AI adoption is high, but integration is where the real work begins

There is no question that Enterprise AI has moved well beyond the experimentation phase. But where it has moved into the business is a harder question. 

In a recent report titled “The Reckoning: Aptean’s 2026 State of AI in Business Report”, Aptean notes that while 98% of organisations are using or implementing AI, only 46% say the technology is integrated into, or is essential to, core workflows and decision-making. The remaining majority are still in the wide middle ground between an interesting pilot and a dependable operational capability. 

This distinction is critical.

A company can deploy a chatbot, automate document summaries, or give employees an AI assistant and truthfully claim to be using AI. That does not mean AI is helping to plan production, improve forecast accuracy, detect inventory anomalies, or make better purchasing decisions.

The real work begins when the technology interfaces well with the systems that run the business.

The integration tax

Aptean’s research found that 82% of respondents consider integrating AI with core business systems more difficult than the AI technology itself—meanwhile, 81% identified data quality or access as their biggest obstacle to successful implementation.

Those figures are not surprising. 

Most organisations lack one clean, accessible store of operational data. They have ERP systems alongside CRM applications, spreadsheets, warehouse platforms, supplier portals, legacy databases and custom integrations built over many years. AI has a way of exposing those weaknesses quickly.

A general-purpose AI tool can produce a useful answer from a prompt. But it cannot reliably support a demand-planning decision without access to trusted sales history, inventory data, supplier lead times, and current production constraints.

The model is only part of the equation.

Why vertical AI has an advantage

Aptean’s study argues that industry-specific AI performs better because it is closer to the data, terminology, workflows and key performance indicators that define a particular business. Organisations using vertical AI outperformed those relying solely on general-purpose tools across seven of eight operational KPIs measured, according to the company. 

However, there is an important qualification. Vertical AI is not a shortcut around poor data or fragmented systems. It may lower the integration burden, but it cannot eliminate it.

That is why the current ERP market is moving towards AI embedded in operational workflows rather than isolated productivity tools. We recently explored how Acumatica 2026 R2 is attempting to make AI useful in everyday ERP, while Unit4’s Data Hub shows why access to governed ERP data is becoming a strategic concern. 

The unglamorous work

AI transformation is often presented as a choice of models, vendors and use cases. In practice, it is also about data cleanup, integration planning and defining processes well enough for automation to make sense.

None of that makes for a flashy product demonstration.

It is, however, what separates an AI pilot from an operational capability.