Operating expense: the ten levers below the delivery line

What is different about overhead in a services firm

In a software company, operating expense is where most of the levers were and the smallest moves. In a services firm, overhead is smaller as a share of revenue (delivery is the big line) but it is shaped around three things that conversion removes: recruiting as the growth engine, time-and-billing as the finance system, and partner compensation as the behavior system. The first two shrink. The third has to be redesigned, and if it is not, every delivery lever meets a partner who is paid to resist it.

The G&A benchmark that anchored the software edition's overhead group carries here too: the gap between median and first-quartile SG&A among large public companies was 8.4 points of revenue before any of these tools existed, and AI is the cheapest way yet to close a gap the best-run firms had already closed.¹

O1. Business development and proposals

What it is. Proposal, RFP and scoping automation; pipeline discipline in a firm that has never had a pipeline, only partners' calendars. The firm's proposals are the first documents the platform should learn to write, because they are repetitive, high-stakes and currently produced by the most expensive people in the building at the end of the day.

Line it moves. Sales and marketing (which appears as a line for the first time, as partner selling time is costed under R7).

Typical range and evidence. Not measured for services firms. The software edition's sales-productivity evidence was Medium (a sponsor's portfolio S&M ratio falling, a vendor's named cases) and the mechanism transfers.² Grade: Weak (not measured in services); Medium by analogue.

How to run it. Cost partner proposal time at close. Capture the last fifty proposals as D4 captures delivery frameworks. Put the first draft of every new proposal on the platform within the first hundred days; the partner edits rather than writes.

Kill criteria. Partner hours per proposal not halved at year one.

Metrics. Year one: partner hours per proposal; proposals per month; win rate. Year three: S&M as a share of revenue, costed.

Horizon and stage. Deploy; day 100 to year one.

O2. Marketing production and demand generation

What it is. Built from near zero. A referral-driven services firm has a website and a conference budget. The platform-delivered firm needs content that explains the product, a demand motion for R6 and R7, and the production capacity to run it without an agency. The software edition's marketing-production lever carried Medium evidence from named vendor cases and carries here unchanged.³

Line it moves. Sales and marketing.

Grade. Medium, carried.

How to run it. Start after R2, because there is nothing to market until the platform-delivered line exists. One person and the platform, not an agency. Measure cost per qualified lead against the costed partner motion.

Horizon and stage. Deploy; year one.

O3. Recruiting reshaped

What it is. In a services firm, recruiting is the growth engine: revenue grows when billable people are hired, so the recruiting function is large, continuous and expensive. After D2's hiring freeze it becomes a small, specialist function that hires reviewers, platform operators and client leads a few times a year. The software edition's recruiting lever (screening and scheduling automation, Medium evidence from named vendor cases, no quality-of-hire data) carries, but the larger move here is the function shrinking rather than automating.⁴

Line it moves. G&A.

Grade. Medium (automation, carried); the function's resize is the firm's own.

How to run it. Size recruiting to D2's plan, not to the firm's history. Keep the people who know how to assess judgment, because that is what the firm hires for now.

Kill criteria. Recruiting cost not down in proportion to production hiring at year one.

Metrics. Year one: hires by role type; recruiting cost per hire; recruiting cost as a share of G&A. Year three: same.

Horizon and stage. Convert; years one to two.

O4. Finance: time-and-billing, WIP, close, FP&A

What it is. A services firm's finance function runs on the time system: hours become WIP, WIP becomes invoices, invoices become receivables, and the month-end close reconciles all three. Conversion removes hours as the unit, which removes most of the reconciliation. The close, reporting and forecasting automation evidence carries from the software edition: benchmark close times, a named vendor case of reconciliation match rates rising from 30 to 90 percent, and FP&A tooling with named customers.⁵ The services-specific move is that unit-based billing (C1) replaces time-based WIP, and forecasting runs on the scorecard rather than on utilization.

Line it moves. G&A.

Grade. Strong (benchmark) / Medium (cases), carried.

How to run it. Retire time-based WIP with D8, in the first hundred days. Rebuild the close around units billed and subscriptions recognised. Put the scorecard in the board pack in place of the utilization report.

Metrics. Year one: close in days; finance headcount per $10M revenue; scorecard produced from the ledger without manual assembly. Year three: finance as a share of G&A.

Horizon and stage. Deploy; day 100 to year one.

O5. Legal and engagement contracting

What it is. Engagement letters, terms, amendments at renewal (P4's calendar), and the AI-disclosure and review-model language that P9 puts into every renewing agreement. The software edition's legal lever carried a randomised trial showing large productivity gains on most legal tasks and named vendor cases on contract cycle time; it carries here.⁶ The services-specific content is that engagement letters are the firm's product terms now, and they change at every renewal for two years.

Line it moves. G&A.

Grade. Strong (trial) / Medium (cases), carried.

How to run it. Template the platform-era engagement letter in the first hundred days (review model, data processing, AI disclosure, tiers). Automate the amendment cycle against P4's calendar.

Metrics. Year one: share of book on the new letter; outside-counsel spend. Year three: outside-counsel spend as a share of G&A.

Horizon and stage. Deploy; year one.

O6. Internal systems consolidation

What it is. A services firm runs on a practice-management system, a time system, a document system, a billing system and a CRM, often from five vendors and often badly integrated. Most of them exist to manage hours and documents, which the platform now holds. The software edition's SaaS-spend lever (a Strong finding that a third of licences go unused, Medium savings cases) carries; the services-specific move is retiring whole systems rather than trimming seats.⁷

Line it moves. G&A.

