Why AI Keeps Stalling: The Legacy CRM Problem Nobody Wants to Fix

Organisations that have run an AI pilot in the last two years tend to tell a version of the same story; the model performed well, but the operational results fell short. Gartner projects more than 40% of agentic AI projects will be cancelled by the end of 2027 due to escalating costs, unclear business value, or inadequate risk controls.

These failures cluster around specific operational touchpoints:

A sales copilot recommending the wrong next action.
A customer service agent unable to view complete history.
A forecasting model relying on incomplete pipeline records.

At the outset, the team believes the solution is within reach. What follows is a confrontation with what actually exists inside the organisation: the CRM, the bolted-on point solutions and the technical debt that were already there. When the project moves toward production, the results that looked strong in isolation become inconsistent, because the model now has to draw on customer data from across the wider organisation rather than the small, self-contained set the team was testing against. That confrontation is what pushes otherwise good projects into permanent limbo, or into the cancellations Gartner’s figures describe.

Decisions about which model to use, which vendor to work with or which use case to prioritise all matter, but every one of them operates inside a boundary set by the state of the legacy CRM the AI is asked to reason across. Organisations fund this work because they are confident AI can solve a problem that spans the whole business.

This problem is a long-standing feature of legacy CRM. What has changed is that AI now reasons across the whole customer relationship at once, rather than one system at a time. This means that it has nowhere left to hide from a problem organisations have found ways to work around.

The foundation of reliable decision support remains a connected CRM, and that foundation still needs fixing before AI can be scaled responsibly.

A connected platform Microsoft Dynamics 365 platform removes much of this constraint at the source. When sales, service and marketing already share one system built on Microsoft Dataverse, the AI is reasoning over one consistent customer record rather than reconciling several. The fragmentation problem below is largely a description of what a legacy CRM makes unavoidable, and what a connected Microsoft platform makes unnecessary.

Fragmented data, fragmented results

AI systems, whether a predictive analytics tool, a large language model, or an agentic workflow that takes actions on someone’s behalf, can only be as good as the customer data they draw on.

When that data is spread across a legacy CRM and the systems bolted onto it, with no shared structure or definition of basic entities like customer, product or transaction, the model is left with two bad options:

working from an incomplete context, drawing conclusions based on whichever fragment of the CRM happens to be accessible to it
depending on an expensive, manual reconciliation process before it can function at all

Both of these defeat much of the purpose of automating the task in the first place and show up directly across the data pipeline:

Where it shows up	What it looks like
Training data	Reflects the structure of individual sales, service or marketing systems rather than the full customer journey, because that is how the CRM data was collected and stored in the first place
Inference pipelines	Break at the exact points where the CRM’s data has to be handed to another system
Model outputs	Accurate when checked against the CRM alone, meaningfully wrong once checked against the service history or marketing engagement held elsewhere

Where data fragmentation shows up specifically in customer-facing AI programmes.

The pattern in this table is a single problem showing up in three places: customer data that was captured, structured and stored without a shared standard. Fixing any one of them in isolation only moves the failure point further down the chain.

A connected CRM has to exist before AI investment can be scaled further.

A structural claim like this should hold up across more than one organisation’s experience. This is worth examining in detail.

The evidence across three independent studies

Boston Consulting Group’s 2024 research, based on a survey of 1,000 senior executives across 59 countries, found that only 26% of companies have developed the capabilities needed to move beyond proof of concept and begin extracting value from AI. The remaining 74% have yet to show tangible value from their AI investment. While enterprise adoption has accelerated rapidly since 2024, with companies increasingly turning toward advanced automation and agentic workflows, the foundational struggle to scale value remains a hurdle for many businesses today.

A second, independent source points the same way. McKinsey’s Global Survey on AI, published in November 2025 from responses gathered across 105 countries, found that 88% of organisations now report regular AI use in at least one business function, up from 78% a year earlier. Despite that growth in adoption, the majority remain in the experimenting or piloting stage, and only about a third say they have begun to scale AI programmes across the enterprise, meaning adoption has grown faster than the value organisations are getting from it.

Gartner’s forecast, cited earlier, addresses the underlying reason described above: agents built without a connected CRM foundation beneath them.

Taken together, these describe a structural condition rather than a temporary delivery issue.

“Enterprise CRM platforms evolve incrementally, with applications optimised for specific tasks rather than cross-functional consistency. AI models operate across functional boundaries by design. Without addressing the foundation, every additional use case adds isolated dependencies, increasing complexity faster than capability.”

This point is particularly relevant for Sales Directors, Customer Service leaders and CX leaders, because it reframes how AI underperformance is interpreted.

It is easy to conclude that a stalled pilot means the wrong model was chosen. More often, it means the CRM beneath it was never built to support one connected view of the customer.

The three studies above describe the size of the problem. Organisations that get past it tend to fix the CRM foundation first before scaling the AI.

Connect the CRM first. Then scale the AI

As AI begins to operate across sales, service and marketing, the requirements change. Systems must share consistent definitions of core entities such as customers, products and transactions. Data must be accessible across platforms without manual reconciliation. Governance must ensure that different parts of the organisation interpret customer information consistently.

Pilots do not need a fully connected CRM to prove initial value. Early deployments help validate use cases and build organisational confidence. The requirement for a genuinely connected CRM only becomes unavoidable once AI is asked to operate at scale, across sales, service and marketing, rather than inside a single controlled pilot.

Organisations that successfully move beyond pilots invest in three areas in parallel:

A connected CRM platform that structures customer, sales and service data consistently. In practice, this means bringing data from sales, service, marketing and other functions into a common structure, so the same customer, product or transaction is represented consistently wherever it appears, rather than existing as several slightly different versions across separate systems.
Integration capabilities that enable reliable data flow across systems. This means building pipelines that move data between systems automatically and consistently, so information created in one part of the business is trustworthy the moment it reaches another, without a manual reconciliation step standing between them.
Data governance frameworks that ensure consistent meaning across business entities. This means agreeing clear definitions and ownership for core concepts such as ‘customer’ or ‘transaction,’ and maintaining that agreement over time, so different teams are not quietly working from different definitions of the same term.

These capabilities work together to create a connected CRM environment in which AI systems can reason consistently across sales, service and marketing.

Where Hitachi Solutions fits in

Hitachi Solutions works with organisations across different sectors to replace the fragmented legacy CRM beneath a stalled AI programme with Dynamics 365 Sales and Customer Service, giving sales, service and marketing one connected, AI-ready customer record instead of a patchwork of point solutions.

Our artificial intelligence approach starts from a simple premise. The CRM foundation comes first, because no model, however capable, can outperform the data it is given to work with.

We support this through Microsoft Dataverse and Microsoft Fabric as the unified data foundation beneath the CRM, and broader Microsoft cloud and AI capability delivery through our Microsoft partnership, so the same connected customer record that powers sales and service also powers the AI built on top of it.

If your AI programme has stalled for reasons that sound familiar, contact our team at Hitachi Solutions.

Built to work as one.

Sources

Back to “Work As One”