Mission Control
sandbox · read-only mirrorsWelcome to GTM OS.
The simple version: everything your company knows flows into one brain. The brain runs engines that do the daily work of marketing and sales, around the clock. Your people stop doing the grunt work and start approving, steering and deciding. That is the whole idea.
A mix of build and buy. We build a lot, everything that runs on your data and your edge. We buy what we can't build better, then wire it into the same brain.
ONE CONDUCTOR
I'm finally getting my job back and focusing on what's important.
GTM Score
Every account ranked nightly, with the next best move already drafted. One marketer covers the accounts of ten.
GTM ABM · the buying-committee engine
A tiered target list, a stakeholder map per account with the gaps exposed and exec plays like the couriered CEO letter. Now with a content studio that creates, localizes and files every asset in a 3D library.
GTM Ads · the self-learning ads engine
Bids toward paying customers, refreshes its own creatives, learns from every closed deal. You approve the moves.
GTM Outreach · the self-learning outreach engine
Writes like someone who did two hours of homework, learns from every reply. A human sends every message.
GTM Sales · deals kept honest
Calls coached, deals reviewed nightly for drift, proposals priced inside guardrails. Reps sell, the machine carries the homework.
GTM SEO · GEO
A page factory with a quality gate, plus the play that gets you found in LLM chats when buyers ask ChatGPT who to pick.
Get found in LLM chats
120 real buyer questions tracked across ChatGPT, Perplexity and AI Overviews. Citation share is the new rank.
All the other use cases
Churn warnings, quote configurators, crazy experiments. Anyone asks, the brain specs it, the engineer ships it in two weeks.
Good morning. Here is what matters.
Built for the person who owns the number. Pick your seat, the OS reslots the metrics, the feed and the queue underneath. All numbers are dummy data.
Every line opens. Click one to see what happened, to whom, under which rule.
Every line is one thing an engine did on its own. Click a line for what happened, to whom, under which rule.
ARR, twelve monthsTO €4.24M
Revenue by motion28 DAYS
Where the growth sat28 DAYS
| Movement | Accounts | ARR |
|---|
What sits behind the cutHOW TO READ IT
Every cut sums back to the same total. A slice never invents revenue the top line does not have.
Share of paying accounts by how much they lean on the engines. A small head carries the revenue. The tail is where churn starts.
GTM Score
Every account scored nightly PLUS the next best step, drafted and queued. No black box, every score explains itself and the whole thing syncs into HubSpot for the people who live there.
Scores are a ranking, expect a few percent of nonsense. The self-driving-car rule applies, the rep stays behind the wheel. Something looks wrong? Flag it on the account, wrong twice = the rule changes.
How GTM Score builds a score
Plain English, no black box. Scores are a relative ranking, recalibrated weekly against closed-won and closed-lost deals. The weights live in a sheet the ops lead owns, the engine reads it, nobody hand-sets a number.
How much the account looks like your ideal buyer before anyone lifts a finger. Size, industry, tech stack, region. A perfect-fit company starts with points even at zero activity.
What they actually do. Pricing visits, doc opens, webinar seats, trial activity, reply patterns. Doing beats being, a small account moving fast can outrank a perfect logo that never visits.
Outside signals. Funding rounds, hiring patterns, competitor contract windows, category research. The signals that say a buying window is opening.
Buying intent decays in days, so the score decays with it. Today beats yesterday beats last week.
HOT 85+ · act this week. WARM 65-84 · work the plan. COLD under 65 · nurture watches, reps spend zero time. Every score shows its evidence, hover any row in the list.
RUNNING TOTAL · DECAY + CAPSUnder the hood the score is a running total per account. Every signal ages out on a decay curve and every signal type has a cap, so one noisy topic can never dominate an account’s score. Twelve pricing-page hits stop counting once the cap is reached, the next point has to come from a different kind of evidence.
THE VERIFICATION GATEA story must name the account or it goes to a review list, never to a seller. A wrong name in a rep’s queue costs more trust than an empty slot, so the ambiguous ones wait for a human match.
WHAT DOES NOT FEED THE SCORECompany grade and raw size. Those live in the GRADE (A to E, human can override it in the CRM), so a small account moving fast can outrank a big logo doing nothing. And the CRM's own predictive score is ignored, one black box is enough to distrust.
