Every CIO of every mid-market or enterprise org I talk to is having the same private conversation: "should we finally replace SAP?" or "is it time to migrate off Dynamics?" The answer, in 2026, is almost always: no, but you do need to modernize around it.

SAP and Microsoft Dynamics are not the bottleneck. They're not what's slowing your AI roadmap. The bottleneck is that your business data is locked inside them, your workflows route around them, and your AI tools can't see them. Fix that and the systems-of-record do exactly what they were designed to do — record. Everything else can happen in a modern layer.

The thesis — Treat SAP and Dynamics as systems of record. Build the systems of intelligence (analytics, AI, agents) and systems of engagement (web, mobile, APIs) around them. The replacement project costs $30M and 4 years. The integration project costs $1-3M and 6-12 months. The economics are not close.

Why the "rip and replace" instinct is wrong

The replacement pitch is appealing on the slide. Modern UI. Cloud-native. AI-ready. Cheaper TCO over 10 years. The reality, in our experience modernizing ERP/CRM environments across the US, GCC, UK and EU:

  • Migration costs always blow out. The "$30M, 4-year" number above is a real client average. The slide-projection is usually $15M, 2 years. Every veteran CIO has lived this delta.
  • The new system rarely matches the configured edge cases of the old one. 15 years of business logic is encoded in custom ABAP, in Power Platform flows, in Field Service rules. Reconstituting those is a multi-year project of its own.
  • Adoption tanks during cutover. Sales teams rebel. Finance escalates. The project becomes a political target.
  • The replacement vendor's roadmap matters more than yours. You become a price-taker on their release cadence.

Compare to the modernization path: keep the system of record, integrate it cleanly, and build the AI + analytics + UX layer on top. Most of what business users complain about (slow dashboards, clunky interfaces, no AI assistance) lives in that top layer — not in S/4HANA or Dynamics itself.

The four-layer modernization architecture

Layer 1: System of Record — keep it

SAP S/4HANA, ECC, Ariba, SuccessFactors, BTP. Microsoft Dynamics 365 Sales, Customer Service, Marketing, Field Service, Finance. These are the source of truth for transactional data. Don't move them. Upgrade them on their own cadence. Run them lean.

Layer 2: Integration Layer — modernize this

This is where most legacy environments are weakest. CSV exports, nightly batch jobs, hand-coded ETL, brittle SOAP integrations. Replace this layer with: event-driven architecture (Kafka, Azure Event Hubs, AWS EventBridge), REST/GraphQL APIs published from S/4 via SAP BTP or from Dynamics via Power Platform, and a managed integration platform (MuleSoft, Boomi, Azure Logic Apps).

The output: every business event in S/4 or Dynamics is published as a stream within seconds. Downstream systems consume it without the original system noticing.

Layer 3: Data & AI Layer — build this

A modern lakehouse on Databricks, Snowflake, or Microsoft Fabric. Ingestion from layer 2. Mastered, governed, queryable. AI/ML training data. RAG corpora for agentic AI. Real-time analytics. This is where AI lives. Not inside S/4HANA, not inside Dynamics — alongside them, fed by them.

Layer 4: Engagement Layer — own this

Custom web/mobile interfaces for customers, partners, and internal users. Built with React, Flutter, or Power Apps depending on use case. AI copilots inside these interfaces (Microsoft Copilot Studio, custom agents). The engagement layer is where business users actually spend their day. It's the layer that determines whether your tech stack feels modern.

What this looks like for SAP environments

For an SAP-heavy environment, the modernization stack we deploy looks like:

SAP S/4HANA + Ariba + SuccessFactors (system of record · keep)
    ↓ via BTP / CDC / IDoc events
Kafka / Azure Event Hubs (integration backbone · build)
    ↓
Snowflake / Databricks lakehouse (data + AI layer · build)
    ↓
SAP Datasphere (governed semantic layer · keep)
    ↓
React/Power Apps custom UIs + GenAI agents (engagement · own)

SAP Datasphere is worth highlighting. It bridges SAP's semantic model with non-SAP data sources, which makes it the natural place to govern the boundary. Use it.

