Most CRM modernization projects are scoped as a platform decision and priced as one. The integration work becomes a line item somewhere near the bottom. Then the project finishes, the new CRM goes live, and the first AI initiative stalls six months later on the discovery that the systems were connected for people, not for data.
There are two different integrations hiding inside every CRM modernization, and conflating them is the most common and most expensive mistake in this work.
System integration keeps records in step so people can do their jobs. A customer created in CRM appears in Business Central. A won opportunity becomes a sales order. A salesperson opening an account sees whether the last invoice was paid. This is operational, it runs continuously, and it is what most people mean when they say the systems are connected.
Data integration lands CRM and ERP data together in an analytics layer where it can be joined, given history, and reasoned over. This is what makes forecasting, margin analysis, and AI agents possible. It is a different pattern with a different cost, and it does not come for free with the first one.
A company can have excellent system integration and still be entirely unprepared for AI. That is, in fact, the normal state of affairs.
Why This Matters More Than It Used To
Your CRM knows who the prospect is, what stage the deal is in, and when someone last called them. Business Central knows whether they pay on time, what their margin looks like, whether the last three orders shipped late, and what they still owe you.
For years, that split produced friction. A salesperson gets caught out by a credit hold. Marketing spends to acquire customers that operations already knew were unprofitable. Annoying, survivable, absorbed.
AI removes the tolerance. Ask an agent to prioritize accounts and it will answer from whichever half of the picture it can reach, with full confidence and no indication that half the picture was missing. Ask it to flag renewal risk and it will use engagement data because payment behavior was never made available to it. The output looks the same either way. That is the problem.
This is why CRM modernization has become a data decision rather than an application decision, and why it now belongs on the leadership agenda instead of being delegated as an IT project.
The Four Decisions
1. Which system owns the customer
Your CRM believes it owns the customer. So does Business Central. Both are partly right, and the ambiguity is where integration projects quietly fail.
The workable answer is a declared boundary. CRM owns the relationship through the point of sale, Business Central owns the customer from that point forward, and specific information flows in one agreed direction. When two systems both believe they own the billing address, something eventually ships to the wrong place, and the argument afterward is about the integration rather than the decision nobody made.
This also determines what AI can be trusted with. An agent cannot resolve conflicting definitions of a customer at runtime. It picks one and proceeds. Settle the boundary before selecting a platform, because it sets the cost of everything after it.
2. Where the CRM lands, judged on integration cost
Three answers are defensible, and the right one depends on where your commercial leverage sits.
Consolidate on The argument is not features, it is fewer moving parts. Dynamics 365 shares a data platform with Business Central and with the Microsoft analytics layer, which means both integrations, operational and analytical, are shorter. One security model, one vendor, and the fastest route from CRM data to AI.
Keep and integrate it properly. If marketing is your engine, forcing a migration often trades a system your team uses for one they resent, and that cost rarely appears in the business case. The honest trade-off is that Microsoft provides no built-in connection to Business Central, so you will pay for an integration platform or a custom build. You will then pay separately to get HubSpot data into the analytics layer. Both are routine pieces of work. Both belong in the budget from the start.
Run both, with an enforced boundary. Marketing in one, sales and service in the other. This works when the boundary is genuinely policed and fails when it is the residue of a migration someone stopped halfway through. It also doubles the data integration work, which is the part that gets forgotten.
The comparison people run is license cost. The comparison that predicts the actual bill is total integration cost, operational and analytical, over three years.
3. What you are willing to spend to connect the systems
Operational integration comes in three cost profiles, and leadership should choose the profile rather than delegating it.
The built-in Microsoft path is included in what you already license and covers common scenarios well. An integration platform costs more and buys flexibility plus a support relationship. A custom build gives you exactly what you want and hands you permanent ownership of maintaining it, including every time either vendor ships an update.
The question to ask is not which option is technically best. It is this: if the person who built this leaves, what happens? That sorts the three faster than any feature comparison, and it is the same question that should be asked of every AI workflow built on top of them.
4. Whether the data ever reaches your AI foundation
This is the decision CRM projects skip, and the one that determines whether the investment enables AI or blocks it for another three years.
A CRM that serves salespeople well and cannot feed the analytics layer is a completed project and a stalled strategy. Landing CRM and Business Central data together, with history, with consistent definitions, and with monitored refresh, is what turns two operational systems into something an AI agent can reason across.
If you land on the Microsoft platform, this connection is short and largely configuration. If you stay on HubSpot, it is real engineering work: routine, entirely feasible, and something that belongs in the platform comparison rather than surfacing as a surprise during your first AI initiative.
Ask for it to be costed during selection. It changes the answer more often than people expect.
What AI Requires From the Integration Layer
AI does not tolerate the ambiguity that reporting has absorbed for years. A dashboard with an inconsistent customer definition produces a number someone eventually challenges in a meeting. An AI agent with the same inconsistency produces a confident recommendation that nobody challenges at all.
Four things have to be true:
- One definition of a customer, encoded once. If "active customer" means something different in CRM, ERP, and marketing, an agent will produce three defensible answers to the same question.
- Operational context available to sales-facing AI. A CRM assistant that cannot see credit status or open orders will enthusiastically recommend pursuing an account that finance has already put on hold.
- History, not just current state. Forecasting, churn signals, and renewal risk all depend on how things changed over time. Today's snapshot supports none of them.
- Governance that survives the move. Once CRM and ERP data land together in a shared analytics environment, who and what is permitted to read it becomes a decision to make rather than a setting to inherit.
Questions Worth Asking
What decision does our sales team get wrong today because of what the CRM cannot see? Are our systems connected for people, for data, or only for people? What would it cost to get CRM and Business Central data into one place with history, and was that number in the CRM business case? If we deployed an AI agent against this next quarter, what would it get wrong?
The Common Mistake
The most expensive pattern is migrating the existing setup into a new platform on the understanding that cleanup happens later, while treating integration as a technical detail to be sorted during implementation. Both choices defer the work that determines the outcome. The abandoned fields and competing definitions arrive intact at a higher license cost, the systems get connected for people because that is what was scoped, and the AI conversation starts from zero eighteen months later.
The better sequence is narrow and fast. Settle the ownership boundary. Agree the two or three commercial decisions your teams need better information to make. Connect the systems operationally for those decisions. Land the data properly so it can be reasoned over. Prove the value, then expand.
The platform decision is the easy part. DStrategyTech is a Michigan-based Microsoft Partner helping SMBs turn Dynamics 365 CRM and Business Central into a foundation for analytics and AI. Our CRM modernization assessment is a short, fixed-scope engagement that evaluates your CRM and ERP landscape, defines the data ownership model, identifies the right system and data integration patterns, and produces a costed roadmap.
to discuss your CRM modernization, Business Central, and Data & AI needs.