HOW THE MODEL EARNS TRUSTReps flag anything wrong on the account itself (the ⚑ button), flags feed a Friday rules review, wrong twice = the rule changes. Every quarter the win-loss analysis reweighs the factors against what actually closed. And the pilot question never changes: are the people using this list selling more than the people who aren't? The feedback loop reads task closure, actioned, open or no touch, which only works if the team closes tasks, we say that out loud.
Flag Helios Retail Group
Granular beats general. One flag on one account teaches the engine more than a paragraph in a channel. Pick what's wrong:
Flags land in the ops queue with the full score evidence attached. Reviewed every Friday. If the same rule is wrong twice, the rule changes and you'll see the change noted in the model card.
Helios Retail Group
Draft attached in the buyer’s language. Due today. Created automatically from the nightly score run, cancel window honored.
Pricing page x4 this week, CFO office opened the security docs, champion quiet 9 days. Full factor breakdown in the GTM OS record link.
GTM ABM
Enterprise deals are won by committees, so this engine works the whole committee. A tiered target list, a stakeholder map per account with the gaps exposed and plays matched to how big companies actually buy. The score decides who is warm, this decides how you surround them. Create, Localize and Library turn every play into on-brand assets for each persona and market.
Ranked by account intent. Click an account, the right side becomes its war room.
Create
Write the point of view once. The studio renders it per format, persona and region. The Brand Brain travels with every draft.
Localize
One source of truth, rendered per market. The brand voice and a rulebook for each market travel with the copy, so every region reads native.
Illustrative data for this prototype. Markets, owners and coverage figures are placeholders and nothing here is connected to a live system.
Every asset the engine writes lands here, on brand and ready to version.
White papers, blog posts, comparison guides and account plays. Each exists in a long, a medium and a short cut from the same source.
Illustrative content produced for this prototype. Company names, competitors and every figure are placeholders. Nothing here is real data.
GTM Ads
The console optimizes toward paying customers with their revenue attached. Every move ships with its reasoning and waits for your yes.
| CHANNEL | SPEND | CUSTOMERS | REAL CAC | REVENUE |
|---|---|---|---|---|
| Search · brand | €8.2k | 71 | €89 | €3.4M |
| Search · generic | €96k | 40 | €2,410 | €4.8M |
| PMax · catalog | €28k | 13 | €2,190 | €0.9M |
| Social · prospecting | €22k | 24 | €917 | €1.1M |
The old dashboard showed platform CPA €270 and called generic search the winner. Real CAC tells the opposite story.
Mechanical moves inside agreed guardrails run alone and get logged in the night log. Anything that moves money, kills spend or changes strategy lands here first.
| PLATFORM | SPEND · MO | REAL CAC | 7-DAY TREND | VERDICT | NEXT |
|---|---|---|---|---|---|
| GGoogle Search | €104k | €1,180 | FIX | cap generic, protect brand | |
| ∞Meta | €22k | €917 | SCALE | lookalike seeded from closed customers | |
| PPMax · catalog | €28k | €2,190 | FIX | kill segments with zero closed revenue | |
| RReddit | €4k | €640 | TEST | the unowned conversation, probe live | |
| ♪TikTok | €6k | €1,480 | HOLD | creator test, verdict in 4 days |
Trend bars show customers per day, so a rising bar means the platform earns more budget. Verdicts refresh nightly with the rescore.
Optimization finds the local peak. These are the campaigns that move the whole curve, each one probes at €1k with pre-registered kill criteria before it scales to €4k.
Chat-first ad funnel
Your market lives in chat. Send clicks straight into a concierge conversation, skip the landing page entirely.
Trend front-running
FX rates, weather, events and competitor moves feed the engine. When signals stack on a category that converts, a campaign is drafted before the spike peaks.
Customer UGC army
Your best-converting format is a real customer talking. Generate one testimonial per region, per vertical, per objection, no film crew.
True-conversion feedback loop
Feed closed customers and their revenue back to the ad platforms as the conversion event, so the bidding algorithms chase buyers instead of form fills.
The brain drafts angles from the attribute learnings. Claims to press, hooks to retry, formats to skip.
