Your agent can call ax1om · scoring is live over MCP, free tier included. Start free Already have an account?

Press kit

Press and research kit

Everything needed to describe ax1om accurately, in one place. The descriptions below are the canonical ones. They are the same text published on our company and directory listings, so quoting them verbatim is the correct move, not a lazy one.

Canonical description Logos and screenshots
01 / Boilerplate

Three lengths of the same description

Pick the one that fits the space. All three say the same thing, so mixing them across a piece stays consistent.

One-liner Tagline and directory subtitle fields.

Trained prediction models your AI agents can call, built only on your own CRM data, with the reason behind every prediction.

Short description Directory summaries, article standfirsts, panel bios.

ax1om is an analytics engine for AI agents, starting with predictive scoring for RevOps teams. It connects to Salesforce or HubSpot, trains machine learning models on your own conversion history - first-party data only, no shared models - validates each one before anything acts on it, and serves conversion likelihood scores with the SHAP factors behind each one: to the CRM, the scoring API, and AI agents over MCP.

Long description Long-form directory profiles and background sections.

ax1om trains prediction models on each customer's own outcome history and exposes them for AI agents to call over MCP and API. Three model families run in production - conversion scoring, customer health, and timing - with scoring live on the MCP surface today. Every model trains on first-party CRM data from Salesforce or HubSpot, with no third-party enrichment and no shared models, and passes validation gates, including leakage detection and label-maturity checks, before anything acts on it. Every prediction ships with SHAP factors showing which CRM fields drove it. Scores write back to Salesforce and HubSpot, and a versioned scoring API serves the same predictions to workflows. Built for RevOps teams and operators putting agents on revenue work, with no data science hire required.

A note on the brand name, because it is the thing that gets corrected most: it is ax1om, lowercase, with the digit one in the middle, including at the start of a sentence and in headlines. Not Ax1om, not AX1OM, not axiom.

02 / Fact sheet

The checkable facts

Company record first, product record second. Anything not listed here is not something we have published.

Company

Legal name
ax1om LLC
Brand name
ax1om, always lowercase, including at the start of a sentence
Founded
2026-03-29, the date the Texas LLC was filed and effective
Headquarters
Texas, United States
Founder
Luis Esquivel, founder and CEO
Website
ax1om.ai, with the product at app.ax1om.ai

Product

Category
Analytics engine for AI agents - trains, validates, and explains models built on each customer's own outcomes · predictive lead, account, and opportunity scoring, the first use case · trained model families for conversion scoring, customer health, and timing.
Who it is for
Operators putting AI agents to work on revenue - RevOps and GTM leaders and technical operators who need the numbers their agents act on to be real - plus RevOps, marketing operations, and GTM teams that own prioritization and do not have a data science function to build it.
Model
One dedicated model per customer. No model is shared across customers and no customer data trains anything outside that tenant.
Algorithm
LightGBM, gradient-boosted decision trees, with SHAP for per-record explanations.
Training data
First-party CRM data only, converted and non-converted records alike. No third-party intent data and no purchased enrichment.
Explainability
SHAP feature importances on every field, per-record top factors on every score, and a Feature Stability Score reporting whether an explanation survives retraining.
Outputs
A score for a lead, an account, or an opportunity. The score is a 0-100 ranking of conversion likelihood, not a percentage chance of conversion, and the per-record factors do not sum to it.
Activation surfaces
Salesforce writeback · HubSpot writeback · CSV export · REST scoring API · scheduled rescoring · MCP server for AI agents
03 / Founder

Luis Esquivel

Luis Esquivel is the founder and CEO of ax1om LLC. He started in marketing operations at a B2B SaaS company and spent more than ten years in marketing and GTM operations roles across private equity backed and publicly traded B2B SaaS companies before starting ax1om.

ax1om is the scoring and insights platform he wanted at those companies: trained on first-party data, with a reason attached to every score, writing back to the CRM where reps already work.

Title for citation: Luis Esquivel, founder and CEO, ax1om LLC

04 / Assets

Logos

Vector where a vector exists. Keep the clear space around the mark, do not recolor it, and do not stretch it. The light and dark labels name the background the file is meant to sit on.

Lockup

Glyph and wordmark together. The default choice.

Wordmark

Type only, for places that already show the glyph.

