A retainer
Account managers and a pod you never meet. Open pricing. Decks, reports, and activity. The motion leaves when they do.
For AI companies · Seed to Series B
The usual alternative is an agency retainer: account managers, open-ended pricing, decks instead of systems. Xeme compiles TAM, ICP, personas, and signals into an engine that prospects, personalizes, and books. Built in Clay, n8n, Python, and Claude Code. You own the stack.
Prices on the pageEngine live in under a weekReply within one business day
01 — The choice
Account managers and a pod you never meet. Open pricing. Decks, reports, and activity. The motion leaves when they do.
A list, a sequencer, a CRM nobody trusts. Someone on your team still has to run every step. The tools do not own an outcome.
You name the revenue line and the guardrails. One pipeline handles signal, copy, routing, and attribution. It lives in your accounts.
02 — The loop
One loop, end to end. Most teams run these as four departments. The lab runs them as one pipeline.
Job posts, stack changes, funding events, site visits. The lab watches who is buying this week, not who exists.
TAM, ICP, personas, and the channel plan become agent config before anything is wired.
Agents research, personalize, send, route, and follow up across email, LinkedIn, and ABM pages.
Every meeting and dollar traces back to the agent that sourced it. What does not attribute gets killed.
03 — The board math
The deck promised pipeline. The budget funds a little more than the team you have. The old answer was an SDR team and an agency retainer. The AI-native answer is a compounding system, with humans on judgment and deals.
A fast engine pointed at the wrong market is expensive noise. Strategy compiles first. Then the roster takes over.
04 — Set in motion
A named process, your data, a human on the decision, and a number measured before and after. A chat licence is not the product. Nothing is migrated off the tools you already run.
Meetings, pipeline, or a closed-won line. Guardrails first: who is a fit, what never gets sent, and what a person still owns.
Clay, the CRM, the domains, and the people who close. The engine connects to systems you already run. No 18-month migration.
Before a live mailbox, it runs against past replies, lost deals, and the calls that closed. You correct it. Then it ships.
When an account heats up, the closer gets the brief in Slack and the CRM. Agents did the preparation. A person takes the meeting.
05 — The roster
Every agent below has shipped in production and been tied to revenue. Humans handle judgment and deals. The roster handles the rest.
Watches job posts, tech-stack changes, site visits, and LinkedIn engagement. Knows who is in-market this week.
Role mapping, intel, gap, positioning, generation. Per-contact emails ship after a second model scores them.
Async fan-out across 100+ domains per batch, including JS-rendered pages. Analyst-days of intel, no analyst.
Visit-tracked pages measuring Share of Answer and AI Share of Voice, plus per-account video in sequence.
An LLM rubric scores fit and intent with written reasoning, pushes context into the CRM, and pings the closer.
Resolves anonymous traffic to a person and an account, enriches through Clay, and sequences the same day.
06 — Inside the engines
These are the production systems behind the numbers. Every stage is named. Every tool is listed.
Personalization is not a first-name merge tag. It is a research pipeline that ends in a sentence only this prospect could receive.
Outcome1.5%+ positive replies across 150+ sending domains. 95%+ deliverability sustained.
Parse title, seniority, team scope, and tenure. Map a persona to the number this person is judged on, and what breaks it.
Crawl the site, docs, pricing, changelog, and careers page. Headless rendering so modern AI-company sites read the same as static ones.
Attach the trigger that put them in queue, with a timestamp. Timing is the personalization.
Compare the current motion with category benchmarks and isolate the one credible gap worth naming.
One angle: displacement, expansion, or timing. One claim, one proof point, mapped to the KPI from stage one.
Subject plus a body under 90 words. First line is the signal. Middle is the gap. Close is one low-friction ask.
A second model grades factuality, specificity, and spam risk from 0 to 10. Below threshold, it regenerates. Above, it enters sequence.
A per-account asset the buyer can visit, a loop that watches the visit, and a sequence that reacts in the same hour. Built for six-figure deals where three stakeholders have to say yes.
OutcomeEnterprise logo closed at $25K MRR. Anonymous visits converted to multi-threaded pipeline.
Tier on signals, not firmographics alone: fresh funding, GTM hiring, stack adoption, engagement history.
47+ data points per account, including a baseline of Share of Answer and AI Share of Voice in their category.
A personalized page per account, visit-tracked. Tier 1 also gets a per-account video recorded against that page.
Email carries the page. LinkedIn carries the relationship. Every touch lands inside the buying window.
Revisits, watch-through, reply sentiment, and colleague visits re-score the account. A hot account pings the closer.
The human enters with the brief: signal history, assets viewed, stakeholders touched, suggested talking points. Agents did the 95%. The closer does the 5%.
OutcomeQualification cut from 48 hours to 9.6 hours.
Forms, replies, and de-anonymized visits land in one queue with context attached.
A rubric grades fit and intent with written reasoning. Disqualifiers get a polite path.
Qualified leads push into the CRM with the reasoning. The right human is pinged in Slack.
A context-aware first reply or booking link goes out while the lead is still on the site.
OutcomeShare of Answer becomes a tracked pipeline channel, for companies whose buyers ask AI first.
Measure how often ChatGPT, Claude, Perplexity, and AI Overviews mention you against competitors on real buying prompts.
Rebuild key pages into self-contained passages: claim, number, and context in one block. Crawlable HTML. llms.txt in place.
