Jon Miller, the cofounder of Marketo, recently posted a framework that’s been circulating across the MarTech world. After talking to over 150 mid-market and enterprise companies about their stacks, he drew a line: some categories of software get replaced by AI and vibe-coding, while others become more valuable. His shorthand is that enterprises will still subscribe to “Systems of Control” even as they DIY everything else.
We think he’s right, and the implications for contact data and data integrity are bigger than his post explored.
Here’s our read on where that line falls, what it means for B2B marketing operations and why the trust layer in your stack is about to matter more than ever.
Miller identifies three categories where AI and agentic coding deliver immediate displacement. We see the same pattern across our customer base.
Workflow UIs When a single engineer can spin up a functional interface in a day using AI-assisted coding tools, tolerance for clunky vendor screens drops to zero. The bespoke form builder you’re paying $40K a year for? That’s a weekend project now. And increasingly, AI agents don’t need interfaces at all. They interact with APIs directly.
Integration plumbing Custom connectors between your CRM, Slack, enrichment tools and routing logic? Agentic coding handles this well. Syncing fields, chaining API calls, building custom webhooks. These were never core product value. They were the tax you paid for using multiple vendors.
Commodity automation Basic ‘if this, then that’ logic. Simple lead scoring rules. Template-driven email sends. If the intellectual property in your tool is a series of if-statements, the moat just evaporated.
This is real. Companies are already pulling spend from tools that fall into these categories. But the wholesale ‘AI kills SaaS’ narrative misses where the value concentrates, as these layers get stripped away.
Miller’s framework identifies three categories that enterprises will still pay for. We’d go further: these categories don’t just survive the AI transition. They become the critical infrastructure that determines whether AI-driven operations create value or create chaos.
Not ‘store all data.’ Your warehouse handles storage. The value layer is what sits between raw inbound data and the systems your teams actually act on: identity resolution, deduplication, consent and permission management, suppression rules and data normalisation.
This is the trust layer. It determines which data is clean enough, compliant enough and accurate enough for your marketing, sales and revenue operations teams to use with confidence.
More AI-generated outbound means more data flowing into your systems. Without a trust layer, you’re scaling garbage at the speed of automation.
Nobody is going to vibe-code their consent management. Nobody is going to trust an AI agent to determine, unsupervised, which leads are real versus fraudulent, which contacts have valid permission to be marketed to, or which duplicate records should be merged versus kept separate. The stakes are too high: regulatory exposure, wasted pipeline and CRM entropy that compounds every quarter.
As AI accelerates the volume and velocity of lead generation, the trust layer becomes the bottleneck that determines whether that acceleration creates value or creates chaos.
This is where Miller’s analysis gets sharpest, and where we think there’s more to unpack.
It’s easy to build an AI agent that makes a decision. Route this lead here. Send this email. Score this contact. The technology for that exists today. What’s hard is building an AI agent that makes the right decision in context.
Why does your team route enterprise leads through a different qualification path than mid-market? Why are certain publisher sources suppressed for specific campaigns? Why did you change your scoring model last quarter, and what broke before you changed it?
That institutional knowledge lives in the heads of your MOps team, the tribal logic of your campaign structures and the post-mortems from last year’s launches. Without it, AI produces what Miller aptly calls ‘confident errors’: decisions that are technically correct but strategically wrong.
An AI agent can pick your highest-converting campaign template. It can’t know that your team learned three launches ago that this particular audience needs a customer story before the product pitch. That insight came from field experience, not from any field in your CRM.
The platforms that win in this environment aren’t the ones that automate decisions. They’re the ones that encode the rules, constraints and institutional logic that make automated decisions trustworthy. Systems of control, not systems of action.
AI can generate the content. It can draft the HTML, write the copy and personalise the message. What it cannot do is manage the infrastructure that gets that content delivered reliably at scale.
Miller uses email deliverability as his example, and it’s a good one. But the principle extends across every channel that touches contact data. For data processing, it means API reliability across dozens of publisher sources, error handling, retry logic and SLA guarantees on data delivery windows. For multi-channel orchestration, it means maintaining consistent identity across every touchpoint without losing compliance.
This is operational infrastructure. It doesn’t demo well. And it’s exactly the kind of thing that breaks catastrophically when you try to replace it with a weekend coding project.
Miller wrote from a MarTech buyer’s perspective, and his analysis is broad by design. But the Data Trust Layer is where all three of his ‘still pay for’ categories converge in a single system.
The Data Trust Layer is trusted data infrastructure: which contacts are real, compliant and actionable? It’s a decisioning system: where do they route, how do they score, what suppression and deduplication rules apply? And its execution infrastructure: how do you process thousands of records from dozens of sources with guaranteed SLAs while maintaining data integrity end to end?
Most of the MarTech stack sits cleanly in one category. Your CRM is a system of record. Your email platform is an execution infrastructure. Your CDP is a data layer. The Data Trust Layer spans all three, which is why it’s uniquely positioned as AI reshapes the stack around it.
As AI accelerates every stage of the demand generation funnel, the Data Trust Layer becomes the control plane. Not because it’s the flashiest part of the stack. Because it’s the part that determines whether acceleration creates pipeline or problems.
The smart move isn’t to rip out your stack. It’s to audit it against these categories and reallocate. Stop paying for workflow UIs you can build yourself. Double down on the systems of control that make AI-driven operations trustworthy instead of reckless.
That’s the bet we’re making at Convertr. Not competing with AI. Building the Data Trust Layer that makes AI-driven marketing operations safe, governed and trustworthy at scale.
Insights