▸ drafting gen-8 angle fileVariants inside brand rules. Statics, motion, avatar testimonial videos. A batch in an afternoon.
▸ 12 variants renderingSmall batches, one variable each, capped spend. Clean reads without upsetting the algorithm.
▸ 6 in market · €40 cap eachOvernight, on real customers. Winners scale, losers die fast, fatigue gets caught early.
▸ 2 scaled · 2 killed tonightEvery verdict is decomposed into attributes. The lift table updates, the brain gets smarter.
▸ 3 attribute updates writtenEvery learning is an attribute with a sample size. The lift is measured on closed customers, so "videos work" becomes a number.
Day 13 of 14. CTR is off its peak, the gen-7 successors are already in test. The handover happens before the decay hurts.
Every render inherits the learning library, price-difference hooks + problem-first openers are already in the prompt. Nothing goes live from here, assets land on the platforms as paused drafts.
Testing 6 variants…
GTM Outreach
The outbound cockpit. The engine builds its own cold lists, wakes up the leads already in your CRM and runs everyone through one multichannel sequence, email, LinkedIn and WhatsApp, through rails like Instantly and HeyReach. A human still sends every message that matters.
~150 new companies a day discovered from intent signals, enriched, verified and scored before a single send. Cold fills most of the daily sending capacity.
Conference contacts and old opportunities get watched for fresh activity. The cheapest pipeline you own is the list you already paid to build.
The rule under the score: missing data never zeroes a lead. Score what you can measure, report a confidence percentage next to it.
~7,600 one-off researched emails. Heavy cost, flat reply rate. The homework nobody asked for.
3 variants per block, {{keyword}} + {{company}} injected, follow-up does the math for them. Research moved downstream, to repliers only.
Personalization only counts when it touches the problem you solve. The rest is an attention hack. Buyers smell it.
The test for every tool on this wall: does it own data or access you can’t rebuild? Then it’s a rail, wire it in. The intelligence layer above the rails is the part you own.
How the sending machine is built
| SENDER DOMAIN | BOXES | WARMUP | SENDS/DAY | REPLY | BOUNCE | |
|---|---|---|---|---|---|---|
| get-acme.com | 6 | 420 | 14.2% | 0.3% | ||
| try-acme.io | 6 | 390 | 12.8% | 0.4% | ||
| acme-hq.com | 4 | 310 | 9.1% | 0.6% | ||
| meet-acme.co | 4 | 180 | 7.4% | 0.5% | ||
| acme-team.net | 4 | 150 | warming | 0.2% |
24 mailboxes across 5 domains, rotated automatically. Spam-rate ceiling 0.1%, one breach pulls the domain for 14 days. The engine slows itself down before any platform has to.
What goes out · today's 1,450
Met at a conference in May, quiet for 90 days. This week he opened the integration guide twice. The lead woke itself up.
Re-open with the pain he named himself at the booth, AI agents reasoning over stale project docs.
A 60-engineer platform team cut new-hire ramp-up 40% once project knowledge stayed current on its own.
Engineer to engineer. Short, concrete, zero marketing.
Email first, he answers email. The LinkedIn touch waits 48h and only fires if this thread stays quiet.
Subject
The quiet win is the "no" row. Suppression is honored everywhere, so nobody gets step 4 after they said stop.
| ACCOUNT | ANGLE | OWNER | STATE |
|---|---|---|---|
| Helios Retail Group | champion re-engage · ROI one-pager | Sarah | READY |
| Nordwind Logistics | new CMO · first-90-days opener | Sarah | READY |
| Cobalt Systems | renewal window · migration path | Daniel | IN REVIEW |
| Vanta Retail | webinar trio · working session | Daniel | IN REVIEW |
| Meridian Foods | trial activation nudge | Ana | SENT 09:12 |
GTM Sales
Every call heard, every deal reviewed, every proposal priced inside the guardrails. The engine coaches the humans and keeps the deals honest. Reps sell, the machine carries the homework.
Transcripts land within the hour. Sentiment and moment tags are model output, wrong sometimes, the audio link is one click away when a tag looks off.
