Why AI decision authority raises the stakes for CRM data

Marketing teams are giving AI decision authority while confidence in the CRM data behind those decisions remains limited. Validity’s “State of CRM Data Report 2026” found that nearly 91% of marketers consider data readiness critical for AI adoption, while only 21% consider their CRM data very well prepared for the AI tools they use or plan to use. Validity sells data-quality products and benefits commercially when companies invest in improving CRM data. Its findings still raise a clear management question: how much automated authority can the available data support?

AI authority is growing faster than data readiness

Validity found that 45% of respondents already use agentic AI that can act without human review, while two-thirds of organizations increased the number of marketing decisions delegated to autonomous agents over the past year. Agentic AI here means systems authorized to take actions rather than simply generate information for a person to consider. As that authority grows, errors in the information those systems use carry greater consequences.

The underlying CRM data remains uncertain. Only 26% of respondents said more than three-quarters of their CRM data is accurate and complete, and nearly half said their organization struggles with CRM data quality. Those figures show a gap between delegated authority and respondents’ confidence in the records supporting decisions. For executives, AI readiness therefore includes the reliability of data used to trigger action.

This creates a stricter operational standard. Human-reviewed workflows provide an opportunity to question an output before execution, although review does not guarantee an error will be caught. Autonomous workflows can execute before that review occurs. Data quality therefore becomes part of the control system for automated decisions.

Bad CRM data has a shorter path to action

An AI agent can use CRM records to make and execute a decision without prior human review. If an incorrect record affects a lead score, for example, that score can determine what happens to the lead before somebody examines the underlying record. The defect has moved closer to execution. That changes the control requirement.

A reporting error can still influence a poor human decision. Automated action adds speed and can remove a review point when the system is authorized to act independently. Executives therefore need controls over AI behavior and the business data supplied to it. The appropriate control depends on the consequence of the authorized action.

This also changes where data controls matter. An agent can act directly from CRM records, making validation, ownership, and monitoring part of decision governance. Validity’s survey shows rising use of autonomous AI alongside reported CRM-quality problems. Those findings establish concurrent conditions; they do not demonstrate that autonomous AI caused a specific financial loss.

The problem is surfacing in business decisions

Nearly 69% of respondents said a revenue, pipeline, or performance number they or their team presented had been challenged or walked back because the underlying data was wrong. Unreliable records are therefore reaching discussions where people assess business performance. When a figure has to be withdrawn, executives may also need to reconsider decisions based on it.

AI recommendations show a related pattern in respondents’ accounts. About 19% said they frequently presented or acted on an AI-generated recommendation they later suspected was wrong due to poor underlying data. Another 43% said this happened occasionally. These figures measure respondents’ suspicions about bad recommendations and their causes; they do not independently verify that the underlying data caused each error.

Seniority does not remove that exposure. Nearly 78% of C-suite executives and 92% of SVPs/VPs said they had acted on an AI recommendation they later suspected was wrong, compared with 41% of individual contributors. Among C-suite executives, SVPs/VPs, department heads, and directors, nearly 75% also reported that a revenue, pipeline, or performance number they or their team presented had been challenged or walked back because the underlying data was wrong. As reported by Validity, these figures show that questionable inputs can reach people with substantial decision authority.

Human oversight therefore depends partly on confidence in the inputs. A senior reviewer can challenge a recommendation that appears implausible, but the survey findings show that seniority itself does not prevent respondents from acting on recommendations they later question. Leaders need to know what information an automated decision uses and what controls apply to it. Higher-consequence actions justify a higher standard of confidence.

Organizations suspect a revenue cost they cannot confidently measure

Validity found that 62% of respondents believe poor CRM data probably or definitely costs their organizations revenue through issues such as missed renewals, inaccurate forecasts, lost deals, and misdirected campaigns. Yet only 28% are very confident that their CRM provides an accurate view of campaign performance and revenue impact. These are vendor-sponsored survey findings, and Validity has a commercial interest in stronger demand for CRM data quality. The figures capture a useful distinction between perceived economic risk and confidence in measuring it.

The first figure reflects respondents’ belief that poor CRM data costs revenue, but it does not quantify the amount lost. The second shows limited confidence in CRM-based measurement of campaign performance and revenue impact. Executives evaluating investment in data improvement should therefore separate evidence of perceived risk from estimates of its financial size.

That distinction matters for ROI decisions. Management may want a precise economic case for data-quality investment while having limited confidence in the system used to measure revenue impact. A defensible business case can acknowledge that constraint. Controls can be justified by the consequence and frequency of observed data problems without assigning them an unsupported revenue value.

AI readiness requires ownership

Only 41% of respondents said their organization has a dedicated data governance team or owner. Respondents also identified poor alignment among marketing, sales, and RevOps and a lack of clear internal ownership as barriers preventing CRM data from reliably supporting marketing. As AI gains authority, somebody must be accountable for the information used by automated workflows. That accountability covers data-quality rules, remediation, and exceptions.

Cross-functional collaboration is also limited by respondents’ own assessment. Only 39% of C-suite executives, SVPs/VPs, department heads, and directors said marketing and IT or RevOps collaborate very well to keep data usable for campaigns. Among senior managers and individual contributors, the figure falls to 27%. Both groups therefore see substantial room for stronger coordination around campaign data.

Automated decisions can cross several organizational boundaries. Marketing may define campaign intent, sales may depend on lead and account records, RevOps may own parts of the revenue process, and IT may control underlying systems. An AI workflow can consume information produced across those functions. Clear ownership establishes who defines quality rules, handles remediation, and decides whether data is fit for a given automated action.

Respondents also identified a preferred way to increase confidence. When asked what would most increase confidence in CRM data for strategy, reporting, and AI, 39% selected continuous automated monitoring that catches and fixes issues in real time. A unified platform ranked second, followed by third-party data validation or enrichment. Continuous monitoring still requires people to define acceptable data, decide when automated correction is appropriate, and assign responsibility for exceptions.

For the 59% of respondents without a reported dedicated data-governance team or owner, accountability is the immediate management issue. Greater autonomous authority makes it more important to know who should prevent, detect, and resolve problems in CRM inputs. Tooling can execute defined controls at scale. Leaders still have to set the rules and assign responsibility for them.

Main highlights

• Match AI authority to data readiness: AI decision authority is expanding faster than confidence in CRM data. Organizations assigning autonomous actions to AI need data controls that reflect the consequence of each action.
• Control CRM data before automated action: Agentic AI shortens the path from a flawed CRM record to a business decision. Data owners can reduce that exposure through validation, monitoring, and clear rules for high-consequence automated actions.
• Treat questionable AI decisions as an operating risk: Respondents across seniority levels reported acting on AI recommendations they later suspected were wrong. Decision owners need visibility into the data behind automated recommendations and the controls governing their use.
• Build ROI cases around measurable data risks: Most respondents believe poor CRM data costs revenue, while relatively few are highly confident in measuring revenue impact through their CRM. Investment cases should use observed data problems and their consequences without assigning unsupported revenue estimates.
• Assign clear ownership for AI data quality: Only 41% of respondents reported a dedicated data-governance team or owner. Marketing, sales, RevOps, and IT need named accountability for quality rules, continuous monitoring, remediation, and exceptions in data used by AI.