The adoption number sounds decisive. A new industry report, Martech for 2026, written by Scott Brinker and Frans Riemersma and sponsored by GrowthLoop, Hightouch, Intuit Mailchimp, MetaRouter, Progress, SAS, and Treasure Data, found that 90.3 percent of marketing organizations now use AI agents somewhere in their operation. The number that gets buried further down the same report is smaller: only 23.3 percent have an agent running in full production. The rest are piloting, testing, or running agents in an assist-only mode where a human checks the output before anything ships.
The budget line tells a less optimistic story. Gartner’s 2026 CMO Spend Survey, based on responses from 401 CMOs, found that martech now accounts for 19.4 percent of the marketing budget, a five-year low and down sharply from 26.6 percent in 2021. Yet 62 percent of those same CMOs said they planned to invest more in marketing technology this year. Both things are true at once because the money is moving: into consumption-based pricing, into general AI initiatives that do not get coded as martech, and into the headcount needed to actually run these tools, not into flat software license renewals.
The expectations gap shows up outside the vendor surveys too. The Australian Centre for AI in Marketing cited a 2026 Adobe survey of 3,000 CX executives and practitioners in which a majority reported no active use of agentic AI at all, and fewer than a quarter were running even a limited pilot. Sixty-three percent still expected agentic AI to free up time for strategic work eventually. Marketers are not rejecting the technology. They are further behind the hype cycle than the adoption headlines suggest.
The report’s own authors point to the actual bottleneck. Among marketers who have tried to put an agent into production, 56.3 percent named poor data quality as a critical obstacle, ahead of budget or staffing concerns. An agent that drafts a campaign brief or qualifies a lead is only as good as the account and contact records it reads from. Stale job titles, dead inboxes, and duplicate CRM records do not just waste a rep’s time anymore; they feed directly into whatever the agent decides to do next. EmailAddress.ai, a related company of ours, works specifically on that layer, verifying catch-all and risky email addresses and keeping contact records current for CRM and outbound use, which is the kind of plumbing this problem actually requires.
None of this means the agent rollout stalls out. It means the next budget cycle probably should not be measured by how many new agents got switched on, but by whether the data feeding them got any cleaner. The 2026 numbers suggest most organizations bought the tool before they fixed the records, and are now finding out in production why that order matters.
Sources referenced in reporting: