Agentic capacity is building faster than governance. Martech Futurist | July 23, 2026
Four data points landed this week, and they describe the same constraint from four angles.
Gartner put 2026 worldwide end-user spending on AI models and platforms at $64 billion, a 63.4% increase over 2025. HCLTech surveyed 500 enterprise decision-makers and found 90% reporting that generative and agentic AI have changed their workflows, while 18% report significant revenue impact. Usercentrics published US findings showing 78% of American consumers would stop using a service after their data was misused, and 66% already have. McKinsey's AI trust research names knowledge and training gaps as the leading barrier to responsible AI practice, cited by roughly 60% of respondents.
Read together, these locate the binding constraint in the same place. Capability is arriving on a purchase order. The capacity to decide what an agent may do, verify what it did, and answer for the result arrives through hiring, training, and organizational design, and it arrives much more slowly. The last three editions worked through the control mechanisms themselves: proving what an agent did, bounding what it was permitted to do, and treating agent identity as the underlying enforcement point. Those mechanisms run on people. This week's research says the supply of qualified people is where the ceiling actually sits.
Spending is shifting toward control surfaces
Gartner's July 20 forecast shows generative AI model spending growing 117% this year and platform spending growing 36.9%. The composition matters more than the total. Gartner's Arunasree Cheparthi describes enterprise AI budgets coming under greater scrutiny, with buyers concentrating on usage efficiency, cost control, and measurable outcomes. The advantage moves to providers that build evaluation, cost transparency, and usage tracking into the workflow, and buyers are selecting platforms that help them monitor performance and enforce policy.
That is a market repricing the Trust and Identity layers of the AI capability stack. Two years ago, procurement compared model quality. This year, procurement compares the instrumentation around the model. Domain-specific language models are forecast to grow 210% in 2026, which points the same direction: narrower models operating inside tighter bounds, chosen because the bounds are legible.
Marketing organizations should read this as permission to change their evaluation criteria. The question a martech vendor should be able to answer in a demo is what the system records about its own actions and who can revoke its access mid-run.
The enterprises scaling agents are the ones investing in people
HCLTech's report, released July 21, separates AI Leaders from AI Followers and quantifies the distance. Leaders are four times more likely to scale agentic and autonomous AI. They define measurable use cases at a rate of 73% against 22% for followers. They secure senior leadership sponsorship at 63% against 36%. The largest single gap sits in workforce transformation: 93% of Leaders run structured upskilling programs, against 20% of Followers.
That last number does most of the explanatory work. The organizations moving furthest into autonomous execution are the ones that spent the most on human capability first. Followers evaluate AI through efficiency and cost, which produces incremental gains and stalls before enterprise scale.
Prachi Gore, Chief Marketing Officer at Asana, described the operating consequence when I interviewed her on The Agile Brand podcast: "I think the biggest change in mindset is we have to think agents-first in everything… The second mindset shift is, at every level in the company, you've all become managers now."
Management responsibility distributed to every level requires management capability at every level. A marketing coordinator who now supervises three agents needs to know what those agents are permitted to touch, what a failure looks like, and when to stop the run. That is a trained skill. The 20% figure describes how many organizations are treating it as one.
This is also where the descriptive and normative split from my second book does real work. Agents produce analysis. Deciding what the organization should do with that analysis stays with a human, and that human has to be equipped to make the call.
Customers have attached a price to the judgment
Usercentrics released US findings from its second annual State of Digital Trust study this week, drawn from Sapio Research fieldwork across 11,000 consumers in seven markets. In the US, 78% say they would stop using a service after their data was misused. 66% already have stopped using a company over privacy concerns. Only 39% of American consumers trust government services with their data, the lowest figure in the study, which leaves brands holding the trust relationship without an institutional floor beneath them.
Globally, 47% of respondents took at least one action with direct revenue consequence in the past six months because of concerns about how their data was used in AI. Canceling, switching, reducing spend. Usercentrics ties the shift partly to agentic AI reaching financial accounts, calendars, and customer records on users' behalf, which makes data handling something consumers experience directly.
