Jon Miller

The $285B SaaS selloff has everyone asking whether AI will kill software. I think that's the wrong question.

The better question: which parts of SaaS get replaced, and which become more valuable? And what’s that mean for #martech?

I've spent the last 18 months talking to 150+ mid-market and enterprise companies about their stacks. They're not preparing to rip everything out. But they are rethinking what to buy versus build.

Here's where I think the lines are forming.

WHAT GETS DIY'd
✅ Workflow UI. When one person can spin up a decent interface in days with Codex or Claude Code, they stop tolerating clunky vendor UX. Chat-based interfaces will replace purpose-built screens for more and more marketing workflows. AI agents will use software, not humans.

✅ Integration glue. Custom connectors between your tools (syncing CRM data to Slack, routing leads between platforms, chaining enrichment steps together) are perfect for agentic coding.

✅ Commodity features. Basic “if this then that” logic is wrapper functionality any competent engineer can vibe code in a weekend.

WHAT ENTERPRISES WILL STILL PAY FOR
I predict enterprises will still subscribe to systems of record, and more importantly, Systems of Control.

✅ Marketing-grade trusted data. Not "store all data", the warehouse will handle that. But the layer that determines what's trusted enough for marketing to act on: identity resolution, permission, suppression rules, deduplication. Nobody's going to vibe-code their consent management.

✅ Governed decisioning. This is essential. It's easy to build an AI agent that decides what email to send. It's very hard to build one that knows why your team sends two webinar invites instead of one, why certain contacts are limited to three emails per week, and why last quarter's campaign structure outperformed the one before it.

This "why" — the campaign patterns, the segment logic, the institutional knowledge that lives in your MOps team's heads — is what makes AI useful versus generic. Without it, you get "confident errors": technically correct decisions that are strategically wrong. The AI picks your best-performing campaign for a product launch, not knowing your team learned last quarter that this audience needs a customer story before the pitch; that insight came from three launches and sales feedback, not from any field in your CRM.

✅ Execution infrastructure. For email specifically, AI can draft the HTML, but it cannot manage sender reputation, IP warming, or domain throttling. Deliverability is a domain expertise problem, not a coding problem. (Other channels will connect over via API/MCP.)


The short version: vibe-coding eats workflow UI. But enterprises still subscribe to the systems that keep them SAFE — trusted data, governed decisions, and reliable execution.


Where's the line forming between build and buy in your martech stack?

I had dinner last night with ten enterprise B2B CMOs in New York. We kept circling back to three problems: measurement frameworks breaking down, executive perception of marketing as a cost center, and trust being restructured by AI.

I realized these aren't separate challenges; they're different symptoms of the same fundamental shifts in how B2B buying works.

THE AI ADOPTION GAP
We began going around the table rating AI adoption: everyone scored themselves 7+ in their personal use, but 3 or 4 at work. Individual capability outpaces organizational infrastructure, especially when IT doesn't invest in marketing.

As a result, most teams remain focused on basic efficiency gains, not effectiveness.

THE MEASUREMENT PROBLEM
One CMO said: "I feel like we've gone back 25 years in measurement, before the MQL, before Marketo, trying to figure out how to connect with the CFO again."

The old model worked because it was simple. Gumball machine math connected spend to MQLs to revenue. But today, buying isn't linear, and several CMOs reported using account engagement, share of voice, and market share instead.

Plus, there is an attribution double standard: "Marketers are always under pressure to attribute what came out of what went in. Sales, on the other hand, as long as they hit the numbers, wherever that came from, it doesn't matter."

One CMO admitted to framing budget requests as "sales enablement" because that's what gets approved. When the CEO sees marketing as a "necessary expense" rather than a strategic lever, you adapt.

THE BUYABILITY PARADOX
Jann Schwarz and Mimi Turner from LinkedIn led a discussion about "buyability". You don't need the best product; you need to be the most buyable product.

Finance and sales believe B2B decisions are rational, while marketing knows they're emotional. "The questions are like, how do I make sure that I'm not making a big mistake? That's the emotion behind."

In other words, buyers aren't optimizing for the best product; they're minimizing career risk.

THE TRUST CHALLENGE
Buyability comes down to trust, but AI is restructuring how trust gets built, especially as AEO replaces SEO.

Email will soon follow social media's path; with Google's AI inbox, algorithms will determine which messages deserve attention. Email will become an earned channel, not an owned one.

Ultimately, this means Marketing's role shifts from generating demand to building trust with buyers — buyers who are increasingly unknown, researching through LLMs, rarely making shortlist decisions based on word of mouth rather than campaign touches.

THE INFRASTRUCTURE SHIFT
This will require rethinking the marketing stack with three distinct layers: Data for ingesting signals, Decisioning using AI reasoning instead of rules-based automation, and Delivery that extends beyond email to wherever buyers pay attention.


LMK in comments if you want more insights from the dinner!