AI has moved from marketing experiment to everyday operating tool, but most teams still cannot show what it is worth. A new Basis survey found that 61.4% of marketing and advertising professionals use generative or agentic AI every day, nearly twice the 33.6% reported a year earlier. Yet just 6.5% of organizations assess AI investments against defined metrics and benchmarks. For business leaders, that gap is the real headline: adoption is scaling faster than accountability.

The fourth annual Basis “AI and the Future of Marketing” report surveyed 184 professionals across agencies, brands, and publishers in August 2026. Its findings are a timely signal, not a census of the entire industry: the public report does not explain respondent recruitment or provide enough methodology to establish representativeness.

AI is now part of the marketing operating model

In the Basis survey, 98.4% of respondents said they use generative or agentic AI in their work. Three-quarters said AI had made them moderately to significantly more efficient over the past year. This is no longer a question of whether a marketing team should try AI; leaders are deciding which workflows to redesign and how much responsibility to delegate.

Agent adoption underscores that shift. The report says 51.1% of organizations now use AI agents, a 155.5% year-over-year increase. Among organizations using agents, common tasks include reporting and analytics (72.3%), drafting campaign briefs (56.4%), creating content (45.7%), and media planning (35.1%). These are consequential workflows, not just experiments in drafting copy.

The ROI gap is a management problem

Only 6.5% of organizations said they evaluate AI investments against defined metrics and benchmarks. Basis found that 65.1% either do not measure AI return at all or do not know whether anyone in their organization does. EMARKETER’s breakdown of the same report puts 46.2% in the “not measuring” group, 23.9% unsure, and about one-quarter measuring ROI anecdotally. That distinction matters: the result is not that only 6.5% ever look at AI outcomes, but that very few use a defined, benchmarked method.

Without a baseline, teams can mistake faster output for better marketing. An agent may produce more campaign variants while increasing review time, waste, or low-quality leads. A content assistant may shorten drafting while creating editing, compliance, or brand-consistency costs. If managers do not track the whole workflow, they cannot tell whether AI improved the business result or merely moved work to another step.

That is why Basis chief marketing and commercial officer Katie McAdams said, “Marketers have moved well past the question of whether or not AI belongs in their workflow. Yet, more work is needed to justify the investment.” The report’s findings make that next stage measurable: connect each deployment to an outcome, not to a usage count.

Measure one workflow before expanding agents

Leaders should resist turning an adoption statistic into a mandate to automate everything. Start with a repeated, bounded process where an AI tool can be compared with the existing workflow. Choose a metric that reflects the business objective, such as cost per qualified lead, campaign production time at an agreed quality level, conversion rate, or revenue per dollar spent. Track the human review and correction required, not just the AI’s raw output.

Where possible, use a control group, holdout, or phased rollout so the team can separate AI’s effect from seasonality, budget changes, and audience differences. Record the baseline before switching tools. Include software, integration, training, oversight, and rework in the cost calculation. If reliable attribution is not available, label the result directional rather than presenting an efficiency estimate as proven financial return.

A practical 30-day AI ROI sprint

  • Week one: Select one workflow and write down its current cost, cycle time, quality standard, and business outcome.
  • Weeks two and three: Run the AI-assisted process alongside the baseline or in a controlled pilot. Log human edits, exceptions, errors, and total labor.
  • Week four: Compare outcomes and full costs. Expand only if the result clears a pre-set threshold and the team can explain the limits of the measurement.

Set a named owner for the metric and a review date before launch. For agentic systems, add approval boundaries for budget changes, customer-facing claims, and sensitive data. This turns experimentation into a repeatable investment decision rather than a collection of disconnected tool trials.

The Basis survey suggests marketers have crossed the adoption threshold; the harder work is proving where AI creates durable value. Real Internet Sales helps organizations build practical AI marketing, measurement, and search strategies that connect technology to business outcomes. Call 803-708-5514 or visit realinternetsales.com.