If your stack still treats the chatbot as the main event, this issue will reset how you think about models, budgets, and full-funnel execution.

This week's edition covers:

  • Kevin's Take: why the chatbot window is a dead end, and where the real AI work is actually happening
  • The Signal: Signal: a blind 64-run bake-off between Sonnet 5 and four rivals, plus why the market is buying AI infra and selling seats (bad news if you price per user)
  • Tools & Tactics: a framework for running full-funnel as a system, not a patchwork of campaigns your pipeline keeps stalling in
  • Quick Links: AI mentions lifting branded search, OpenAI's tiered GPT-5.6 partner drop, agent usage up 56x inside OpenAI

Kevin's Take

The chatbot window is not where the real work is happening anymore

I think we're all still designing marketing workflows around a UI that's already becoming obsolete. Ethan Mollick's piece on the twilight of chatbots argues that the chat window, the prompt-and-response back-and-forth we've all trained ourselves on for the last three years, is giving way to agents that just go do the work while you're doing something else.

The easy thing to do right now is to keep getting better at prompting, building prompt libraries, running prompt training sessions, arguing about which model gives better copy. My worry is that the actual leverage has moved somewhere else and a lot of that work is going into the wrong layer.

What I've noticed in our own work is that the moments where AI actually changes what we can ship are almost never in the chat window anymore. They're in the background. It's a skill we wrote once that runs every time we build a component, it's a script that pulls transcripts out of a folder and turns them into positioning notes overnight, it's an n8n workflow that watches for something to happen and then does five things on its own. The chat is where we set the thing up, and then the thing runs.

If I'm honest, the days where I'm most productive are the days I barely touch Claude directly at all, because the work is happening while I'm in a meeting or asleep (my favorite thing to do is code at the gym using Claude Code /remote control).

The implication for marketing leaders is that hiring for "prompt fluency" is aiming at the last war. What actually matters going forward is whether your people can define a job clearly enough that something else can go do it, and whether they can check the work when it comes back.

That's a different muscle than writing a clever prompt. It's closer to being a decent manager of a junior team, which most marketers already know how to do, they just haven't connected it to AI yet. I'd rather have someone on my team who can scope a piece of work and review what comes back than someone who can prompt beautifully but still has to sit in the chat window to get anything done. The chat window is where AI started. I don't think it's where it's going to live.

That's it for this week. Talk soon.

— Kevin Kerner, CEO, Mighty & True

The Signal

Sonnet 5 tested against 4 rivals across 64 blind runs (1 min read)

Claire Vo built a live benchmark and ran 5 frontier models through 64 blind tests including prototypes, PRDs, and agent voice tasks. The results ranked Sonnet against peers with surprising outcomes on quality and cost tradeoffs.

Why it matters: If your team is picking a default model for content, prototyping, or agent workflows, this head-to-head gives you real data to challenge your current stack choice.

The market is buying AI infra and selling seats (1 min read)

Across 87 public SaaS companies, only infrastructure/dev tools (+68.5% YoY) and security (+17.6%) are up while seat-priced application software is down. The market is buying the AI stack and selling the application layer, forcing CIOs to reallocate toward AI infrastructure.

Why it matters: If you sell seat-priced software or pitch your CIO for more licenses, the budget math has already turned against you and you need to reframe your value in AI-outcome terms.

Tools & Tactics

A framework for running full-funnel marketing as a system (1 min read)

Andrei Zinkevich outlines the key pillars of building a full-funnel marketing operational system for B2B teams. The framework covers how to structure marketing ops across awareness, demand, and revenue stages.

Why it matters: If your full-funnel motion is a patchwork of campaigns instead of a system, your pipeline will keep stalling at the handoff between marketing and sales.

Score your B2B marketing — free

The Blueprint Score benchmarks your brand's competitive velocity (content activity, market presence, search momentum) against 50 mid-market B2B tech companies. Enter your domain, get your score in 5 minutes.

▶ Run your Blueprint Score

Want a second set of eyes?

Running a marketing team right now feels like rebuilding the plane mid-flight. If you want a second set of eyes on what is actually working (and what you can safely cut), we do free 30-minute marketing audits. No pitch deck, just a honest conversation. Reply to this email and we will find a time.

▶ Book a 30-min audit →