Why Your Data Is likely Your Biggest AI Constraint — And How to Find Out

Most AI implementation failures get blamed on the wrong thing. The tool wasn’t right. The vendor oversold it. The use case was too ambitious. Rarely does the post-mortem land where it should: on the data.

Data Readiness is the second pillar of the AI Maturity Index, and consistently the one that surprises organisations most when they see their score. It sits at a research baseline of 20% — exactly in line with the overall average — but it carries disproportionate weight. More AI initiatives stall or fail because of data problems than any other single cause. IDC found that 45% of AI projects in the Asia-Pacific region fail to hit ROI targets specifically due to poor data foundations.

The good news is that data readiness is diagnosable. Unlike strategy gaps (which require leadership behaviour change) or culture gaps (which take sustained effort), data problems are concrete and fixable — if you know what you’re actually looking at.

What does data readiness for AI actually mean?

Data readiness for AI means your customer and commercial data is clean enough, connected enough, and accessible enough for AI to actually use it — in real time, across the systems that matter to your revenue operation. It is not simply a question of how much data you have. Volume without quality, accessibility, or governance is not an asset. It is a liability that produces confident-sounding AI output built on a broken foundation.

The g2m AI Maturity Index measures data readiness across four dimensions: the quality and completeness of customer and prospect data; real-time accessibility of that data to AI tools; the capture and usability of unstructured commercial data; and data governance across the customer journey. Each dimension represents a distinct failure mode, and organisations typically have uneven scores across them — reasonable CRM hygiene but no governance framework, or accessible structured data but thousands of emails and call recordings sitting completely invisible to AI.

What is the difference between structured and unstructured data — and why does it matter for AI?

Structured data is the data living in your CRM, ERP, and financial systems — fields, records, rows, relationships. It is queryable, reportable, and if maintained well, directly usable by AI. Most organisations have some structured data, though quality and completeness vary considerably.

Unstructured data is everything else: emails, call recordings, meeting notes, proposals, support tickets, contracts, chat transcripts. IDC estimates that 80% of enterprise data is unstructured. It is also where most of the commercially valuable knowledge actually lives — what your customers said in the last three discovery calls, what objections came up in the deals you lost, what your best account managers know that your worst ones don’t.

For AI, the distinction matters enormously. A well-structured CRM tells AI what happened. Unstructured data tells AI why, and how. The highest maturity level on this question in the Index is: AI can search and retrieve our unstructured commercial knowledge by meaning — not just keywords. That capability — semantic retrieval across call recordings, proposals, and email histories — is where AI starts to compound commercial knowledge rather than just automate administrative tasks. Most B2B organisations are nowhere near it.

80% of enterprise data is unstructured — and it’s where most commercially valuable knowledge actually lives.

— IDC

Why does data quality kill AI initiatives?

AI models do not compensate for bad data — they amplify it. A model trained on or querying incomplete, inconsistent, or inaccurate data will produce output that sounds authoritative and is wrong. That combination — confident-sounding, wrong — is more damaging than no AI output at all, because it erodes trust in the tool, the team that recommended it, and the strategy behind it.

In B2B commercial contexts, the data quality problems that most commonly kill AI initiatives are:

• CRM records that are incomplete, duplicated, or inconsistently maintained across teams and regions

• Customer interaction data (calls, emails, meetings) that exists but is siloed and unsearchable

• Pipeline data that reflects what salespeople recorded rather than what actually happened in deals

• Multiple systems holding conflicting versions of the same customer record with no single source of truth

• No governance framework — so data quality degrades over time regardless of initial clean-up efforts

Each of these is a specific, diagnosable problem. None of them requires replacing your technology stack to fix. But all of them need to be identified before AI investment compounds on top of a broken foundation.

What does ‘not started’ look like on data readiness?

The lowest maturity level on the data quality question in the AI Maturity Index is: our customer and prospect data is fragmented, incomplete, or largely unreliable. Most organisations reading that description will recognise at least partial truth in it, even if they wouldn’t describe their situation that starkly.

In practice, ‘not started’ on data readiness looks like:

• Sales teams maintaining their own records outside the CRM because they don’t trust the central system

• No one able to answer ‘how many active customers do we have?’ without first qualifying which system they’re pulling from

• Call recordings sitting in a telephony platform that no one reviews and nothing can query

• Governance existing as a policy document that no one has read since it was written

• AI tools that have been purchased but can’t connect to core data systems without manual extraction or IT involvement

This is not a technology failure. It is a data culture and governance failure — and it cannot be solved by buying a better AI tool.

What does ‘embedded’ look like on data readiness?

The highest maturity level across the data readiness pillar represents a state that very few B2B organisations have reached — but which is entirely achievable with deliberate investment:

• Customer data is comprehensive, well-governed, and consistently maintained across all systems

• Marketing, sales, and customer data is fully accessible to AI tools in real time via integrated systems — no manual extraction, no IT intermediary

• AI can search and retrieve unstructured commercial knowledge by meaning — semantic retrieval across emails, calls, proposals, and support history

• Data governance is fully embedded: access controls, audit trails, and quality standards are actively maintained, not just documented

• Customer behaviour data drives real-time action — AI is identifying signals and triggering responses without waiting for a human to notice

The commercial consequence of reaching this state is not incremental. It is the difference between AI as an administrative efficiency tool and AI as a genuine revenue intelligence capability — one that compounds knowledge over time rather than resetting with every staff change.

How does data readiness connect to the other four pillars?

Data readiness is the constraint that most quietly undermines the other four pillars. You can have strong executive mandate (Pillar 1) but if the data feeding your AI strategy is unreliable, the strategy produces bad decisions at scale. You can invest in sophisticated technology architecture (Pillar 3) but if the data flowing through those tools is fragmented, the architecture amplifies the fragmentation. You can redesign processes around AI (Pillar 4) — but if the data inputs are wrong, the redesigned process produces wrong outputs faster.

The maturity question for data readiness is not ‘do we have data?’ Almost every B2B organisation has data. The question is: ‘is our data accessible, connected, and trustworthy enough that AI can actually use it — and do we have the governance to keep it that way?’

How does your organisation score on data readiness?

The AI Maturity Index measures data readiness across five questions: customer data quality, real-time accessibility to AI tools, capture and usability of unstructured data, governance framework, and real-time use of behavioural data. Your score on this pillar — and the pattern it creates alongside your other four pillars — is where the useful diagnostic starts.

For most organisations, data readiness is the gap that’s easiest to deny and most expensive to ignore. The index puts a number on it.