Grade. Strong (waste) / Medium (savings), carried.

How to run it. Inventory at close. Retire the time system with D8. Retire the document system as D4 moves the method onto the platform. Keep finance and CRM until the platform can replace them without a project.

Metrics. Year one: systems retired; software spend per employee. Year three: software spend as a share of G&A.

Horizon and stage. Deploy; year one.

O7. Partner compensation redesign

What it is. The behavior system. A services firm pays its senior people on origination (revenue they bring in) and billable hours (their own and their team's). Both reward the thing conversion removes. A partner paid on hours will protect hours; a partner paid on origination will sell projects rather than subscriptions; neither has any reason to spend a month writing the method down (D4) or to hand a client to a platform tier (R1). The lever is to change the formula in the first compensation cycle after close to pay on what the scorecard pays on: retention of the book (R4), conversion of pricing (R1), review quality (D3), expansion (R3) and, through rollover equity, the exit.

Line it moves. G&A (partner compensation is often the largest overhead line), and it governs behavior across the D- and R-groups.

Typical range and evidence. Not measured as an outcome anywhere public. The mechanism is the critics' second reason in the Fortune op-ed, that under hourly economics firms revert to maximizing hours, which is a statement about incentives as much as about pricing.⁸ A conventional PE structure in accounting, as presented to partners in 2025, uses deferred payments and rollover equity to keep partners through the hold period; it does not, as presented, tie anything to conversion.⁹ Grade: Weak (no outcome data); Strong as a mechanism, because the incentive is arithmetic.

How to run it. Design the formula before close, with P7, because it is part of the structure. Pay for D4 time explicitly in year one. Pay for premium-tier ownership (R11) and for group revenue (R9) rather than own-firm origination. Publish the formula; a compensation system partners do not understand is one they will assume is against them.

Conditions. P3's map, so the formula is designed for the people who hold the book. Rollover equity in the structure, so the exit is where the partners' upside lives.

Kill criteria. Formula not changed by the first compensation cycle after close. Named-person retention below 90 percent in the quarter after the change (the formula was wrong, or it was communicated badly; find out which).

Metrics. Year one: share of partner compensation tied to scorecard metrics; named-person retention. Year three: partner compensation as a share of revenue; retention.

Horizon and stage. Convert; year one.

Across the five firms. A: partners are consultants with books; the formula moves to retention and expansion. B: partner comp is the most formalised and the hardest to change; the APS structure (P5) may put partners in a different entity, which complicates it. C: origination culture is strongest; this is the archetype where O7 decides the deal. D: engineers with stamps are paid for the stamp; less change. E: fewer partners, more managers; O7 is a management-bonus redesign.

O8. Shared services across the roll-up

What it is. One finance, one HR, one IT, one legal function for all the acquired firms, running on one platform (B3). In a conventional roll-up this is the synergy in the deck that takes three years to appear because each firm keeps its systems. Here the conversion has already retired most of each firm's systems (O6) and most of its billing complexity (O4), so the shared function is smaller and arrives sooner.

Line it moves. G&A across the group.

Typical range and evidence. Roll-up synergy data is thin and mostly private. The best public caution is the opposite case: an acquisition-led platform with reorganisation costs in every quarter and interest expense tripling in a year.¹⁰ Grade: Weak (no data); the cautionary comp is Strong.

How to run it. Only after the second firm is on the platform (B3). Shared services before a shared platform is two integrations at once.

Kill criteria. Group G&A per dollar of revenue above the first firm's standalone at the second anniversary of the second deal.

Metrics. Year two onward: G&A per dollar of group revenue; shared-function headcount.

Horizon and stage. Convert; years one to three.

O9. The executive operating system

What it is. The register, the weekly kill review, and the board report that carries lever outcomes and the scorecard rather than the utilization report. Carried from the software edition with its evidence: a majority of PE leaders reporting AI benefits inside twelve months while under a third call their implementation efficient, and an analyst forecast that a large share of agentic projects will be cancelled for unclear value.¹¹ The gap between those numbers is what the register is for.

Line it moves. G&A (small); governs everything.

Grade. Strong (mechanism), carried.

How to run it. The register goes live at day 100 (P8). Every lever on it has a baseline, an owner and a kill date. The kill review is weekly and the board sees the scorecard monthly. The first lever killed on schedule is the proof the system is real; a register where nothing is ever killed is a wish list.

Metrics. Day 100: register live. Year one: levers killed on schedule; board pack carries the scorecard. Year three: the same.

Horizon and stage. Convert; day 100 to exit.

O10. G&A benchmark closure

What it is. The sum of O1 to O9: the whole overhead stack taken from where a services firm sits toward where the best-run companies already are. The anchor carries from the software edition: a 2026 benchmark of the thousand largest North American public companies put median SG&A at 16.2 percent of revenue against 7.8 percent for the first quartile.¹ A services firm's overhead is a smaller share of revenue than a software company's because delivery is larger, so the absolute prize is smaller; the direction and the discipline are the same.

Line it moves. G&A.

Grade. Strong, carried.

Metrics. Year one and three: G&A as a share of revenue against the firm's own close baseline and against the benchmark quartiles.

Horizon and stage. Convert; years one to three.