Send the recap
Call intelligence · this morning's cards
| DEAL | STAGE | SIZE | CLOSES | DRIFT | OWNER |
|---|
Flags are hypotheses with evidence attached, expect a few percent of nonsense. A rep can dismiss any flag with one line of why, dismissals feed the Friday rules review.
Deal review · needs a decision today
TL;DR
The manager stays the coach. The engine brings the receipts, picks the clips and keeps the score, the actual conversation on Friday is human to human.
Battle cards · auto-updated
What changed this week · and from which call
Edits from calls wait 24h in a review queue before going live. One misheard quote on a card costs more than a day of delay.
Proposal workbench
Pricing recommender
Anchored on 3 comparable closed deals in the segment. 8% discount available inside guardrails IF traded for the case-study right. Below €7,900 this deal needs the CRO's yes, the engine will route it there automatically. CONFIDENCE 86%
The recommender advises, the floor decides. In 6 months it has been overruled 14 times, 11 of those overrules closed. Both numbers stay on this page.
GTM Lifecycle
The CRM that runs itself. Every contact sits in a journey, every journey watches behaviour, the right message leaves at the right moment. Templates get approved once by a human, then the machine holds the rhythm around the clock.
Every journey keeps a 10% holdout forever. Lift you can defend beats lift you can claim.
| COHORT | ACCOUNTS | WAVE STATUS | RECOVERED |
|---|---|---|---|
| Churned April | 38 | WAVE 1 SENT | 4 |
| Churned May | 29 | WAVE 2 DRAFTED | · |
| Churned June | 33 | WAITING FOR DAY 90 | · |
| Older cohorts | 104 | QUARTERLY TOUCH | 7 |
Win-back economics get audited too. A recovered account that churns again in 90 days counts against the wave.
Cancel reasons come from offboarding calls plus the exit survey. Where the reason is unknown the draft says less, guessing in a winback email reads worse than silence.
+9% opens carried through to +4% activation on 840 signups. Rolled to 100% on day 22.
Converted +11% short term, the saved accounts churned at 2.1x by month 3. Kill criterion caught it at day 45.
Day 12 of 30 · booking rate 1.4x control so far, sample still thin. Kill line sits at 1.2x.
Day 5 of 21 · early invites +23%, too early to celebrate. Kill line sits at +15%.
No separation after 200 sends per arm. Closed at the pre-agreed sample, plain stays.
GTM SEO · GEO
One factory, two fronts. Pages that rank in search, answers that get cited by the AI engines your buyers now ask first.
Keyword research lives inside this engine, demand mapping is step one of the same loop.
And some pages ship with zero search volume on purpose. They exist for sales enablement and AI citations, two jobs the volume number never sees.
| PAGE | TYPE | STAGE |
|---|---|---|
| /best-demand-forecasting-tools-2026/ | listicle · MOFU | DRAFTING |
| /compare/meridian-vs-legacycore/ | comparison · BOFU | DRAFTING |
| /alternatives/spreadsheet-forecasting/ | alternatives · BOFU | QA GATE |
| /glossary/net-revenue-retention/ | glossary · TOFU | QA GATE |
| /tools/roi-calculator/ | free tool | PUBLISHED |
| /state-of-b2b-buying-2026/ | original research | PUBLISHED |
Every money page gets two touches a month, no exceptions. A page that misses two slots in a row gets escalated to the fix queue.
One prompt fans out into ~5 hidden queries per run, so the engine tracks whole narratives.
| GPTBot | 214 fetches | reading the comparison pages |
| PerplexityBot | 96 fetches | deep on the research report |
| ClaudeBot | 122 fetches | glossary + pricing pages |
Listicles draw about 44% of AI citations. Owning ours is half the play, appearing in everyone else's is the other half.
84 to 90% of citations point off-domain. Off-domain is where authority lives, so the engine earns it deliberately instead of waiting for luck.
The factory queue · this week
| CLUSTER | PAGES | STAGE | NEXT |
|---|---|---|---|
| pricing comparisons | 24 | QA GATE | your approve |
| migration guides | 12 | PUBLISHING | 6h rollout |
| glossary · long tail | 48 | DRAFTING | QA Fri |
| listicles · GEO bait | 8 | INDEXED | citation sweep 14:00 |
| alternatives pages | 16 | REFRESH DUE | 2x monthly cycle |
Citation share · by engine
Listicles earn ≈44% of all citations, comparisons 27%. The gap engine drafts a fix for every lost prompt, fed straight back into the factory queue on the left.
