“AI” is table stakes. Trust, identity, and orchestration are the moat. Martech Futurist | July 3, 2026

Most of the AI work inside marketing organizations right now generates activity that never reaches the bottom line. Two releases dated July 1, plus two deeper pieces from the past two weeks, explain why, and where leaders are pulling ahead. Gartner quantified how quickly the software interface is thinning. HBR diagnosed why most AI programs stall before they reach the bottom line. McKinsey mapped the capability architecture replacing the campaign model, and Forrester detailed what starts to break when agents run long enough to act on their own. The four converge on a single shift: the hard part of AI in marketing has moved from acquiring capability to rewiring the operating model that puts capability to work.

That shift settles the argument that has run since generative tools went mainstream. Adoption is effectively universal, and captured value is rare. McKinsey finds 90 percent of CMOs experimenting and fewer than one in ten capturing value across their workflows, with only 28 percent attempting a genuine rewiring of teams and processes. The distance between those two numbers is where the next few years of competitive separation will happen.

In the enterprise conversations I have across the year, the pattern holds. Leaders can list their pilots. Far fewer can describe the operating model that would let those pilots compound. This week's research gives that instinct a spine.

Software is becoming invisible

Gartner's projection that up to $234 billion of enterprise application spending sits exposed to "agentic arbitrage" by 2030 lands directly on the martech stack. When agents complete tasks across systems, the user interface stops being the thing a buyer pays for. Gartner describes the endpoint as software receding into the background, with the consequence that seat growth and revenue come apart for many vendors. For a CMO, the practical read is that a tool's value migrates from its screens to the outcomes it can execute and the customer context it can retain. This is the interface layer receding in real time, and it changes how stack decisions should be justified.

McKinsey's companion point is that the same recession runs on the demand side. More than half of consumers already lean on AI to guide purchases, which moves brands from an attention economy into what McKinsey calls a "trust economy," where the recommendation system determines which brands appear at all. Once a machine does the choosing, being findable stops being sufficient.

The bottleneck is the operating model

HBR's diagnosis is that leaders keep aiming AI at whatever feels most urgent, which produces motion without durable value. It cites an MIT finding that 95 percent of generative AI projects fail and an NBER survey of more than 6,000 executives in which roughly 90 percent reported no measurable productivity gain over three years. HBR names this the "urgency trap" and prescribes anchoring AI to a clear purpose and to long-term value rather than to the nearest visible pain point.

McKinsey's data explains the mechanism. Most organizations bolt AI onto existing workflows as point solutions, so savings never reach the bottom line and freed-up time goes unused. McKinsey states plainly that the campaign-era model has stopped working, and it defines the alternative as marketing redesigned into a continuous engine where insight, creativity, personalization, agentic commerce, and orchestration run in real time. That is an operating-model change that technology purchases alone cannot deliver.

Trust, identity, and orchestration are the moat

Forrester's read on agentic AI is that the capability has arrived and enterprise readiness has not. Once an agent runs for hours or days, it behaves like a distributed system, which demands orchestration, identity, and persistent context to stay reliable. Forrester describes a "trust tax": every autonomous action has to be logged and defensible to an auditor, and that cost keeps most enterprises stuck in pilots. Its prescription is governance enforced as code that runs while the agent runs, in place of policy that is written down and hoped for.

Set beside McKinsey's new operating roles, the pattern is clear. The capabilities that decide who wins sit in layers a company cannot buy ready-made: verification and trust, identity and permissions, memory and context, and the orchestration that connects them into a loop. Tools are commoditizing. The wiring between them is the differentiator.

Featured insights

Gartner: $234 Billion in Enterprise Application Software Spend Is at Risk from Agentic AI (July 1, 2026) Gartner projects that agents working across systems will expose roughly a fifth of enterprise SaaS spending by 2030 and will erode the interface as a point of differentiation. Practitioner takeaway: reassess martech renewals on outcomes executed and context retained, and ask each vendor how its agents earn their place once the dashboard stops being the product. Read on Gartner Newsroom

HBR: When Developing an AI Strategy, Beware the Urgency Trap (July 1, 2026) HBR argues that aiming AI at the most urgent visible problem produces activity without results, citing high project-failure and flat-productivity findings, and prescribes anchoring AI to purpose and to long-term value. Practitioner takeaway: before the next initiative, write down the durable outcome it serves; if the honest answer is speed on a visible pain point, expect it to stall. Read on HBR

McKinsey: From Campaigns to Continuous Growth: AI Capabilities Shaping Marketing (June 22, 2026) McKinsey defines five capabilities (continuous insights, scaled creativity, hyperpersonalization, marketing to AI agents, and always-on orchestration) and argues their value appears only when they run as one continuous system. It reports 90 percent of CMOs experimenting, fewer than 10 percent capturing value, and 28 percent rewiring workflows. Practitioner takeaway: choose one capability where AI shifts the economics quickly, instrument it to track value daily, and design it to connect to the others so it does not become a dead-end pilot. Read on McKinsey

Forrester: The State of Agentic AI, 2026: Companies Are Chasing, Few Are Catching (June 12, 2026) Forrester finds the technology ahead of enterprise readiness and traces the constraints to ROI uncertainty, governance gaps, and a "trust tax" on every autonomous action, recommending identity and policy enforced as code alongside the agent. Practitioner takeaway: treat governance, identity, and audit trails as design requirements for any agent you place in the funnel, and budget for the trust tax before you scale. Read on Forrester

Key takeaways

• Adoption has stopped being the differentiator. The operating model that turns capability into captured value has become the differentiator.

• Justify martech on outcomes executed and context retained, since the interface is losing its status as the thing worth paying for.

• Compete for machine selection. As recommendation systems choose, brands need machine-readable credibility alongside human-facing visibility.

• Build the layers a company cannot buy: verification and trust, identity and permissions, memory, and orchestration wired into a continuous loop.