McKinsey's research shows the internal counterpart. Across nearly every risk category, active mitigation lags behind stated risk awareness, and the gap is widest for personal privacy and intellectual property. Organizations know where the exposure sits. Fewer have built the controls. McKinsey also found that organizations with explicit accountability for responsible AI score materially higher on maturity, 2.6 against 1.8 for organizations with no clearly accountable function.
Kip Havel, Chief Marketing Officer at Dexian, put the labor market read on this when I interviewed him: "The companies building the frontier are hiring for judgment and point of view. That tells you something about where the scarcity is."
Respecting customers and their data has been the fourth Agile Brand principle since I wrote it. The Usercentrics numbers convert it from a stated value into a line item.
Featured Insights
BigDATAwire (Gartner release) | Gartner Forecasts Worldwide AI Platforms and Models Market to Grow 63% in 2026 | July 20, 2026
Gartner projects $64 billion in worldwide end-user spending on AI models and platforms in 2026, up 63.4% from $39 billion in 2025, with generative AI model spending growing 117% and domain-specific language models growing 210%. Senior Principal Research Analyst Arunasree Cheparthi describes buyers concentrating on usage efficiency, cost control, and measurable outcomes, and selecting platforms that help them choose tools, monitor performance, enforce policy, and contain cost.
My takeaway: Add policy enforcement, action logging, and permission revocation to your martech evaluation criteria this cycle. The market has already started pricing them.
HCLTech | The Blueprint for AI Leadership | July 21, 2026
A global study of 500 enterprise decision-makers conducted with Raconteur found 90% reporting workflow transformation from generative and agentic AI, 91% citing improved data access, and 18% reporting significant revenue impact. AI Leaders are four times more likely to scale agentic and autonomous AI, and separate from Followers most sharply on structured upskilling programs, 93% against 20%.
Practitioner takeaway: Benchmark your own upskilling coverage before your next agent deployment. If the number is closer to 20% than 93%, the deployment will stall at the same place everyone else's does.
Usercentrics | July 21, 2026
Sapio Research surveyed 11,000 consumers across seven markets. In the US, 78% would stop using a service after data misuse and 66% already have. Globally, 47% took an action with direct revenue consequence in the past six months over AI data concerns, and 52% now trust AI less than humans with their personal data, up from 48% in 2025.
My takeaway: Map every agent in your stack to the customer data it can reach and the consent basis that authorizes that reach. Where the chain breaks, you are carrying churn risk you have not priced.
McKinsey | State of AI trust in 2026: Shifting to the agentic era | March 25, 2026
The 2026 AI Trust Maturity Survey covers approximately 500 organizations. Roughly one third reach maturity level three or higher in strategy, governance, and agentic AI controls. Nearly two thirds cite security and risk concerns as the top barrier to scaling agentic AI. Around 60% name knowledge and training gaps as the leading barrier to responsible AI implementation, up from about 50% the prior year. Organizations with explicit accountability for responsible AI average 2.6 on the maturity scale against 1.8 for those without.
My takeaway: Name a single accountable owner for agent governance inside marketing. The maturity difference between having one and not is larger than most tooling decisions you will make this year.
Key Takeaways
Budget is moving toward instrumentation. Gartner's 63.4% spending growth is concentrating around evaluation, usage tracking, and policy enforcement. Bring those requirements into your vendor scorecard now.
Upskilling coverage predicts agentic scale. The 93% against 20% gap in HCLTech's data is the sharpest divide between organizations scaling autonomous AI and organizations stuck in efficiency pilots. Treat training budget as agent infrastructure.
Assign the owner before the agent. McKinsey's maturity difference between organizations with explicit responsible AI accountability and those without is 2.6 to 1.8. Marketing needs a named person who decides what agents may touch.
Consent is a revenue variable. With 78% of US consumers prepared to leave over data misuse and 47% globally having already acted, the permission an agent inherits carries a measurable price when it is exceeded.
Judgment is the scarce input. Capability arrives on contract. The ability to bound it arrives through hiring and training. Plan the second on a longer lead time than the first.