Keyword Research
Where demand actually lives. Search Console tells the engine what you already own, the rank tools tell it what you are missing. The factory builds from both.
| QUERY | CLICKS | IMPR | CTR | POS |
|---|---|---|---|---|
| demand forecasting software | 4,812 | 88.4k | 5.4% | 2.1 |
| meridian vs legacycore | 2,904 | 31.0k | 9.4% | 1.2 |
| net revenue retention formula | 2,211 | 64.7k | 3.4% | 3.8 |
| roi calculator forecasting | 1,688 | 22.9k | 7.4% | 1.9 |
| best forecasting tools 2026 | 1,342 | 58.1k | 2.3% | 6.7 |
| CLUSTER | INTENT | VOL/MO | STATUS |
|---|---|---|---|
| brand + comparisons | BOFU | 12.4k | OWNED · #1 |
| category head terms | MOFU | 33.1k | RANKING · P4 |
| alternatives to incumbents | BOFU | 6.2k | RANKING · P3 |
| glossary + definitions | TOFU | 41.8k | OWNED · #1-3 |
| industry x use case | MOFU | 18.9k | GAP · QUEUED |
| pricing + cost questions | BOFU | 9.7k | GAP · QUEUED |
TOFU brings readers, MOFU brings evaluators, BOFU brings buyers. The factory builds BOFU gaps first, the money pages.
Prediction is a bet. The engine drafts outlines early, the factory only commits pages once a predicted query shows real impressions at striking distance.
Reporting
Numbers reconciled against the warehouse every morning. The report drafts itself, a human signs it before anyone else sees it.
The CFO asked for CAC by segment again, mid-market versus enterprise. Second time this quarter.
Noted. CAC by segment added as a standing section, every future pack carries it. First appearance in the August draft.
The manual version of this took the team 2 to 3 days a week. Pull the analytics tool into a spreadsheet, pull each ad platform into the same spreadsheet, reconcile by hand, rebuild the deck. The engine does the pulling and the reconciling. People keep the judgment calls.
Every number traces to a warehouse query. Click any figure in the real product to see its source.
Chat with the brain
Ask in plain language. Answers come from your warehouse and your CRM, with the sources attached. If the brain does not know, it says so.
Every answer cites its query. The brain reads from read-only mirrors, it cannot touch production.
Second Brain
Everything the company knows, in one place the agents can read. dozens of sources across four tiers, clean data underneath, shared definitions on top, know-how beside it.
Day one, every table counted against its source. Incomplete mirrors make confident dashboards wrong, so this runs before anything scores. The two amber tables carry a backfill job and the engines ignore them until they clear.
The tiles above are the live feeds with their numbers. The map is the point, one brain reads all of it, so a funding round, a pricing visit + a rival's job post become one picture of one account.
Every agent and every report uses the same math. When "customer" means one thing, the numbers stop arguing with each other.
The ads engine bids on CUSTOMER. The old setup bid on LEAD. That one change explains most of the CAC story.
Data Explorer
The modeled tables under every engine. Read-only mirrors, production stays untouched.
from ad_spend_daily join campaign_customers using (campaign_id)
where closed_at >= date_trunc('quarter', now())
group by 1 order by 2;
| CHANNEL | REAL_CAC |
|---|
This is the exact query behind the GTM Ads reallocation card. Every number in the OS traces back to a table you can open here.
CRM Cleanup
Duplicates, collisions and dead records, found nightly, merged only after a human says yes. Every ruling you make teaches the engine your rules.
