Rule of 40 in the AI era: what a software company should check after a PE deal

In board meetings at software companies after a private equity deal, I keep hearing the same sentence: the AI features are coming, and the score will hold. Usually one line is missing from the plan, the gross margin of the new features.

What is the Rule of 40?

The Rule of 40 is a rule of thumb for software companies: revenue growth in percent plus profit margin in percent should add up to at least 40 ^[1]. A company growing 25 percent with a 15 percent EBITDA margin sits exactly on the line; one growing 10 percent needs a 30 percent margin. Which margin is meant varies by source: EBITDA margin, operating margin or free cash flow margin. Anyone comparing figures should settle the definition first.

In the German Mittelstand, the mid-sized and often family-owned companies, I mostly meet the Rule of 40 in two places: when software companies are valued at a sale, and in the value creation plan afterwards. Buyers read the score as shorthand for one question: does growth come from the business itself, or is it bought with money? Where the margin comes from, the score hardly shows. That is where AI pushes gross margin down.

A typical case: the software company after the PE deal

The following example is constructed. It condenses a pattern that is common among German mid-sized software companies and describes no real firm.

A software company in southern Germany has spent twenty years building software for trade businesses such as electricians, plumbers and carpenters: quotes, job planning, time tracking, invoicing. It has revenue of around 20 million euros, a little over half from subscriptions, the rest from implementation and maintenance. Customers stay for years because the software maps their working day. The founder sells a majority stake to a private equity platform that brings several vertical software vendors together in a buy-and-build model.

The value creation plan has three levers: raise prices, integrate acquisitions, add AI features. One assistant writes quotes from photos and voice notes taken on site, a second one schedules the technicians. Let's assume customers like both. At the start the company grows 15 percent at a 25 percent EBITDA margin, so the score is 40. For the AI module, the plan shows rising revenue at last year's gross margin.

Gross margin sets the ceiling

EBITDA margin and operating margin always sit below gross margin. Classic subscription software often runs at a gross margin around 80 percent, because one more licence costs almost nothing to deliver. AI features behave differently: every request to a language model costs compute, and that cost grows with usage. Surveyed AI vendors expect their AI products to reach a gross margin of around 52 percent in 2026 ^[3].

For the software company, that means (my own calculation, assuming an 80 percent gross margin on existing revenue): once the AI module reaches 30 percent of revenue, the blended gross margin falls by a little over eight points. If all other costs stay the same and the AI module adds no extra growth, the score drops from 40 to about 32. If the quote assistant, the feature customers value most, also uses the most compute, the gap grows with every success of the feature.

Who pays for the AI costs?

The answer sits in the pricing model. If the company bundles the AI features into the existing contract, it carries the volume risk itself: every technician who uses the assistant more often costs margin. If it charges through credits or by usage, the volume risk moves to the customer. 29 percent of the surveyed software companies already use such AI credits ^[4].

There is limited room. 70 percent of respondents say AI is paid for out of customers' existing tech budgets ^[4]. A trade business with twelve employees has no separate AI budget. I would therefore start with tiers: basic functions in the subscription, heavy use through an allowance that is billed visibly. How to build such tiers is something I describe in the pricing ladder for AI pricing models.

For the software company the numbers work out like this: if the pricing model passes half of the margin gap on to customers, about four points come back, and the score stands at 36. Without that step it stays at 32.

What retention adds

The score is a snapshot. Retention shows whether growth will last. I measure it on the original cohort: how much of the revenue from customers who were there two years ago is still there today, without new customers and without acquisitions? On a buy-and-build platform this is one of the first questions, because acquisitions inflate growth. Without that split, the score mostly measures the acquisition list.

Customer lifetime value is calculated on gross profit, and AI pushes gross profit down. The same customers are worth less at a 72 percent gross margin than at 80 percent, even if they stay a little longer. Anyone selling the AI features as a lever for retention should run that calculation as well.

Which bar applies: fund or long-term owner?

How strictly to read the 40 depends on who owns the company. Dave Kellogg works through a scenario for debt-financed buyouts: if the valuation multiple halves, the Rule of 40 no longer delivers the target return, and the bar rises to a Rule of 60, for example 20 percent growth plus a 40 percent EBITDA margin ^[2]. The halving is an assumption in his model; the calculation itself is open and easy to follow.

Bain argues from management's point of view: companies that want to compete with AI-native vendors may have to accept a Rule of 30 for a while ^[1].

For the software company in this case, that means: the fund has debt in the structure, interest is running, and the holding period is limited. For the fund, a score of 32 over two years is a real problem. A family business or a long-term investor can carry the dip more easily, as long as the plan shows a dated path back to 40.

How should a board check the Rule of 40?

The check needs a P&L by product line and the pricing logic of the AI features.

1. Fix the margin definition. Which margin is in the score, and was it the same last year?
2. Split gross margin by revenue type. Report subscription revenue and AI revenue with their own gross margins. Check whether the cost of model calls is booked under hosting or under cost of revenue.
3. Lay the pricing model against the cost curve. When usage rises, does the price rise too, or only the cost?
4. Measure retention on the original cohort. Separate organic growth from acquisitions.
5. Calculate the score two years out. Insert the planned AI share, recalculate with the target gross margin and discuss the gap to today's figure with management.

My Take

I still find the Rule of 40 useful for mid-sized software companies, as a starting point for board discussions. It forces growth and margin into one sentence. With AI, it needs a second figure next to it: the gross margin of AI revenue.

An AI assistant on its own does little to protect the software company from competition, because every competitor can buy the same language models. What remains are the data from twenty years of trade jobs and the firm place the software has in customers' working day. AI features built on that data support higher prices. A language model that is merely plugged in creates costs and no pricing argument.

In due diligence I watch for one pattern in the plan: AI revenue rises while gross margin stays at last year's level. Then the cost side of AI is missing. The plan has to explain how the margin will hold, through prices, cheaper models or savings elsewhere.

If you want to break down the score of a portfolio company this way, the Rule of 40 method on convios.com sets out the steps and the limits of the metric. For the cost side, the AI cost calculator helps.