What "AI-ready" actually means

The distance between the data an organization holds and data a model can use.

28documents on this topic
18organizations represented
3issues named
7sourced citations
0sourced statistics

The state of it

One of 5 topics within Data readiness.

Two thirds of enterprise data is not in a state a model can use. That is the finding, and it is unusually well evidenced: across 2,002 executives, only 33% of data is rated AI-ready and 53% of functional data is rated low quality. Data debt is scored high severity by 56% of them - level with technology debt and worse than process or talent.

What makes it a debt rather than a project is the stated cause. It is not one bad system. It is fragmented sources, legacy architecture, weak governance and ownership, poor quality management and tooling gaps, accumulated over years when patchy data was a tolerable inefficiency. What changed is not the data; it is that a model trained on it converges confidently on the wrong answer, and an agent acts on that answer faster than anyone can check it.

The trap in this band is the obvious response. "Fix the data" as an enterprise-wide programme is expensive, slow, and finishes long after the use cases that justified it were cancelled. The distinguishing move is sequencing - remediate along the workflows actually being rebuilt, and accept that most of the estate stays untouched for now.

The issues, by agreement

How many independent organizations name each issue as a problem. An issue is only as real as the number of separate publishers that identify it, so the count is the ranking. Bars are organizations, not documents. Where the count reads ours, no publisher here states the issue and the analysis is our own.

names it as a problemdisputes itqualifies it
Two thirds of the data is not in a state a model can use
2
Up to two fifths of the working week already goes on fixing the data
1
Remediating the estate, or remediating the paths the work runs on
ours

Who takes which position

The chart above counts positions; this shows whose they are. Read down a column for what one organization holds across the whole topic, and across a row for who lines up on one issue. Where a cell carries more than one position, the strongest is shown and the rest are in the tooltip.

Ddisputes it Qqualifies it Nnames it as a problem Pproposes a fix
What "AI-ready" actually means: 3 issues against the 3 organizations cited on them. The number under each name is how many of these issues it is cited on.
Issue Genpact · 3 Capgemini · 1 McKinsey & Company · 1
Two thirds of the data is not in a state a model can use N Q N
Up to two fifths of the working week already goes on fixing the data N · ·
Remediating the estate, or remediating the paths the work runs on Q · ·

A dot means this organization is not cited on that issue. It does not mean they are silent on it: an organization is cited where its document takes a position we could locate, and the absence of a citation is the absence of a finding, not a finding of absence. Who is represented lists everyone working on this topic, including those not cited above.

Where they disagree

No contradictions recorded on this topic yet.

The issues in full

Each issue carries the organizations that name it, the numbers behind it, and the remedies proposed - with the concrete steps under each. Every citation points at a section of a named document, so any count here can be checked.

Issue 012 organizations name it1 qualifies it2026 evidence

Two thirds of the data is not in a state a model can use

The organization has plenty of data and believes that is the same as being ready. A third of it is fit for AI; half of the functional data is rated low quality by the people who own it.

The figures are self-assessed, which if anything understates the problem - people grade their own estates generously. The value of the number is that it reframes the conversation from acquisition to remediation: nothing here is solved by collecting more. And the causes named are structural rather than technical - fragmentation, legacy architecture, weak ownership, absent quality management - which is why buying a tool does not move it.

How to fix it — 1 approach, 3 steps

Grade the data behind each live use case before funding more

Score the specific datasets each funded use case depends on, and publish the score alongside the business case rather than after delivery slips.

Done when Every funded use case names the datasets it depends on, each carries a grade in the business case itself, and each grade shows the date it was last taken.

  1. For each funded use case, name the datasets it depends on and grade each one.0-30 daysChief data officer
  2. Put the grade in the business case, so a weak dataset is visible before approval.30-90 daysCFO
  3. Re-grade on a fixed cycle, so the number moves rather than being taken once.ongoingChief data officer
The evidence — 4 documents
OrganizationDocumentPosition
GenpactConsultancy · June 2026$18 trillion in trapped AI valueOur reading Reports only 33% of data rated AI-ready and 53% of functional data rated low quality, with 56% scoring data debt as high severity, from a survey of 2,002 global enterprise executives.Data debt severity and qualitynames it
GenpactConsultancy · June 2026$18 trillion in trapped AI valueOur reading Attributes the state to fragmented source systems, legacy data architectures, weak governance and ownership, poor quality management and tooling gaps - accumulated when patchy data was merely an inefficiency.Why it is a debt rather than a projectnames it
McKinsey & CompanyConsultancy · June 2026Data readiness for scaling AI impactOur reading Names the stakes as the publisher frames them, tying the bottleneck to the small share of companies reporting AI fully scaled across the organization.What it costs downstreamnames it
CapgeminiConsultancyHow AI’s reshaping CXO decisionsOur reading Puts numbers on the gap between wanting and having: clear governance and accountability protocols are wanted roughly twice as often as they are in place, and systems to validate and explain output twice again.What executives say they are missingqualifies it

Issue 021 organization name it2026 evidence

Up to two fifths of the working week already goes on fixing the data

Data debt is discussed as a blocker to future AI. It is also a cost being paid now, in salaried hours spent reconciling, correcting and preparing data before anyone can use it.