What this looks like for Dynamics environments

For Dynamics 365 environments, the stack is similar but the names change:

Dynamics 365 (Sales · Service · Marketing · F&O) (system of record · keep)
    ↓ via Dataverse + Power Platform
Azure Event Grid / Service Bus (integration · build)
    ↓
Microsoft Fabric / Synapse (data + AI · build)
    ↓
Power BI semantic models (governed analytics · keep)
    ↓
Power Apps + custom React/Next + Copilot Studio (engagement · own)

The Dynamics path benefits from Microsoft's tight integration story — Dataverse is genuinely the easiest path from a CRM transactional system into a data lake, and Copilot Studio is the most pragmatic way to deploy AI agents on top of Dynamics data.

Where AI actually fits

The modernized stack unlocks four classes of AI use cases that are hard or impossible inside the system of record itself:

1. AI-native analytics

Natural-language query over your business data. "Show me revenue by region by quarter, broken down by product line, vs same period last year." With the lakehouse layer fed in real-time from S/4 or Dynamics, this works. Without it, you're waiting for someone in finance to build a Power BI dashboard.

2. Agentic customer service

A service agent that can answer customer questions about their account, their orders, their invoices — by reading from the system of record in real time, with appropriate access controls. The system of record stays clean. The agent reads from a sanctioned API.

3. Copilots inside business workflows

"Draft this customer renewal email based on their account status." "Summarize this prospect's last 6 months of engagement." These work when AI has read access to the relevant data — and that requires the integration + data layer to exist.

4. Predictive analytics on transactional patterns

Churn prediction, demand forecasting, dynamic pricing. These require historical transactional data in a queryable, governed form. SAP and Dynamics generate the data; the lakehouse turns it into a feature store.

The ROI conversation

For a typical mid-market SAP/Dynamics modernization in our portfolio:

  • Year-1 investment: $1.2M-$2.8M (integration platform, lakehouse build, first 3 AI use cases).
  • Year-1 returns: $1.5M-$4M (productivity gains in finance/sales/ops, reduction in manual data work, faster reporting cycles, first AI-driven revenue lifts).
  • Year-3 cumulative ROI: Typically 3-5x the investment.

The "rip and replace" alternative for the same scope: $25M-$45M over 3-4 years, with disruption costs that are difficult to quantify but real. The modernization path wins on every dimension that matters to a CIO making this call.

The five integration projects to start with

If you're at the beginning of this and want concrete first moves, here's the order we recommend:

  1. Stand up the event integration layer. Pick Kafka or Azure Event Hubs. Define the event schema for the 5-10 most important business events. Get one event flowing end-to-end.
  2. Stand up the lakehouse. Snowflake, Databricks, or Microsoft Fabric. Ingest the events. Build the first three governed datasets (typically: customers, orders, invoices).
  3. Pick one AI use case. Service agent, NL analytics copilot, or predictive churn. Whichever delivers the cleanest ROI story. Ship it.
  4. Build one engagement-layer surface. Customer portal, partner portal, or internal ops dashboard. Whichever business team complains loudest. Ship it.
  5. Then scale. Now that the four layers exist, every new use case is incremental, not net-new architecture.

What to demand from your integration partner

If you're picking a partner for this work, the questions to ask:

  • "Show me an event-driven SAP integration you've shipped to production. Walk me through the data lineage."
  • "How do you handle master data harmonization across S/4 and a non-SAP source?"
  • "What's your stance on SAP BTP vs an external integration platform? When do you use each?"
  • "How do you secure a Dynamics 365 → Fabric pipeline so the AI layer can't accidentally exfiltrate PII?"
  • "Show me a Dynamics + Copilot Studio deployment that's live, with metrics."

Generic systems-integrator answers won't survive these questions. You want a partner who has done this — not one who is about to learn on your dime.

The wider point

SAP and Microsoft Dynamics are still the right systems of record for most mid-market and enterprise environments. The question isn't whether to keep them. The question is whether the layers around them are letting your business move at the speed it needs to. In 2026, "modernize, don't replace" is the durable answer for almost everyone.

If you're scoping an SAP integration, a Dynamics modernization, or a hybrid (we have plenty of clients running both), book 30 minutes. We'll show you what your modernized architecture could look like specifically for your business — with the actual integration patterns and AI use cases that earn their keep.