Publish the assets those engines cite: comparisons, benchmarks, teardown data.
Re-run the prompt panel monthly. The number gets its own attribution line.
07 — The build
The Sprint is not a discovery phase. It is a build schedule with dated deliverables. Three greenfield engines have shipped on this cadence.
TAM, ICP, personas, signal selection, channel plan. Nothing is wired before this ships. About two hours of founder time.
Domains, mailboxes, warm-up, SPF, DKIM, DMARC, CRM instrumentation. The signal layer goes live.
Personalization and ABM engines shipped. First sequences in flight. QA rubrics tuned on live sends.
First qualified meetings land. Speed-to-lead goes live on inbound. Iteration follows reply data.
Attribution, runbooks, and handoff docs. Every dollar traceable to an agent. You own everything. The Lab is optional from here.
08 — Operating system
TAM, ICP, personas, and signal selection are decided before a tool is wired.
Static lists say who exists. Job posts, stack changes, visits, and engagement say who is buying this week.
If a client pipeline needs a babysitter, it is not finished. Agents score, retry, branch, and ship. A person still owns judgment and the deal.
Reply rates are diagnostics. ARR is the metric. Every agent attributes to closed-won or gets killed.
09 — The difference
| Agency retainer | Xeme | |
|---|---|---|
| Who | Account managers, handoffs, a pod you never meet. | The engineer who builds the system. |
| Pricing | Open-ended retainers. Pricing revealed on a call. | Fixed builds and flat subscriptions, printed on this page. |
| Deliverable | Decks, reports, and activity summaries. | Production systems running in your infrastructure. |
| Ownership | The system lives in their accounts. You rent it. | Your Clay, your CRM, your domains. |
| Metric | Sends, opens, replies. | Closed-won revenue, attributed per agent. |
| When you leave | The engine leaves with them. | The engine stays and keeps running. |
10 — Engagements
Fixed scopes. Prices on the page. You own what ships.
The full engine, greenfield: strategy, signal layer, personalization, sending infrastructure at 95%+ deliverability, CRM instrumentation and routing.
Ongoing operations after the Sprint: signal expansion, campaign iteration, ABM assets, inbound routing, and a monthly attribution review.
A paid audit of the current GTM stack and signals, delivered as a node-by-node build plan.
11 — Proof
One engineer, one system. Fastest build live in under 3 days. 300+ mailboxes orchestrated. Every dollar attributed from signal capture through closed-won. Full case study on request.
ARR in 6 months for a $12M ARR services firm. 250+ domains at 95%+ deliverability. Qualification cut from 48 hours to 9.6.
Targeting accuracy. Asset-led ABM and outbound for an early-stage cybersecurity startup. NDA signed.
Engine live in under 3 days. Revenue attributed end to end across nine months. 300+ mailboxes orchestrated.
250+ sending domains, qualification cut from two days to under ten hours, $180K+ new ARR inside six months, 40% pipeline-efficiency gain.
Targeting accuracy up 45%. Asset-led ABM opened enterprise accounts the team had chased for a year.
12 — Lab notes
When a champion looks up the lab, they should find the workflows: node diagrams, prompts, and the numbers already on this page. The first essays are not up yet. The drop is one production workflow a week, with no gate.
Work email. It lands with ops@xeme.co. No sequence is attached.
13 — The loadout
If your stack already includes some of these, the lab builds around what you have.
14 — The operator
Xeme is run by Rasul Shaikh. Three greenfield GTM builds attributed to revenue, shipped in Clay, Python, n8n, and Claude Code.
You work with the person who builds the system. Scoping, architecture, shipping, attribution. Capacity is capped at two builds a month for that reason.
15 — Questions
Venture-backed AI/ML companies from Seed to Series B with a sales-led or hybrid motion. Founders doing founder-led sales, first GTM hires, or revenue leaders whose SDR math stopped working. Not a fit: pure PLG with no sales motion, or anyone who wants rented SDR bodies.
Hire one later, for the judgment work. An SDR doing manual research and sends costs more than a Sprint per quarter and produces less pipeline than an instrumented engine. The teams that win give their best person agents, not a team to manage.
Yes. Everything is built inside your infrastructure: Clay, HubSpot or Attio, n8n, sending tools, domains. When we part ways, the engine stays and keeps running.
The engine goes live in under a week. First qualified meetings typically land by week 3. The full system is instrumented at 30 days. The schedule is above.
About two hours of founder or GTM-lead time in the first week for the strategy compile, access to the CRM and domains, and sign-off on the signal plan. After that the roster runs. You take the meetings.
Usually the most useful case. Most Sprints start from a half-built stack. The Teardown maps what to keep, and the Sprint rebuilds the engine around it. Existing tools lower the build time, not the value.
Fixed-fee Sprint, flat monthly Lab subscription, optional performance kicker on closed-won revenue. Prices are printed on this page on purpose.
Multi-domain setup, SPF, DKIM, DMARC, warm-up, and inbox rotation. 95%+ deliverability sustained across 150+ domains in production. Volume scales only as reputation allows.
The engineer. No juniors, no offshore pods, no handoffs. That is also why capacity is capped at two builds a month. The agents do the volume. One human does the judgment.
The engine stays. It lives in your accounts with runbooks and handoff docs. Lab clients get a 30-day wind-down with a full transfer session.
Open channel
One call to scope the engine. We reply within one business day.
Accepting Q4 clients · 2 build slots a month