Your GTM software is not for you anymore

Buy GTM tools for agents, not for interns staring at dashboards.

Adam Schoenfeld (adamGTM) posted the stack receipt: across the GTM tech he surveyed, 52% had an API, 38% had MCP, and only 13% had a CLI. Clay, Apollo, ZoomInfo, and HubSpot sat in that 13%. Twelve months earlier he said he could not spell CLI. Now the buying question changed. Can an agent call this tool without a human clicking through screens?

If your stack only has a dashboard, your agent is locked out of the channel.

Why did dashboards win for so long?

Because the buyer was a human ops person.

You logged in. You filtered. You exported a CSV. You uploaded it somewhere else. The UI was the product. That made sense when the operator was a person with a mouse. It fails when the operator is software that needs to enrich, score, sync, and write back without waiting for someone to free up an afternoon.

GTM engineering is the shift from theater dashboards to systems that move state. AI as operator only works if the tools expose a machine interface. API, MCP, CLI. Not a pretty chart the agent cannot touch.

What do API, MCP, and CLI actually buy you?

Three different doors into the same room.

1. API. Programmatic read and write. Scripts, backends, and agents call endpoints. Most mature GTM tools have some version of this. Adam's 52% is the floor of "an agent can talk to you at all."
2. MCP. Model Context Protocol. The tool shows up inside the agent's workspace as callable functions with schema. Ops builds once. Reps and agents consume in natural language. Good for day-to-day agent use without custom glue every week.
3. CLI. Command line. Coding agents and humans in a terminal can run the same plays. Adam's 13% is the scarce layer. Clay, Apollo, ZoomInfo, HubSpot made that cut in his list. Scarcity is the signal. Teams that can drive GTM from a terminal compound faster than teams waiting on UI clicks.

You do not need all three on day one. You need at least one path that is not "open the app and click." Prefer vendors that ship API plus MCP or CLI. Treat dashboard-only tools as human-only tools. Price them that way.

How does this change the buy decision?

Stop asking "does the UI look modern?"

Ask:

1. Can my agent authenticate and call the core verbs (search, enrich, sync, write)?
2. Is there an MCP or CLI so a coding agent can build plays without a custom integration project every time?
3. Are writes scoped and auditable, or is every call full admin?
4. Can I kill a runaway job without logging into a support ticket?

That last one belongs next to kill-switch thinking for outbound and write permissions. Machine access without rails is how you wake up to a CRM full of junk.

Staff agents as a roster with roles, not as one mega-bot with every login. Each role needs tool access that matches the job. Research agents get read. Sync agents get scoped write. Nobody gets a dashboard password and a prayer.

What about tools I already paid for?

Audit them this week.

Pull your GTM stack into a table: tool, has API (Y/N), has MCP (Y/N), has CLI (Y/N), agent can write (Y/N), audit log (Y/N). Anything that is dashboard-only and sits on a critical path (enrichment, CRM sync, ad audiences, sequencer) is a bottleneck. Renewals become bargaining moments. "We renew if you ship MCP or CLI this quarter" is a real sentence in 2026.

Ehrenberg-Bass / LinkedIn B2B Institute: about 95% of B2B buyers are out of market. Your agents will thrash if they can only operate through screens humans babysit. Machine interfaces are how you keep signal volume high without hiring more people to click export.

How does this fit Signals?

Signals is where you decide what enters the system and what agents may notice.

If enrichment and intent tools are dashboard-only, your signal layer is a human queue. That is not a signal system. That is a part-time job. Buy for agents. Keep dashboards as a window for humans who need to spot-check. Do not let the window be the only door.

Adapt or fail. Pretty UI for the intern is not the product anymore. Callable surface for the agent is.

Start Signals, Convert, Grow

FAQ

Do I need MCP if I already have an API?

API is enough to start. MCP lowers the cost for agents and reps to call ops-built functions without new glue. Prefer both when you can. Do not block a buy that has a solid API and a roadmap for MCP.

Is CLI only for engineers?

No. CLI is how coding agents and power users run the same plays. Founders who live in Cursor or Claude Code will use CLI more than a classic ops UI. The 13% scarcity Adam flagged is the edge.

Should I rip out dashboard-only tools tomorrow?

Not overnight. Rank by critical path. Replace or demand interfaces on tools that block enrichment, sync, ads, or send. Keep niche human tools if they are not on the agent path.

How does this fit Signals, Convert, Grow?

Signals owns what agents can ingest. Convert owns first-touch systems that call those tools. Grow owns scale. A dashboard-only stack stalls all three because the operator cannot act without a human in the loop.