Received wisdom: classic frameworks under AI pressure 02: Crabtree's LER

By Garrett, 05 Mar 2026

What happens to the labour efficiency ratio when labour isn’t the bottleneck?

Greg Crabtree has spent decades telling agency owners something they did not want to hear. Your business is not a creative enterprise with a financial problem. It is a labour arbitrage operation with a brand. Every dollar you spend on people either earns you more than a dollar back in gross profit, or it does not. If it does not, you do not have a business. You have a hobby with a payroll.

Crabtree’s core metric, the Labour Efficiency Ratio, is deceptively simple. Divide gross profit by total labour cost. For most service businesses, Crabtree says, a healthy ratio sits between 1.5x and 2.5x, meaning every pound spent on people, including the owner, priced at genuine market rate, should return at least $1.50 in gross profit. Below 1.5x, you are funding your clients’ projects with your own capital. Above 2.5x, you are probably leaving capacity money on the table or grinding your team into powder.

What makes the LER worth taking seriously, eighteen years after Simple Numbers, Straight Talk, Big Profits! first appeared, is the discipline it imposes on the instinct to hire ahead of revenue. Agency owners, especially those who have grown by winning, tend to staff for ambition rather than actuals. The LER punishes this immediately and legibly. When you bring on a senior strategist six weeks before the project, they are supposed to work on materialising, the ratio drops. The number forces the conversation that the P&L alone does not.

The argument that has aged well

Crabtree’s deeper thesis in Scaling Up, co-written with Verne Harnish, is that profitable growth in service businesses requires expanding gross profit faster than you expand headcount. The maths is simple, but resisting the cultural gravity of agency life is challenging. Growth in agencies is often measured by people, team size, and org charts that appear hefty in pitch decks. The LER serves as a counterforce, questioning whether the org chart genuinely generates a return.

This idea remains relevant because it is fundamentally true. A boutique agency with twelve people making $2.4m in gross profit and $1.2m in labour costs is more profitable and resilient than a thirty-person agency with $3.6m in gross profit and $2.8m in labour costs, although the latter looks more substantial externally. The first has a 2.0x LER, while the second operates at 1.29x, probably deceiving itself about whether the senior hire last March was necessary. The discipline holds firm; what is beginning to shift is the understanding of what constitutes labour.

Where AI creates the crack

AI presents a challenge for Crabtree’s framework. Artificial intelligence tools can produce outputs that boost gross profit without appearing in the labour cost denominator. They are embedded in software subscriptions, SaaS line items, and the tools budget that most agency finance directors classify as overhead rather than cost of goods sold. The LER does not account for them. The ratio remains clean while the underlying model shifts underneath.

The ratio Crabtree would likely invent today

If Crabtree were writing Simple Numbers today, having watched what AI does to agency cost structures, my guess is that he would extend the denominator. Call it the Production Efficiency Ratio, or keep the LER name and widen what counts as labour cost. The principle holds, and it is worth restating plainly. Measure gross profit relative to the total cost of generating it, but extend the cost bucket to include all capacity-generating expenditures, not just payroll.

In practice, this means pulling AI tooling subscriptions out of the overhead bucket and treating them as production costs. It means asking, for every AI platform you pay for, whether the capacity it generates is priced into your billing model. The discipline Crabtree applied to headcount decisions applies equally here. Are you paying for production ability you cannot bill against? Is that $500-per-month AI subscription generating $4,000 in additional billable throughput per month? That is an 8x return, considerably better than most humans and worth knowing explicitly rather than letting it dissolve into an overhead line.

There is an uncomfortable flip side. Some agencies are buying AI tools because they feel they should, running them intermittently while generating no measurable increase in throughput, and feeling vaguely modern for having done so. A Crabtree-style analysis of those tools would show them as a pure cost drag. The ratio would punish them exactly as it punishes a hire who never quite hit their billing target.

What the LER tells us about AI augmentation

There is a version of the AI-and-agencies conversation that treats the technology as a question of quality or creativity, focused on better output and faster iteration. That conversation is incomplete. Crabtree would redirect it immediately towards the ratio. The question is not whether AI makes your work better. It is whether AI makes your labour more efficient, and by how much.

