When every signup runs inference on your dime, growth and gross margin pull against each other. The acquisition motion that works for a zero-marginal-cost SaaS tool quietly bankrupts an AI company. You need a model that brings in users who convert before they burn through your compute budget.
The Problem
Every free user runs up a real GPU bill
Classic SaaS acquisition assumes near-zero marginal cost per user, so the playbook is to acquire as many as possible and monetize later. AI products break that assumption – every active user runs inference that costs real money, and a generous free tier can lose more on compute than it ever recovers. A viral spike that would be pure upside for a normal SaaS app can be a margin emergency for an AI company. Acquisition strategy has to account for unit economics that most growth playbooks were never designed to handle.
You have to sell to the engineer and the executive at once
AI buying decisions split across a technical evaluator who tests the model and an economic buyer who signs the contract. The engineer cares about latency, accuracy, and whether it breaks on their edge cases; the executive cares about cost, risk, and whether this is a defensible bet. Acquisition content and funnels built for one of them stall with the other – a developer-led motion that wins the eval still dies if the VP sees no business case. Most AI companies optimize for whichever buyer their founder resembles and leak the other half of every deal.
Trust is the real gate, and demos do not clear it
Buyers are nervous about accuracy, hallucination, data handling, and whether your model does what the landing page claims. A slick demo proves the happy path; it does nothing for the buyer worried about the failure case in their actual workflow. Acquisition that leans on demo magic gets lots of trials and few conversions because the trial is where trust gets tested and lost. The hard part of AI acquisition is not getting attention – it is getting a skeptical buyer to believe the thing works on their data, not yours.
The category shifts faster than your funnel can keep up
A new foundation model, an open-source release, or a competitor's launch can reset buyer expectations overnight, and a base model getting cheaper or commoditized can erase your differentiation between one quarter and the next. An acquisition engine tuned to last quarter's positioning starts converting worse and no one knows why. The pace of change means a funnel is never finished – the message that worked when you were the only option fails the moment a buyer has three credible alternatives. Acquisition for AI has to be rebuilt continuously, not set and left.
How We Help
We start with the unit economics, because for an AI company acquisition strategy and gross margin are the same conversation. The first thing we look at is your real cost to serve – inference cost per active user, how that scales with usage, and where your current funnel brings in users who cost more than they pay. We map that against who actually converts and retains, so we are not optimizing top-of-funnel volume that quietly destroys margin.
From there we build the acquisition strategy around the two-buyer reality. We design the motion so the technical evaluator can prove the model works on their own data and the economic buyer gets a business case in the same flow, rather than picking one and losing the other.
Execution is where we make trust the center of the funnel instead of an afterthought. We help you replace happy-path demo theater with proof a skeptical buyer believes – the ability to test on their data, honest framing of where the model is strong and where it is not, and accuracy and safety signals that survive a procurement questionnaire.
Measurement is built on margin-aware acquisition metrics, not raw signups. We track cost to acquire against cost to serve and lifetime value, so growth that destroys gross margin gets caught before it scales. We watch conversion split by buyer type to see whether the engineer-side and executive-side motions are both working. And we measure how fast the funnel adapts when the category moves, because in AI a static funnel is a decaying one.
The work succeeds when you are acquiring customers who clear their own compute cost, when both buyers in the deal are getting what they need, and when a trial converts because the buyer believes the model – not because the demo dazzled them.
What we deliver
- A margin-aware acquisition model that accounts for inference cost per user
- A two-track motion that serves the technical evaluator and the economic buyer in one flow
- A trust-first funnel: test-on-your-data proof, honest accuracy framing, procurement-ready signals
- Channel and message tuning to the segments that actually convert and retain
- A process for rebuilding the funnel fast when a model or competitor resets the market
For an AI company, customer acquisition is a gross-margin decision disguised as a marketing one. The cheapest user to acquire is often the one who costs you the most to serve – and the funnel has to know the difference.
Our Methodology
Acquisition work for AI companies starts where most growth playbooks never look – at the cost to serve. The first phase audits inference cost per active user, how it scales with usage, which segments convert and retain, and where the current funnel brings in users who lose money. That defines the segments worth chasing and the ones to gate or reprice, so we are never optimizing volume that destroys margin.
