transformation and harmonization are necessary
to allow the sharing of transactional data, and
for analytical processes and AI functions to
take place.
• One-way data transformation. Data
warehousing, as practiced by SAP and by other
vendors in the past, focused on extraction
and one-way transformation, storage, and
processing of data, decoupled from original
data structures and business process context.
Effective, enterprise-wide AI, however, requires
two-way transformation and re-integration of
data into business processes as they run.
What organizations need
AI agent networks need a solid data foundation
if they are to process queries across systems
effectively. The organization’s data structures—its
master data and system architecture—need to
be harmonized so AI can understand the different
systems in a consistent and accurate way.
When processes are integrated, consistent data
can be stored in the systems. Integrating systems
and maintaining consistent data will overcome
the problem of isolated, fragmented systems
and allow AI to execute end-to-end business
processes.
Business leaders believe AI has the potential to transform their organizations.
They see it accelerating cycle times, reducing errors, performing processes
automatically, and yielding insights. Their initial experiments and projects have
reinforced this expectation by delivering productivity improvements, albeit in
isolated areas.
1. “The GenAI Divide: State of AI in Business 2025”, Preliminary Findings from AI Implementation Research from Project Nanda, MIT, 2025.