- SAP Managed Tags
Generative AI has changed how we interact with systems. Writing emails, summarising documents, chatting with copilots - all of that is now table stakes.
But when I work with enterprises, I keep seeing the same pattern: The most important business decisions are not made in text. They are made in tables.
Employees, sales deals, payments, supply chains, logistics, finance - these are structured datasets. Rows. Columns. Relationships. And this is where many AI initiatives quietly struggle.
Gartner puts some hard numbers behind this reality:
- 57% of organisations admit their data is not AI‑ready
- Less than 30% of CEOs are satisfied with the returns from current AI investments
- By 2027, task‑specific AI models are expected to outperform general‑purpose LLMs by 3x for decision accuracy
The problem isn’t AI ambition. It’s using the right kind of AI for the right kind of data.
Why Generative AI Alone Falls Short for Enterprise Decisions?
Large language models are excellent at natural language tasks. They predict the next word incredibly well. But enterprise systems don’t need the next word. They need the next outcome.
When applied to structured data, general LLMs often struggle with:
- Numeric accuracy
- Understanding relationships across multiple fields
- Explainability and governance
That’s not a failure of the technology. It’s simply a mismatch. Structured data requires relational intelligence, not linguistic fluency.
The Case for Relational Foundation Models
This is where SAP RPT‑1 becomes especially relevant.
Play around with SAP RPT-1 at https://rpt.cloud.sap/
SAP RPT‑1 is not an LLM. It is a relational pre‑trained transformer, built specifically for tabular enterprise data. Instead of predicting text, it predicts business outcomes by understanding how columns and rows influence one another.
In practice, that difference is profound. Let me illustrate using a few real‑world use cases I’ve been working with.
Use Case 1: Employee Attrition Risk Intelligence
One of the most powerful scenarios is employee attrition prediction. I have built this app EARI - using SAP RPT-1 and vibe coding again using SAP BTP.
Attaching the sample screenshots for reference. I will be sharing a blog on how to build it soon.
Here, employees are not reduced to sentiment or survey comments alone. Instead, the model looks at structured signals such as:
- Tenure
- Salary vs market benchmark
- Overtime hours
- Months since last promotion
- Engagement score
- Performance rating
- Training hours
- Manager changes
Based on these inputs, the model produces:
- A quantified attrition risk score
- Clear risk categories (low, medium, high)
- Key risk drivers contributing to the score
- Actionable retention recommendations
For example:
- Below‑market salary and excessive overtime surfaced as dominant risk drivers
- The system recommended compensation review, workload redistribution, and targeted learning interventions
This moves HR conversations from gut feel to data‑backed decisions, without treating people like black‑box probabilities.
Use Case 2: Deal Win Probability in Sales
Another scenario focuses on sales decision‑making.
Sales teams typically juggle:
- Deal value
- Time in pipeline
- Competitor presence
- Product fit
- Decision‑maker engagement
- Budget confirmation
- Historical win/loss patterns
Instead of reacting late in the cycle, a predictive model can estimate win probability early and explain why.
In this scenario:
- Structured deal data was analysed
- A probability score was generated
- Clear recommendations followed, such as focusing on ROI articulation or escalating executive engagement
The result is not automation replacing sellers, but AI augmenting judgement — exactly where enterprise AI should land.
Use Case 3: From Dashboards to Decisions
Across HR, sales, finance, and supply chain, there’s a common shift happening.
Static dashboards answer: “What happened?”
Predictive, relational models answer: “What is likely to happen next — and what should we do about it?”
Gartner refers to this evolution as Decision Intelligence, and predicts that by 2027, AI‑driven decision systems will augment or automate 50% of enterprise decisions.
This is not about replacing human decision‑making. It’s about improving decision quality at scale.
Trust, Governance, and Enterprise Reality
Another important aspect often overlooked in AI discussions is trust.
- Enterprise AI must be:
- Explainable
- Governed
- Secure
- Compliant
Relational models implemented within enterprise platforms make this feasible. Data stays within controlled boundaries, predictions are traceable, and outcomes can be validated — something that becomes critical when AI directly influences people, revenue, or compliance‑sensitive processes.
Gartner consistently stresses that AI without trust is a liability, not an advantage.
So Where Does This Leave Generative AI?
Generative AI is not going away - and it shouldn’t.
It excels at:
- Conversational interfaces
- Summarisation
- Documentation
- Knowledge assistance
But for enterprise decisions, generative AI alone is not enough.
The future is hybrid:
- Generative models for interaction and experience
- Relational foundation models for prediction and decisions
When these are combined thoughtfully, AI stops being a demo and starts becoming a business capability.
Final Thought
The biggest shift I’m seeing isn’t about technology. It’s about mindset.
- From experimentation to outcomes
- From dashboards to decisions
- From generic AI to domain‑specific intelligence
Enterprises that recognise this early will move faster, decide better, and waste less time chasing AI hype that never translates into value.
And in my experience, that’s where the real transformation begins. Even for me as a developer for almost a decade and then another decade being a customer champion, it took a lot of time to accept, adapt and leverage such potential of AI. Now for me the Business Technology Platform is a sleeping bed - From starting my day to closing I observe and leverage different AI advancements in SAP BTP and help our customers win with SAP.
Happy to have a discussion on how does AI help you and your business.
Bluesky| User | Count |
|---|---|
| 25 | |
| 19 | |
| 17 | |
| 13 | |
| 11 | |
| 11 | |
| 10 | |
| 9 | |
| 6 | |
| 4 |