The second phase builds the motion around the technical-buyer and economic-buyer split and makes trust the center of the funnel. We design an eval or self-serve path that lets the engineer prove the model on their own data, a parallel track that arms the champion with a business case for the executive, and proof that clears a skeptical procurement review rather than dazzling on the demo. We then tune channels and message to the converting segments and build the capacity to rewrite the funnel quickly when the category moves.
What makes this different from a growth agency is that we treat unit economics, the two-buyer split, and trust as the acquisition problem rather than chasing signup volume and top-of-funnel traffic. An agency optimizes for cost per lead. We optimize for customers who clear their own compute cost, deals where both buyers are satisfied, and a funnel that survives the next model release.
How We Work
Initial engagements typically run three to six months because building a margin-aware acquisition engine means auditing real unit economics, rebuilding the funnel around two buyers, and proving the new motion converts before it scales. The first 30 days are the audit – cost to serve, segment conversion and retention, and where the current funnel leaks or loses money. The next phase rebuilds the strategy and the funnel around margin, the two-buyer split, and trust. The final phase is execution and tuning against live conversion and margin data.
Our team brings a growth strategist who owns the acquisition model and unit economics, working alongside whoever runs your demand gen, your product team for the eval and self-serve motion, and your finance or ops side for the cost-to-serve picture. From your side we need real inference-cost data, conversion and retention by segment, and access to how trials and evals currently run. We cannot fix margin-blind acquisition without the actual cost numbers.
The cadence is weekly working sessions through the build and into execution, with a regular review of acquisition cost against cost to serve and lifetime value so margin-destroying growth gets caught early. Because the category moves fast, we build the rebuild-the-funnel muscle into the engagement rather than treating the funnel as finished. Initial engagements run three to six months, with the option to extend into an ongoing growth partnership as the market shifts.
- Acquisition that brings in customers who clear their own compute cost
- A funnel that converts both the technical buyer and the economic buyer
- Trials that convert on believed proof rather than demo theater
- A funnel that adapts quickly when a new model or competitor resets the market
If your ai / machine learning company needs customer acquisition leadership, we should talk.
Expand your marketing team output with our experts
Let us take a custom approach to your growth goals by assembling and leading the best-in-class marketing team to support your next stage.
Frequently asked questions
How much does a customer acquisition engagement cost for an AI company?
Initial engagements typically run in the $20K-$45K per month range depending on scope, channel breadth, and how much of the funnel needs rebuilding. That is well below the loaded cost of a full-time VP of Growth and gets you an operator who has built acquisition engines before rather than someone learning on your budget.
How long before we see results from an acquisition engagement?
Early signals – leaks plugged, the two-buyer motion taking shape, trust proof replacing demo theater – usually appear inside the first 60 days. Margin-aware acquisition results, where cost to acquire and cost to serve are both moving in the right direction, typically show across a full quarter as the rebuilt funnel runs at volume.
How does the acquisition team integrate with our product and finance teams?
We work with product on the eval and self-serve motion that earns the technical buyer's trust, and with finance or ops on the cost-to-serve numbers that make acquisition margin-aware. The engineer-facing proof and the executive-facing business case both depend on those teams, so we run reviews together rather than working in a marketing silo.
What makes Winston Francois different from a typical growth agency?
A typical agency optimizes cost per lead and signup volume, which for an AI company can actively destroy gross margin by acquiring users who cost more to serve than they pay. We treat unit economics, the technical-buyer and economic-buyer split, and trust as the core of the acquisition problem.
How do you measure ROI from a customer acquisition engagement?
We measure acquisition cost against cost to serve and lifetime value, so growth that erodes gross margin is caught before it scales rather than celebrated as a signup win. We track conversion split by buyer type to confirm both the engineer-side and executive-side motions are working.
What type of AI company is the right fit for this service?
Companies between Series A and growth stage, with a product in market and real inference costs, where acquisition and gross margin are pulling against each other get the most value. If you are seeing strong signup numbers but weak conversion, or growth that is not improving the bottom line, that gap is usually a margin-blind funnel and a two-buyer mismatch.