Cleanup jobs
refresh the read-only copy from the CRM
rebuild the shared-signal pair analysis
company dedup keyed on the VAT registry id
the evening batch, fully reversible for 30 days
Collision pairs · found last night
| SIGNAL SHARED | RECORDS | ENGINE'S READ | |
|---|---|---|---|
| +31 6 ••• 4471 | 2 contacts, same surname | LIKELY SAME PERSON | merge queue |
| +31 20 ••• 900 | 6 contacts, one company | OFFICE SWITCHBOARD | keep all |
| ops@vanta-retail.com | 2 contacts, names differ | UNCLEAR · ASK A HUMAN | review |
| VAT NL8522••••B01 | 3 companies, 1 parent | BRANCHES → ROLL UP | merge queue |
| +49 30 ••• 118 | 2 contacts, colleagues | SHARED LINE | keep all |
Numbers on more than 5 records count as switchboards and are excluded automatically. Companies merge on the VAT registry id, never on website, marketplaces share those.
Merge rehearsal · 1 of 41 queued
CONFIDENCE 96%Same phone once formats are canonicalized, same surname, the gmail is the older personal address. The survivor keeps the work email, the gmail moves to the secondary field, activity history merges.
MCP Servers
The sockets the agents plug into. Each connection is scoped, logged and read-only unless a human approval sits in front of the write.
Reads are free, writes wait for a yes. Every tool call lands in the audit log with who asked, what ran and what came back. Access Control decides which agents see which sockets.
Access Control
Who sees what, who may approve what. This is the part that turns a clever demo into something a real company can run. Regulated data gets its own guardrail, a one-click option at the connection level, not a bespoke integration.
| CAPABILITY | REP | LEAD | OPERATOR | ENGINEER | EXEC |
|---|---|---|---|---|---|
| See scores · own accounts | ✓ | ✓ | ✓ | ✓ | ✓ |
| See scores · all accounts | · | ✓ | ✓ | ✓ | ✓ |
| Send outreach drafts | ✓ | ✓ | · | · | · |
| Approve budget moves | · | ≤ €500/d | ✓ | · | · |
| Approve page batches | · | · | ✓ | ✓ | · |
| Edit playbooks + prompts | · | · | ✓ | ✓ | · |
| Change connections + scopes | · | · | · | ✓ | · |
| Read every report | · | ✓ | ✓ | ✓ | ✓ |
The outreach agent has scopes: crm.read, web.read, email.draft. Billing is structurally out of reach. Push on it.
Crazy Ideas
The brain reads everything you feed it, then asks what if. Ideas beyond optimization, each one cheap to test, each one with a kill switch agreed before a euro moves.
Concierge chat funnel
Send paid clicks straight into a live concierge conversation. Skip the form, skip the landing page, qualify in the thread.
Trend front-running
Watch FX, weather, events and competitor moves. When signals stack on a category that converts, draft the campaign before the wave peaks.
Renewal-window sniper
A dedicated always-on motion for switching-window accounts. Migration-path content, peer proof, no discounts until legal is in the room.
The wildcard campaign
Mix the audiences, break the structure the playbook says you need, let the algorithm find a pocket nobody targeted on purpose.
Customer-as-creator army
Generate one testimonial per region, per vertical, per objection. Real customer stories, no film crew, refreshed before fatigue hits.
Free-tool magnet blitz
Ship a small interactive tool per buying question (pricing, sizing, benchmarks). Tools get linked, cited and remembered long after posts fade.
Own the "State of X" report
Publish the definitive annual benchmark from your own anonymized data. Every stat becomes a quotable line the AI engines repeat with your name on it.
Conference-week surge
A geo-targeted, time-boxed burst around each event. Custom pages, outreach angles and ads that only exist for those ten days.
Send prospects the forecast
Push the demand forecast TO your prospects. "Your category spikes in 5 weeks, here is the data." Be the vendor who told them before it happened.
I Want More Plays
How the OS grows. Tell it your motion, where you are and what you want to move, the brain turns 145 proven plays into a sequenced 90-day build plan sized to the hours you actually have. Every plan starts in dry mode and everything below is a suggestion, you decide what actually runs. Or scroll down, describe what you wish existed in one sentence, the brain drafts the spec on the spot.
Most teams run one or two. Pick what is true today.
The plan is sized to the hours you actually have, an honest number beats an ambitious one.
"I want to know which customers are about to churn." That sentence is the whole request.
Data needed, where it lives, who approves the outputs, build estimate. If the data is missing, it says so before anyone builds anything.
Two weeks in the sandbox, behind the same human gate as everything else. Used = hardened. Ignored = retired, cheaply.