In data-intensive functions this reaches up to 40% of employee time - not a projection, a description of the current week. It changes the business case materially, because remediation stops being an AI investment that pays back through some future model and becomes the recovery of capacity already being consumed. It also explains why the debt persists: the work is distributed across dozens of people doing it quietly inside their own jobs, so it never appears as a line item anybody is accountable for removing.

How to fix it — 1 approach, 2 steps

Count the hours going into data preparation before funding anything

Measure what reconciliation, correction and preparation currently cost in hours across the functions that depend on the data, and put that figure in the remediation case.

Done when A stated hours-per-week figure for data preparation and correction exists for at least three data-intensive teams, and the recovered-capacity number derived from it appears in the remediation business case.

  1. Ask three data-intensive teams to estimate weekly hours spent preparing and correcting data.0-30 daysCOO
  2. Put the recovered-capacity figure in the remediation business case alongside the AI benefit.30-90 daysCFO
The evidence — 2 documents
OrganizationDocumentPosition
GenpactConsultancy · June 2026$18 trillion in trapped AI valueOur reading Reports up to 40% of employee time in data-intensive functions going to reconciliation, correction and preparation, and ranks higher costs and wasted effort at 15% of cited business impacts, behind poor decision quality and unreliable insights at 18%.What data work consumes todaynames it
GenpactConsultancy · June 2026$18 trillion in trapped AI valueOur reading Frames remediation as reclaiming capacity already being spent rather than as new investment justified by a future return.The cost case for remediationproposes a fix

Issue 03Our analysis1 qualifies it2026 evidence

Remediating the estate, or remediating the paths the work runs on

Data debt invites an enterprise-wide cleanup: scoped across the estate, years long, delivering nothing until it is largely complete. Every source here that addresses sequencing recommends the opposite - remediation tied to the workflows being rebuilt.

This is the sequencing error, and it is the one that separates the organizations that get something from the ones that fund a data programme twice. The alternative is unglamorous: pick the workflows being rebuilt, remediate the data on those paths to the standard those workflows need, and leave the rest. It feels like under-solving. It is the only version that produces a working system inside a budget cycle, and it builds the pipeline patterns that make the next path cheaper. No organization in this index names the horizontal cleanup as the failure. The argument is ours, inferred from what these sources recommend instead, and the consensus count is zero for that reason.

How to fix it — 1 approach, 3 steps

Remediate along the workflows being rebuilt, not across the estate

Scope data remediation to the paths of the workflows already funded, to the standard those workflows need - and say plainly that the rest waits.

Done when A written remediation scope lists the workflows being rebuilt this year and the data on their paths, and names explicitly what is being left unremediated.

  1. List the workflows being rebuilt this year and the data each one touches.0-30 daysChief data officer
  2. Scope remediation to those paths only, and write down what is deliberately left.30-90 daysCOO
  3. Build each fix as a reusable pattern so the next path costs less than the last.90-180 daysHead of data engineering
The evidence — 1 document
OrganizationDocumentPosition
GenpactConsultancy · June 2026$18 trillion in trapped AI valueOur reading Frames the four debts as sharing root causes and compounding into one structural ceiling, so fixing any one of them alone is judged insufficient - the remedy it draws from that is a coordinated enterprise-wide mandate rather than a narrower one.Why the debt framing mattersqualifies it

Who is represented

This dossier is drawn from 18 organizations working on the subject, 3 of which are cited directly in the issues above.

Consultancy — 8

McKinsey & Company 5 Capgemini 3 Genpact 1 Deloitte 3 Accenture 1 EY 1 IDC 1 UST 1

Institution — 3

NIST 2 Cloud Security Alliance 1 IAB 1

Academic — 2

Carnegie Mellon SEI 1 MIT Technology Review 1

Hyperscaler — 4

IBM 2 Google Cloud 1 Lenovo 1 OpenText 1

Vendor — 1

IntuitionLabs 1