SAP and AI: why most projects generate no ROI – and how to change that

AI projects keep multiplying. So do the budgets. But measurable ROI is still rare. Why?

Not because AI does not work. Because the conditions are not in place. That is the picture we set out with SAP and Snowflake at SAP Connect on 19 March.

And it is the starting point for thinking we have been doing with our clients for several years: Data & AI ROI does not come from deploying a model. It is built, stage by stage, on solid foundations. 

The problem: SAP data trapped in its silos

In most organizations, ERP data is both the most valuable and the hardest to put to work.
Finance, customers, products, supply chain – it is all there, structured, reliable. But walled off. 

So when you want to cross-reference: 

• Data SAP with IoT data, and,

• Marketing data or external feeds.

Projects run into expensive ETL pipelines, loss of business context, patchy governance and limited scalability. The cost of extraction destroys part of the value before the AI has even started working. 

That is exactly the problem the SAP × Snowflake partnership addresses. 

The two conditions for AI ROI

We now frame our answer around two ideas we use systematically on our engagements: 

1. AI Readiness – Is your data ready to feed AI?

That covers governance (certified, traceable, reliable data), building consumable data products, and enriching the foundation semantically. Without AI Readiness, AI trains on noise.

1. AI Gravity – Does the AI run where your data is?

The principle is simple: moving data is expensive and creates risk. AI should run as close to the data as possible, not the other way round. That is what Cortex AI on Snowflake and Joule on SAP make possible – native execution, with no duplication. 

When either condition is missing, the AI initiative fails. No ROI. The equation we put forward: AI Readiness + AI Gravity + integrator expertise = measurable Data & AI ROI. 

A concrete case: €4.6M a year in heavy industry

To show what those conditions make possible, we presented the case of a European heavy equipment manufacturer whose ambition is to grow its after-sales revenue from $120M to $180M by 2029. 

-> The flow works like this 

• connected equipment (IoT units) sends telemetry and health monitoring data in real time.

• A machine learning model calculates how long before a potential failure. 

• As soon as an alert is raised, the system cross-references it with the SAP data – is the contract active? Is there a maintenance service in place? If not, an alert goes automatically to the sales rep covering that client, to propose a proactive intervention.

-> That flow: semi-structured IoT data + structured ERP data + AI + a sales process – generates an estimated €4.6M a year across three use cases:

• Operational maintenance,

• On-site support,

• Remote support. 

   

What makes that case possible? The combination of SAP BDC (business context, governance) + Snowflake  (bringing in external IoT data, running the AI) + KPC (architecture, integration, roadmap). 

The Snowflake Zero Copy connector: what changes structurally

The SAP × Snowflake partnership rests on one central technical mechanism: Zero Copy. SAP data is available inside Snowflake with no physical duplication. No ETL pipelines to maintain. The SAP business context is preserved. Governance gets simpler. No extra silo is created. 

It changes the foundation. For 20 years, Data and Process have grown separately inside organizations. AI is what brings them together – provided you have the right base to do it on. 

The roadmap we recommend: Open → Enrich → Activate

We do not believe in big bangs. What we recommend to our clients is a three-stage path: 

• Stage 1 -> Open. 
  Make the ERP data available through SAP BDC (zero-copy). An immediate quick win, with no rebuild of the existing system.

• Stage 2 -> Enrich
  Connect the external data – IoT, marketing, logistics – onto Snowflake. 
  The information scope widens, and the AI models have more to work with. 

• Stage 3 -> Activate. 
  Launch the first cross-cutting AI use cases and feed the results back into SAP. The processes become augmented. 

At every stage: governance, no silos, quick wins combined with building foundations that last. 

KPC's role

KPC comes in as the third pillar of that equation. Our dual expertise – the SAP world and the data & AI world – lets us activate both conditions at once:

building AI Readiness (governance, data products, semantics) and deploying AI Gravity (Snowflake architecture, Cortex AI, SAP BDC integration). 

Our concrete way in:

• an AI Readiness Assessment engagement

• assess how ready your data foundation is for AI,

• identify the quick wins, define the roadmap.