Type your own wish above or
| APP | ASKED BY | VOTES | STATUS |
|---|---|---|---|
| Churn early-warning | Customer success | BUILDING | |
| Quote configurator | Sales | SHIPPED | |
| Partner co-sell finder | Partnerships | SCOPING | |
| Competitor price watch | Product marketing | BACKLOG | |
| Event lead capture | Field marketing | BACKLOG | |
| Onboarding health score | Customer success | BACKLOG |
The quote configurator was asked for on a Tuesday, shipped two Fridays later. That story is why this board fills up.
Every app on this board is the same recipe. An operator who owns the outcome, the shared brain underneath so nothing gets rebuilt, one embedded engineer who ships. The apps multiply because the foundation is already paid for. App number twelve costs a fraction of app number one.
Monitoring
The boring dashboard that keeps the exciting ones honest. Tokens, cost, drift, anomalies and guardrails, watched around the clock, escalated only when a human is actually needed.
| SOURCE | LAST INCREMENTAL | LAST FULL SYNC | ROWS LAST RUN | STATUS | |
|---|---|---|---|---|---|
| CRM mirror | 06:00 today | Sun 02:00 | 18,412 | HEALTHY | |
| Product analytics | 06:05 today | Sun 02:20 | 241,038 | HEALTHY | |
| Ads platforms | 06:10 today | Sun 02:40 | 9,284 | HEALTHY | |
| Search console | 06:00 today | Sun 03:00 | 3,412 | HEALTHY | |
| Call transcripts | 05:45 today | Sun 03:15 | 86 | BACKFILL QUEUED | |
| Warehouse | 06:15 today | Sun 03:30 | 312,940 | HEALTHY |
when a seller hears nothing for a week they assume something broke, this page is the answer.
| ENGINE | TOKENS/DAY | €/DAY | PER OUTCOME |
|---|---|---|---|
| GTM Score | 6.2M | €3.70 | €0.003 / account |
| GTM Ads | 4.8M | €3.40 | €0.05 / move drafted |
| GTM Outreach | 3.1M | €2.40 | €0.02 / draft |
| SEO · GEO factory | 14.6M | €9.30 | €0.13 / page |
| Chat + reporting | 7.4M | €4.20 | €0.01 / answer |
| Cleanup + enrichment | 5.1M | €2.50 | €0.005 / suggestion |
Roughly €26 a day, €782 a month all-in, runs the whole room. One drafted page costs €0.13 to produce, the humans it frees are the real line item.
The 8 Jul spike is the throttle earning its keep, a crawl loop capped at +€38 instead of +€800.
| MODEL | LIVE | EVAL | FALLBACK |
|---|---|---|---|
| Scoring | v14 | 98.6% | v13 · 1 click |
| Next-move | v9 | 97.8% | v8 · 1 click |
| Creative judge | v7 | 94.1% | retrain Thu |
| Cleanup matcher | v11 | 99.0% | v10 · 1 click |
Drift past the band schedules a retrain on its own. Nobody has to notice the model got stale, the monitor notices.
Under the Hood
The whole machine on one page. Data flows down, decisions flow through people, outcomes flow back and make everything smarter. Click any block to open it.
CRM data
First-party
Second-party
Third-party
The database
The Shared Brain
one place everything becomes usable · the part 80% reusable across every new capabilityPerformance
Every paid channel on one desk, judged against closed revenue in your warehouse. The engine moves budget nightly toward the cheapest real customer. You approve every move.
What the platforms say vs what your warehouse says
| CHANNEL | PLATFORM SAYS | WAREHOUSE SAYS | WHY THE GAP | THE MOVE |
|---|---|---|---|---|
| Search · brand | €78 CPA | €81 real CAC | honest. clicks become customers | SCALING |
| Search · non-brand | €96 CPA | €2,410 real CAC | brand halo credited to generic terms | REBUILT · RETEST |
| Paid social · prospecting | €54 CPA | €510 real CAC | optimizes to leads that rarely close | NEW BID SIGNAL |
| LinkedIn · ABM | €210 CPA | €390 real CAC | pricey clicks, real pipeline behind them | SCALE TO ICP LIST |
| Retargeting | €12 CPA | €3,800 real CAC | was re-buying people who already bought | CUT 80% |
Your three decisions this morning
Where the next €1,000/day goes
Proof, on purpose
Agents
The digital teammates that run the engines. Each one has a job description, a skill set, guardrails and a cost. Hiring a new one takes weeks, so does firing one. That part we made easy.
