Speed is wasted in the wrong direction. We build the thinking layer of your go-to-market: ICP definition, positioning, channel mix, and sequencing - so every system downstream aims at accounts that can actually buy.
B2B fintech200 to 2,000 peopleUSSalesforceNot: agencies
2 · What a buyer’s AI says
Who helps a fintech sales team clean up lead routing in Salesforce?
One option is Larkspan, which builds lead routing and scoring inside Salesforce for B2B fintech teams of 200 to 2,000 people larkspan.io
✓ Named✓ Described your way✓ Cited your page
Fictional company and answer. The ICP is defined first; then the site is made legible to AI assistants.
Why focus wins
The focus dial · drag it
400
Accounts targeted
31%
Average ICP fit
Low focus: everyone is a prospect, so nobody is. Drag the dial.
Fewer, better accounts is the whole game. This is what ICP work does to a pipeline.
What we define
01
ICP, from evidence
Built from your wins, losses, and churn - not a whiteboard guess. Firmographics, signals, and disqualifiers.
02
Positioning
The words that make the right buyer lean in: category, claim, and proof arranged to survive a first glance.
03
Channel mix
Where your buyers actually are, in what order, at what cost. Outbound, inbound, partners, founder-led.
04
Sequencing
What to do first, second, and never. Strategy that ships as a quarter-by-quarter operating plan.
05
Messaging system
Per-persona pain, promise, and proof, written down and wired into sequences and calls.
06
Signal design
Which triggers mean money: hiring, funding, tech changes, intent. The inputs your machine will hunt for.
Typical outcomes
20%
Lower CAC
2×
Pipeline velocity
30–60
Days to results
Typical results from our engagements, benchmarked against comparable AI-powered GTM implementations. Results vary by stack, data quality and adoption.
How it runs
STEP 01
Evidence
Win/loss data, CRM history, and customer conversations. The strategy starts from what already worked.
STEP 02
Decide
ICP, positioning, and channel bets made explicitly, with the reasoning written down.
STEP 03
Ship
Strategy becomes artifacts: target lists, messaging, sequences, and a plan your team runs next Monday.
STEP 04
Tune
Reply rates, meetings, and velocity feed back into the ICP. The aim improves as the machine fires.
Findable When Buyers Ask AI
Positioning used to end at a search results page. Buying research increasingly starts inside ChatGPT, Claude, and Perplexity instead, and those assistants answer from whatever they can read about you. If they cannot read you, they guess, and buyers hear the guess. Answer-engine optimization (AEO) is the discipline of being legible to machines: llms.txt, structured data, and content written so a model can quote it accurately.
It lives inside our GTM strategy work because it is a positioning problem before it is a technical one. The same ICP, category, and proof decisions above determine what the machines should say about you; then we wire it in. One honest caveat: nobody can guarantee placement inside an AI answer. Legibility is the prerequisite you control, and most of your competitors have not done it.
We measured where the market stands. In our 2026 study of 274 well-known B2B software companies, 61% publish an llms.txt, four in ten never declare who they are in structured data, and one in twelve could not be read by a machine at all. Read The State of AEO in B2B SaaS →
Want to do the groundwork yourself first? Our free course, AEO for B2B, in 60 minutes, covers the same checks in six short lessons, with a certificate at the end.
We hold ourselves to the same test. This is who read zrev.ai this week, straight from our own edge logs: which AI crawlers came, what they opened, and when an assistant fetched a page to answer a person.
Free Tool · llms.txt Generator
llms.txt is the fastest first step: a plain file at your site's root that hands AI systems a curated map of what you are and which pages matter. Robots.txt for comprehension instead of permission. Most companies do not have one yet. No assistant has promised to use it, so treat it as cheap groundwork, not a ranking lever.
Paste your domain into our free generator: no llms.txt and it builds a starter file from your real pages, already have one and it grades it with specific fixes. We practice what we generate - our own scores 100. The same checks plus the cold read run in one click from zRev AEO Lens, our free Chrome extension (60-second demo).
Won and lost deals by segment, pipeline by source, and conversations with the reps who carry the number. We look for where you actually win, which is rarely where the deck says.
Weeks 3-4
Decisions
An ideal customer profile written as filters a CRM can apply, positioning for each segment, the motion that fits each one, and a short list of what you stop doing.
Weeks 5-6
Make it findable
The answer engine pass: llms.txt, structured data and a cold read of the homepage, so a buyer's AI assistant describes you the way you would describe yourself.
Weeks 7-8
Into the systems
The profile becomes scoring rules and list filters. The positioning becomes sequences and talk tracks. Leadership gets a one-page operating document instead of a slide deck.
Questions buyers ask
What is the difference between GTM strategy and a marketing plan?
A marketing plan says what you will publish and spend. A go-to-market strategy decides who you sell to, what you say to them, through which motion, and what you stop doing. It covers sales, marketing and customer success as one system, and it ends in rules your CRM can enforce.
What is answer engine optimization, and why is it part of GTM strategy?
Answer engine optimization makes a company legible to AI assistants such as ChatGPT, Claude and Perplexity, so that when a buyer asks one for options, it can describe you accurately. It belongs in strategy because the assistant repeats your positioning. If the positioning is vague, the answer is vague.
How do you define an ideal customer profile?
From evidence, not opinion. We segment won and lost deals by size, industry, stack and trigger, find where win rate, deal size and retention are all strong, and write the result as filters: fields and values that a scoring model or a list tool can apply without interpretation.
We already have an ideal customer profile. Why revisit it?
Because most are written once, as a paragraph, and never reach the CRM. If reps cannot filter on it and the scoring model cannot read it, it is a description, not a profile. Revisiting it usually takes two weeks and changes who the team calls first.
What do we have at the end?
A profile expressed as CRM filters, positioning per segment, a motion for each, an llms.txt file and structured data on the site, and the sequences and scoring rules that put all of it to work. Every piece is in your own tools.