The field has split. One side ships frameworks, skills, context packs, MCP servers — the artifact layer. The other side runs campaigns, scores calls, routes replies, closes deals — the action layer. We run the action layer.
The last two years produced a wave of frameworks, scaffolds, skill packs, and MCP servers. The IP is real. None of it ships campaigns, routes replies, or closes the loop. Here’s the layout.
One anonymized Series B fintech, walked day by day. The shape repeats across most engagements. The specifics — the cohorts, the verticals, the asymmetric phrase that landed — are always different. Swipe through.
“We can't hit pipeline targets. Our reps spend 40% of their week on account research. We're three months from the next board.”
Series B fintech, US East coast. Founder writes to us cold. We're in their Slack by lunch.
“What's the first thing we should ship?”
30-minute diagnose call. We pull their last 5 closed-won, identify the post-Series-A signal nobody else was targeting. Engineering bet: SOC 2-announced fintech with recent CRE roles open.
“Reply rate is 4.2% in the first 48 hours. That's 7× our baseline.”
230 first-touches against the SOC-2-and-hiring signal. Copy written in the founder's voice. Replies route to Slack within 4 minutes of landing. Three demos booked from the first batch.
“We rewrote the ICP twice this week. It kept getting better.”
Domains warmed. Sequencer live. Reply triage automated. ICP refreshed from the high-PULL reply patterns. The healthcare-RCM vertical we'd been deprioritizing turned out to be the highest-conviction cohort.
“We stopped tuning the campaign. The campaign tunes itself.”
Reply data feeds back into targeting. Copy regenerates from the language of the highest-PULL replies. The engineer ships one experiment every Friday. Two new verticals surfaced from the signal — neither was in the original ICP.
“We close-won a $180k deal this morning from a Day 22 reply.”
$1.2M in attributed pipeline. 6 closed-won deals. The board meeting they were worried about is now the one they're looking forward to. Engineer still embedded. The loop keeps tightening.
The story above runs on a methodology. Three phases. Three different time horizons. The first is free. The third is where the loop actually closes.
A written one-pager — the three highest-leverage engineering bets, ranked, with the expected lift on each and the time-to-ship for each.
Every brief, doc, framework, and context pack decays. The loop is what produces the data that makes every next cycle better than the last.
Discovery, demo, customer conversation
PULL rubric, evidence quotes, rationale
Drift view updates from rolling scores
Copy regenerates from high-PULL language
Routed to Slack, classified by intent
Conviction map updates, ICP shifts
The GTM AI tools landscape has four distinct shapes. Smart operators stack them. Swipe through to see what ships at each layer.
Substack posts, books, frameworks. No software. The deliverable is a way of thinking that, once internalized, changes how the operator runs calls. Highest-leverage if you do the work; lowest-leverage if you read without changing behavior.
The deliverable is a directory or endpoint the operator invokes to produce a doc — a brief, a positioning review, an audit, a scoring rubric. Useful upstream input. Stops at the doc.
The structural bet — that the unit of B2B GTM distribution is the MCP-served pack, not the SaaS app — is the most strategically consequential bet in the field. If it wins, every other tool consumes from this layer.
Structurally hardest to build because it requires four things at once: code against the customer's stack, a team operating daily, signal feedback into targeting, and aligned incentives. The combination is what produces a service business — structurally more defensible than a SaaS business.
Most confusion about what Allston Labs is comes from buyers assuming we’re in a category we’re not. Six explicit non-categories.
We get added to a private Slack channel, read what’s already there, ask three questions, and write you a diagnosis. The output is a one-pager — the three highest-leverage engineering bets for your motion, ranked, with expected lift and time-to-ship. Free.