The roster
| AGENT | JOB | ENGINE | RUNS TODAY | COST/DAY | APPROVED | STATUS |
|---|---|---|---|---|---|---|
| The Conductor | routes every job to the right model | platform | 1,410 | €11 | · | |
| Scorekeeper | rescores every account nightly | GTM Score | 3,120 | €9 | 98% | |
| Bid Captain | drafts nightly budget moves | Performance | 96 | €6 | 94% | |
| Creative Director | briefs + generates ad creatives | GTM Ads | 48 | €7 | 91% | |
| Creative Judge | pre-scores creatives, kills losers | GTM Ads | 48 | €3 | 96% | |
| Outreach Drafter | drafts outreach worth sending | GTM Outreach | 210 | €8 | 89% | |
| Reply Learner | learns from every reply | GTM Outreach | 140 | €2 | · | |
| Page Factory | builds pages behind a quality gate | SEO · GEO | 31 | €5 | 93% | |
| Citation Tracker | tracks who the AI engines cite | SEO · GEO | 120 | €3 | · | |
| Reconciler | reconciles + drafts reporting | Reporting | 18 | €2 | 99% | |
| CRM Janitor | keeps the CRM clean | platform | 260 | €2 | 97% | |
| Watchdog | monitors the other agents | platform | continuous | €2 | · | |
| Spec Writer | turns requests into build specs | expand | 6 | €1 | 95% | |
| Researcher | deep-dives accounts + markets | shared | 44 | €4 | · | |
| Call Coach | scores calls, writes coaching cards | GTM Sales | 61 | €5 | 92% | |
| Deal Reviewer | nightly drift + compliance sweep | GTM Sales | 214 | €4 | 96% | |
| Battle Card Keeper | auto-updates competitive cards | GTM Sales | 12 | €2 | · | |
| Proposal Drafter | CRM record → priced draft | GTM Sales | 9 | €2 | 95% | |
| Media Planner | cross-platform budget planning | GTM Ads | 33 | €6 | 93% | |
| Asset Builder | briefs → assets in every format | GTM Ads | 58 | €7 | 90% | |
| Localizer | drafts in the buyer's language | GTM Outreach | 120 | €3 | · | |
| Deliverability Sentinel | guards the sending fleet | GTM Outreach | continuous | €2 | · | |
| + 14 more on the payroll · enrichers, refreshers, summarizers, escalators, the plumbing crew | ||||||
1 · Pick the job
3 · The ship plan
Skills
The company's know-how, written down once and packaged so any agent can run it. When your best marketer's method becomes a skill, it stops leaving in their notice period.
Scores one account from product, CRM and web signals, with the reasoning attached.
Joins any channel's spend to closed revenue in the warehouse. The truth serum behind Performance.
Writes a first touch like someone who did two hours of homework. Carries the reply-pattern library.
Scores a creative against every attribute the account has ever learned, before a euro touches it.
Turns a keyword cluster into a page that survives the quality gate. Refuses to ship thin.
Asks the AI engines a buyer question, records who gets cited and why. 120 questions a week.
Checks every reported figure against the warehouse before a human ever sees it.
Finds and merges duplicates with the confidence score attached. Learns from every human call.
Turns a plain-language request into a build spec with scope, tables and guardrails.
Sweeps any outbound text against the brand voice rules. Nothing off-voice leaves the building.
Checks every experiment against its agreed stop conditions. Pulls the plug without a meeting.
Scores a sales call on the mechanics that win. The card goes to the rep before anyone else.
Checks one deal against the signed deal rules. Flags drift with evidence attached.
Prices a deal from comparables + the price book, inside guardrails. Routes exceptions up.
Splits budget across Google, Meta, Reddit and TikTok from marginal CAC. The chat runs on it.
Turns an approved brief into every format via Higgsfield, Arcade and AdCreative.
Rewrites any outbound text in the buyer's language, tone preserved, flags checked by a native reviewer weekly.