Stop building GTM chatbots. Ship skills.

Persona chatbots rot. Skills compound.

On Sep 28, TechCrunch's Sarah Perez reported that Google is killing Gemini Gems in favor of skills, with auto-migration set for Nov 17, 2026. Slash-command skill UX replaces the gem-as-chatbot pattern. That product move points at a GTM truth I keep seeing in teams: the "AI SDR buddy" and "persona GPT" become stale prompts nobody trusts, while a small set of reusable skills (research brief, DQ check, AE handoff) get invoked every day.

A skill you can invoke beats a chatbot you babysit.

What is a GTM skill versus a chatbot?

A skill is a named job with inputs, tools, and a done definition. A chatbot is a personality with a prompt.

"Act as a skeptical VP of Sales" is a chatbot. "Produce a one-page account brief from CRM + last 10 site sessions + open opportunities, flag DQ, stop if exclusion hits" is a skill. The first invites endless chat. The second returns an artifact another human or agent can use. Google moving Gems to skills is the consumer-shaped version of that shift. Grow teams should make the same cut on purpose.

This is AI as operator, not AI as mascot.

Why do GTM chatbots fail after the demo?

Because nobody owns freshness, and the job was never specified.

The persona drifts. Pricing changes. ICP narrows. The chatbot still sounds confident. Reps stop trusting it, then use it for jokes, then forget it. Skills fail differently. If the research brief skill breaks, the missing artifact shows up in the AE's morning stack. You fix the skill. You do not "have a conversation" with a ghost.

Staff agents like a roster. Roster rows are jobs. Chatbots without jobs are costumes.

Which skills should a GTM team ship first?

Four that pay rent without needing a science fair.

1. Research brief. Account, contacts, last meaningful touches, open risks. Hard stop on exclusion and DQ. Scope tools on purpose, same rule as agent research tool scope.
2. DQ check. Explicit rules: segment, geo, competitor, bad fit product use. Returns pass/fail plus reason. No essay.
3. AE handoff. When a meeting books, package context the AE will actually read: why now, what was said, what not to pitch.
4. Message test. Before you buy more seats, run five variants against a frozen ICP sample and score meetings, not vibes.

Each skill gets an owner, a version, and a place it is invoked (slash command, workflow step, ticket button). That is rails for Grow: reuse beats reinventing the prompt in every chat window.

How does this fit Grow?

Grow is compounding what already works.

A chatbot is a one-off conversation. A skill is an asset. When you improve the DQ check once, every outbound path inherits it. When you improve the handoff once, every AE meeting starts less cold. That is how AI stops being a demo tax and starts being an operating margin.

Ehrenberg-Bass / LinkedIn B2B Institute put most buyers out of market. Skills that protect AE time and kill bad fits matter more than a chatbot that will cheerfully "nurture" the 95%.

What should I do Monday?

Kill or archive one persona chatbot that has not produced a meeting-linked artifact in 30 days. Write one skill card: name, input, tools, output, fail conditions. Implement it where reps already work (ticket, slash command, or workflow). Measure invocations and whether AEs used the artifact. Then add the next skill. Do not build a sixth chatbot.

Adapt or fail. If Google is migrating gems to skills on Nov 17, 2026, you can migrate your GTM stack without waiting for a calendar invite.

Start Signals, Convert, Grow

FAQ

Are chat interfaces always wrong?

No. Chat can invoke a skill. The failure mode is making the chat persona the product instead of the skill output.

How many skills should we start with?

Two or three that hit research, DQ, or handoff. Ten half-built skills is another form of chatbot sprawl.

Do skills replace agents?

Skills are what agents (and humans) call. Agents still need roster jobs, tool scope, and rails. Skills are the reusable unit of work.

What is the Nov 17 date?

Sarah Perez's Sep 28 TechCrunch coverage of Google killing Gemini Gems for skills cites auto-migration on Nov 17, 2026. Use it as product proof of the skills pattern, not as your GTM deadline.