AI / ML companies bill on consumption that swings week to week, but the CRM still tracks revenue like a flat SaaS contract. We close the gap between what customers actually use and what the board thinks you booked, so the forecast survives contact with reality.
The Problem
Usage revenue and CRM bookings tell two different stories
A customer signs a $200K commitment, then burns through tokens at twice the rate they projected, or half. The CRM shows the booking. The billing system shows the consumption. The two numbers never match, and nobody owns the reconciliation. So your reported ARR is a guess dressed up as a fact, and finance and sales argue about whose number is right instead of fixing the pipe between them.
The forecast cannot survive a consumption swing
When revenue is driven by how much a customer runs through the product, a single large account changing its usage can blow your quarter up or down by double digits. The forecast was built on contract values and close dates, which means it has no idea this is coming. You find out from the billing dashboard after the fact, not from a model that saw the trend building.
PLG signups and sales-led deals share one broken funnel
Engineers swipe a card, start consuming, and expand on their own while your sales team works enterprise deals through the same pipeline. A self-serve account that is about to become a six-figure logo looks identical to a tire-kicker in the CRM. There is no product-qualified lead definition, so the signal that should trigger a sales motion gets lost in a queue nobody watches.
Product usage data never reaches the CRM
The data that actually predicts revenue – API calls, seats activated, tokens consumed, rate limits hit – lives in a product database that the CRM cannot see. Sales works blind on renewals because they have no view of whether an account is ramping or churning. Expansion and contraction are invisible until they show up in the invoice, which is the worst possible time to learn about them.
How We Help
We start by tracing the money, not by pitching a tool. For an AI / ML company that means following revenue from the moment a customer consumes the product to the moment it lands in your reported ARR. We map the billing system, the CRM, the product usage tables, and every sync and spreadsheet in between. We document where the consumption number and the bookings number diverge, and why. You cannot reconcile two figures until you can see exactly where they split apart.
From that map we build the reconciliation layer that should have existed from the start. Usage-based revenue gets tied back to the commitment in the CRM, so a customer's actual consumption sits next to what they signed for. Overage, drawdown, and ramp become visible as they happen instead of at invoice time. When the board asks what your real ARR is, the answer is one number with a defined methodology behind it, not three dashboards that disagree.
Then we rebuild the forecast so it respects how your revenue actually behaves. A consumption business cannot forecast on close dates alone, because the deal closing is only the start of the revenue, not the size of it. We blend committed contract value with usage trend – run rate, ramp curves, and the accounts whose consumption is accelerating or falling off. The forecast becomes something you can defend in a board meeting because every assumption is written down and tied to a signal, not a sales rep's optimism.
We fix the PLG-to-sales handoff next. We define what a product-qualified lead actually is for your business – the usage threshold, the account shape, the behavior that says this self-serve user is ready for a human. Then we wire the trigger so that when an account crosses it, the right rep gets the account with the usage context attached, not a bare email. Self-serve expansion keeps running on its own, and the accounts worth a sales motion stop slipping through.
Underneath all of it, we unify product usage data with the CRM so sales and finance work from the same picture. Consumption, seat activation, and account health flow into the CRM on a schedule, with alerts when an account's usage drops in a way that predicts churn. Renewals stop being a surprise. Expansion gets caught while there is still time to act on it.
The Winston Francois difference is that we are operators who have run revenue inside companies that grew the way yours is growing, not an agency renting you hours. We build the systems, train your team to run them, and leave documentation so the work survives your next finance or ops hire. You own everything we build.
And we know when to stop. Not every AI / ML company needs a full reconciliation platform on day one. Sometimes the fix is one well-defined usage feed into the CRM and a forecast model that finally accounts for consumption, and we say so even when it shrinks the engagement.
What we deliver
- Revenue data-flow map – a documented trace from product consumption through billing to reported ARR, including every place the numbers diverge
- Usage-to-bookings reconciliation layer – actual consumption tied to CRM commitments so overage, drawdown, and ramp are visible as they happen
- Consumption-aware forecast model – committed value blended with usage run rate and ramp curves, with every assumption documented and defensible
- Product-qualified lead definition and trigger – a usage threshold that hands self-serve accounts to sales with context attached, not a bare email
- Product-usage-to-CRM pipeline – API calls, seats, and consumption synced into the CRM on a schedule, with churn-risk alerts on usage drops
- Single ARR methodology – one defined definition of revenue the board, finance, and sales all work from
- Operator runbook – documentation and training so your team runs the systems after we leave
In a consumption business the contract is the floor, not the forecast. Most revenue operations problems at AI / ML companies are not forecasting problems. They are reconciliation problems – the moment your usage number and your bookings number stop agreeing, every report downstream is fiction.