For agencies in the £1m to £5m revenue range, where labour typically accounts for 55% to 70% of revenue, even modest LER improvements compound significantly. If an eight-person strategy firm moves from a 1.6x to a 1.9x LER through AI augmentation of research and client reporting, and holds revenue steady, the gross profit improvement over twelve months is material enough to change what the owners can pay themselves, invest in, or hold in reserve.

Crabtree would say it plainly: “Show me the number”. Not the case study, not the productivity anecdote, not the AI tool’s marketing page. Show me the gross profit before and after, along with the total production cost that generated it. That is a much more demanding test than most AI adoptions are currently being held to.

The benchmark question

One thing that has not yet happened but should is a systematic update to LER benchmarks for AI-augmented businesses. Crabtree’s 1.5x to 2.5x range was calibrated against a world where every unit of productive capacity was a person. If AI tools can generate meaningful capacity at a fraction of the per-unit cost of a human, the ceiling on attainable LER should move.

An agency with a genuinely AI-woven delivery model, where AI handles first-draft production and research synthesis with humans doing the judgment work and final quality pass, might sustain a 3.0x or 3.5x LER on a stable headcount. That would have been borderline exploitative in Crabtree’s original framework, because the only way to hit those numbers with humans alone was to overwork people. With AI absorbing the volume load, it might simply be what a well-run operation looks like. The benchmarks will shift. The ratio will not.

The owner’s compensation problem, revisited

One of Crabtree’s most debated points is that the owner must be included in the labour cost at a genuine market-rate salary, even if they choose to pay themselves differently. The logic is sound. Remove owner labour, and the ratio is biased in favour of businesses that run on founder subsidy. You need to see the true cost of the capacity being deployed.

AI creates a similar related issue. If the founder is personally managing the AI toolset, prompting the systems and reviewing outputs before billing the capacity as if it were delivered by a team, the cost of their attention does not appear in the denominator. The LER looks neat, but the owner is exhausted. The business is not truly scalable as the numbers suggest.

Crabtree’s original point, that you must include the cost of the thing actually generating the output at the rate it would cost to replace it, applies directly to AI-augmented founders undertaking AI work they have not yet operationalised. If you spend ten hours a week managing your AI systems and that management is central to your delivery model, the cost of that time belongs in the denominator, and the ratio will decrease. This is not bad news; it’s accurate news, which is Crabtree’s main point.

What this looks like in practice

At Better Than Good, we work with clients where, increasingly, the discussion is not “should we use AI?” (that one is mostly settled) but “how do we account for it honestly so the numbers still reflect what we think they reflect.”

The pattern we see most often is a firm that has enthusiastically adopted AI tooling, watched its output volume climb, and concluded it is more efficient than ever, without having reclassified a single line item in its P&L. The LER looks healthy. But when we reconstruct the true production cost, including tooling, founder time managing the systems, and the quality-check overhead that shifted from writers to principals, the picture shifts. Sometimes the efficiency gain is real, and the ratio, correctly calculated, is even better than the headline number suggested. More often, there is a gap between the reported ratio and the actual one.

The solution is straightforward and requires the same discipline Crabtree has always advocated. Get the denominator right before trusting the ratio. If you want to see how that applies to your business, use our simple calculator and get in touch.

The number that never lies

Crabtree’s appeal, across two books and a generation of business owners who have attended his workshops, is not about intricate analysis. The LER works because it is difficult to manipulate and harder to dispute. A decrease in the ratio indicates an issue with the cost structure. The reasons for the decline and whether a new hire will pay off are secondary; the fact remains that it has fallen.

AI does not alter this principle. What AI does is demand a more honest accounting of what should be included in the denominator. An agency partner billing AI-generated capacity without accounting for the tooling that produced it is doing the same as the founder working 60 hours a week, but counting their salary as 40. The figure appears more favourable, but the business remains fragile.

The LER will endure beyond this AI hype cycle because it poses the right question. The necessary upgrade is modest. Widen the denominator to encompass all production capacity costs, not just those linked to a human. Calculate the ratio monthly. If it declines, investigate the cause before the cash flow becomes a pressing concern. If it increases, verify that the rise is genuine before hiring based on it.

Crabtree’s metric was always more than just labour productivity. It was about whether the business had earned the right to its size. That question becomes more urgent, not less, as capacity costs decrease and the temptation to grow billing before achieving real efficiency grows. The cost of tools may change, but the discipline remains constant.