Glyph

The square mark on its own, for avatars and favicons.

Square

Pre-cropped social and directory profile image, 400px.

The full brand directory is served at /brand/. Only the files listed above are current; anything else in that directory is a working variant and may change.

05 / Screenshots

Product screenshots

Real views from the product, captured on the sample CRM dataset that ships with every account. They are not customer data and they are not a result we are promising. If you publish one, the caption should say sample data, the way ours does.

The model performance view in the ax1om console: a row of four metric cards labelled AUC-ROC, Lift at top 20 percent, Precision at top 20 percent, and Records trained, each with a link to its explainer; a Key insights panel below them listing plain-language findings about which job titles convert above or below the baseline rate; and two charts side by side, a conversion-by-score-bucket bar chart with a dashed baseline reference line running from highest to lowest scores, and a score distribution histogram banded into hot, warm, and cold
The model performance view in the ax1om console: a row of four metric cards labelled AUC-ROC, Lift at top 20 percent, Precision at top 20 percent, and Records trained, each with a link to its explainer; a Key insights panel below them listing plain-language findings about which job titles convert above or below the baseline rate; and two charts side by side, a conversion-by-score-bucket bar chart with a dashed baseline reference line running from highest to lowest scores, and a score distribution histogram banded into hot, warm, and cold
Model performance, from a training run on sample data
The Feature analysis table in the ax1om console, subtitled that fields are rolled up from individual feature values and that clicking a row expands it into those values, with columns for rank, field, SHAP impact, splits, FSS, and direction; the ranked rows name CRM fields such as LeadSource, Title, NumberOfEmployees, Status, and Industry, each showing an impact bar, a splits bar, a stability badge, and an up, down, or mixed direction arrow, above a footnote distinguishing how often the model uses a field from how much that field actually moves the score
The Feature analysis table in the ax1om console, subtitled that fields are rolled up from individual feature values and that clicking a row expands it into those values, with columns for rank, field, SHAP impact, splits, FSS, and direction; the ranked rows name CRM fields such as LeadSource, Title, NumberOfEmployees, Status, and Industry, each showing an impact bar, a splits bar, a stability badge, and an up, down, or mixed direction arrow, above a footnote distinguishing how often the model uses a field from how much that field actually moves the score
Feature analysis · on a sample-data workspace
The first screen of score setup in the ax1om console, headed Create a Score, offering four outcome cards labelled Lead conversion, Won business, At-risk customers, and Renewal / expansion, each naming the question it answers and the records it trains on, above a collapsed More starting points section and a Start from scratch option
The first screen of score setup in the ax1om console, headed Create a Score, offering four outcome cards labelled Lead conversion, Won business, At-risk customers, and Renewal / expansion, each naming the question it answers and the records it trains on, above a collapsed More starting points section and a Start from scratch option
Score setup, first screen · on a sample-data workspace
06 / Substance

Where the detail lives

This page is the summary. These four are the primary sources, and they are the right things to cite.

Model performance figures

AUC, lift at the top of the list, precision, and conversion by score bucket, from a training run on the sample CRM that ships with every account. Real dashboard, sample data, and labelled as such.

Trust center

Security posture, subprocessors, data residency, and where the compliance work currently stands. The right citation for any security or privacy question.

API reference

The scoring contract: endpoints, auth, field tables, the error envelope, and the versioning policy. Generated from the OpenAPI artifact.

MCP server

How an AI agent reaches the same predictions as tools rather than raw HTTP, so scoring logic never has to live inside the agent.

07 / Disambiguation

Which ax1om this is

ax1om is the B2B predictive lead and account scoring platform at ax1om.ai, operated by ax1om LLC in Texas. The name is spelled with the digit one.

Several unrelated things share or resemble the name, including a browser-automation product at a similar domain and a recording artist publishing under the same spelling. We are not affiliated with either, and this page makes no claim about them beyond that they are separate. When a citation needs to be unambiguous, the domain ax1om.ai or the legal name ax1om LLC is the safe identifier.

Press and research enquiries

There is no press desk and no agency in between. Send it through the support form or by email and it reaches the founder. A human reads every message and replies within one business day.

Send a message [email protected]

Fact-checking a specific number before you publish is welcome and encouraged. If the figure is not on this page or in one of the four sources above, ask rather than infer it.