Our Methodology
We work in a 90-day sprint because a revenue operations project rots if it drags. Days 1 to 30 are the audit and the map – we trace revenue from product consumption to reported ARR, document every sync, and find exactly where the usage number and the bookings number split. We do not build anything in the first month. We earn the right to build by proving we understand your revenue better than the people reporting it.
Days 31 to 60 are the core rebuild: the reconciliation layer and the product-usage pipeline into the CRM. We ship in small pieces, test each against real customer records, and keep the old reporting running until the new numbers are proven. Nothing goes live on faith, because a wrong revenue number is worse than a missing one.
Days 61 to 90 are the forecast model and the PLG-to-sales handoff – the layers that only work once the data underneath is clean and reconciled. We end the sprint with your team trained and a runbook in their hands. The hard stop is on purpose: if something is not working by day 90, we want it visible, not buried in a retainer that quietly renews.
How We Work
The first 30 days are diagnostic. We embed with your finance, sales, and data people, get read access to the billing system, the CRM, and the product usage tables, and build the map. You get a documented picture of how revenue actually flows through your company – usually the first time anyone has seen it whole. We surface where consumption and bookings diverge and confirm scope before any building starts.
Days 30 to 60 are construction. We build the reconciliation layer and the usage-to-CRM pipeline inside your existing systems, and we do not force a rip-and-replace unless your tools genuinely cannot do the job. We work in your environment, on your tickets, with your team watching, so the knowledge transfers as we go instead of in a handoff doc at the end.
Days 60 to 90 are the forecast model, the product-qualified lead handoff, and enablement. We stand up the consumption-aware forecast, wire the PLG trigger, and train your operator to run and extend the model. By the end you have a team that can defend the number without us in the room.
The team is senior and small – the operators doing the work are the ones who scoped it, not a junior pool behind a project manager. Cadence is a weekly working session plus an async channel for the in-between. We move at the speed of decisions, not the speed of a status report.
- Reported ARR backed by a single reconciliation methodology, so finance and sales stop arguing over whose number is right
- A forecast that accounts for consumption swings, so a large account changing its usage no longer blindsides the quarter
- Product-qualified leads routed to sales with usage context, so self-serve accounts ready to expand stop slipping through the funnel
- Product usage flowing into the CRM with churn-risk alerts, so renewals and contractions are caught while there is still time to act
- A documented revenue system and a trained operator, so the work survives your next finance or ops hire
If your ai / machine learning company needs revenue operations 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 revenue operations engagement for an AI / ML company cost?
Most engagements run $30K-$80K for the 90-day sprint, depending on how tangled your billing and CRM are and how many revenue motions you run. A single-motion company with a clean usage feed sits at the low end.
How long before we can trust the forecast?
The audit delivers value in the first 30 days, because a clear map of where consumption and bookings diverge is often the first one anyone has had. The reconciliation layer ships in days 30 to 60, which is when your reported ARR starts being a number instead of a guess.
Do you replace our billing and CRM tools or work with what we have?
We work with what you have whenever your tools can do the job, which is most of the time. AI / ML companies usually have capable billing and CRM platforms that were wired together badly under time pressure, not the wrong platforms.
What makes Winston Francois different from a revenue operations agency?
We are operators who have run revenue inside scaling companies, not an agency renting you billable hours. The senior people who scope your work are the ones who build it.
How do you measure ROI on a revenue operations project?
We define the metrics in the first 30 days against your current baseline, so improvement is measured, not claimed. The usual numbers are forecast accuracy against actuals, the gap between reported and reconciled ARR, speed from product-qualified signal to sales touch, and how much expansion and churn you catch before the invoice.
Is this a fit for a Series A company that has not built revenue operations yet?
Yes, and building it right the first time is cheaper than untangling it later. Early-stage AI / ML companies often have consumption billing flowing into a CRM with no reconciliation, no product-qualified lead definition, and a forecast built on close dates alone.
