Everything I'm seeing at the moment. This comes from a stream of thought and then I applied it to this little post. Hope it helps. Context: I run an AI venture studio and help B2C/B2B companies build their GTM engines. So you're getting the AI-native perspective and the incumbent perspective as well. This is what I'm seeing going into 2027:
1. There are two funnels now. The human one still works (landing pages, SEO, community, referrals). The second one is for agents. It matters more every month. People already hand small purchases to an agent, especially anything under $500. I call it agent economicus. If your pricing isn't on your site, the agent can't match you to the buyer's budget and ranks you lower. SEO turned into GEO. The next step is getting the agent to pick you. You get one key and the agent picks which data provider to call based on quality and cost.
2. 80% of companies run the same motions. Lead scoring, ABM, content, SEO, ads that learn on their own, outreach. The code behind all of it is close to worthless now because anyone can build it in a weekend. What makes you different is the other 20%. Your definitions, your tone of voice, your data, your rules. Your risk taking, your leaps of faith. LLMs are still crap at that. To be honest, so are most marketers.
3. So build it yourself. The biggest waste in a B2B budget right now is a multi year contract with a software vendor. Most of what you pay for is the interface. The real value sits in clean data, deduplication, the semantic layer and access control. Recently a CEO showed me the lifecycle CRM he vibe coded over a weekend. His engineer needs three to four weeks full time to build the real version, plus every application they need on top. It will beat anything they could buy.
4. Marketing departments now have (or should have) more engineers than ever. They are turning into engineering departments. I expect at least one, two or three engineers working full time inside every marketing team, often more. These are AI native full stack people who know Postgres and React.
5. With that setup the marketer works from the command line. One prompt finds 250 target accounts with real buying signals. It maps the buying committee (economic buyer, champion, end user, the IT person who will block you). It pulls email, LinkedIn and phone, then scores each person from 1 to 100. Five minutes, about $4. The same session builds the ABM plays for each account and writes the content around them. Performance ads and retargeting run from there too. The agent launches them, reads the numbers and moves budget to what works. This is how we work now. If we need a user interface to be able to communicate with other humans, we just whip it up ourselves…
6. Here's what AI native looks like in our own ventures. One of them runs on 0.1 FTE. A developer looks at it half a day a week. Outreach, content, programmatic SEO, ads, support and onboarding all run by themselves. I get a ping when a guardrail trips, like a discount request or a bug.
What surprised me most is the work that never shows up in the first place.
No weekly status meeting. Every agent writes a log, so the status is always live.
No ticket queue. The agent answers in minutes and sends me the odd ones.
No chasing invoices. Customers pay at checkout and the books match every day.
No marketing to sales handoff. The same agent researches, writes and follows up.
No approval rounds. The guardrails get written once and I check a sample.
No hiring plan to grow. More customers just means a bigger compute bill.
7. AI native fails when one of three things is missing. Leadership that gets it and gives a mandate. A team that can set it up. The right motions picked first.
by the way, this is how I do transformations to “ai-native” BUs… (I’ll write about this soon)
8. Distribution is the hard part now. Building got easy. Once everyone runs the same automated playbook there's no friction left. Anything without friction loses its value. So we go back to the craft. Hooks, angles, positioning, an audience that trusts you, maybe even physical stuff again. The teams I work with tell me "I finally feel like I'm doing marketing again." They used to spend most of their week on reports and Salesforce fields.
9. Agents are as creative as you let them be. LLMs are great at pattern matching and bad at judgment and risk. They can't come up with a truly new billboard idea yet. Your output is only as good as your brief, because prompting is just briefing. It's the revenge of the creative brief.
10. What to learn. At my previous company we trained 50,000 people. Today my answer is to skip the course. Learn the command line (Claude Code, Codex). Build a second brain that holds your context. Stay culturally curious so your briefs are good. Then go grow something, a side project or one ABM campaign. You learn it by doing it.
David
