A firm crosses some invisible line on its way from eight people to thirty. The work the founder once cleared personally — the client documents that need chasing, the intake that needs sorting, the follow-up that never goes out on time — stops fitting into the margins of anyone’s day. It is still getting done, mostly: late, or by someone senior who should be doing something more valuable, or not quite completely. Nobody decided to let it slide. The firm grew, the operational work grew with it, and at some point the arithmetic of who-does-what quietly stopped balancing.
Then comes the question every managing partner at this size eventually asks: is there something built for a firm like ours? Not for a solo shop that runs on a founder’s evenings, and not for a five-hundred-person company with a budget for an operations department — for the firm in between, with real recurring work and no obvious place to put it.
This article is about that in-between. Our interest in the answer is worth stating plainly: Gridex operates managed work for firms of exactly this size, so this is an argument from an interested party — which is why every external figure below is named, dated, and checkable.
The Work That Doesn’t Fit a Job Description
Start with what’s actually piling up. A 22-person accounting firm hits busy season and the bottleneck isn’t the returns — it’s chasing the documents needed to start them, the same organizer requests going out three times before the client responds. A 30-person engineering firm pursuing a federal contract spends a week rebuilding the same past-performance and compliance materials it assembled for the last proposal, because no one owns keeping them current. A growing practice of any kind accumulates intake, review prep, and follow-up that scales with the client base but never appears on an invoice.
This work has two features that make it hard to place. It is recurring — it arrives every week, every busy season, every proposal cycle — and it is no one’s whole job. It’s a slice of three people’s roles, which means when those people are busy with billable work, it’s the slice that gives. That is the work a 10–50 person firm is trying to find a home for.
The Four Answers, and Why Each One Pinches
By the time a firm asks the question, it has usually weighed the familiar options. Each one fits some firms; none fits this size cleanly.
Hire someone. A dedicated operations or admin hire turns variable work into a fixed cost. But the work is spiky — it surges at busy season and proposal deadlines and goes quiet between — so the role is hard to scope and easy to over- or under-hire. The fully loaded cost of an employee runs roughly 1.25 to 1.4 times base salary once benefits, payroll taxes, and overhead are counted, per SBA’s SCORE guidance — a real commitment for work that doesn’t fill a calendar evenly. And if the hire leaves, the knowledge leaves with them.
Outsource it. Handing the work to an outside provider moves the labor but keeps the responsibility. Someone senior now writes the instructions, answers the questions, and checks the output — supervising a team they can’t see. For many firms the oversight costs more in partner time than the work it replaced.
Buy a tool. Software promises to absorb the work, but the firm has to configure it, adopt it, and keep using it. At this scale that rarely holds: small-business AI adoption actually fell from 42 percent to 28 percent between 2024 and 2025 in NEXT Insurance’s survey of 1,500 owners, and in Goldman Sachs’ 10,000 Small Businesses survey, 49 percent of small firms using AI named lack of technical expertise as a barrier. The tool arrives; the work stays. We’ve made that case in full in “The hidden adoption cost of new software.”
Pile it on the team. The default answer is no answer: absorb it. That works until it doesn’t — backlog, overtime, and the quiet erosion of billable hours as the best people spend evenings on operational work nobody is paying for.
None of these is a mistake. They’re the available shapes, and at this firm size each one pinches.
The Valley
Step back and the pattern is structural. The 10–50 person firm sits in a valley between two models that genuinely work — for everyone but it.
Below the valley is the owner-run operation, where the founder absorbs the operational work personally. That model is fine until growth outpaces one person’s bandwidth, somewhere in the low double digits of headcount for most professional service firms. After that, the work backs up.
Above the valley is the enterprise, which can justify a dedicated operations department or an embedded engineering engagement — the “forward-deployed” model the large AI firms now use to make their systems actually work inside a client. That model delivers, but it is expensive by construction. A capable forward-deployed engineer commands a salary of $135,000 to $200,000 and up; spread across the few clients one can embed with, the all-in cost clears $75,000 per deployment a year, as SaaStr’s Jason Lemkin documented in August 2025. His conclusion is the whole point: “The very businesses that need AI transformation most may be priced out of the implementation model that actually delivers results.” We unpacked that economics in “The forward-deployed model was built for the enterprise.”
This is not a niche gap. According to the SBA’s 2024 Small Business Profile, 99.9 percent of US businesses are small businesses, and they employ nearly half — 45.9 percent — of the private-sector workforce. The firm in the valley isn’t an edge case. It’s the typical firm. It just happens to be the one neither the owner-absorbs model nor the enterprise model is built to serve.
Why the Shape Fits, Not the Price
The reason managed AI capacity fits this band isn’t that it’s a cheaper version of the enterprise model. It’s that the shape of the delivery matches the shape of the problem.
The problem is one recurring, judgment-bounded block of work that arrives unevenly and belongs to no single role. The model answers it the same way: AI systems absorb that block — intake and document chase, review preparation, the follow-up queue — and return it as a review-ready packet, queue, or brief. Your team opens it, checks it, and decides. A client document packet at the accounting firm; a compliance matrix on the engineering firm’s proposal; a qualified follow-up queue behind any practice’s shared inbox — same pattern, three shapes of firm.
Nothing about that requires a fixed hire for variable work, an enterprise budget for embedded staff, or a software rollout the team has to sustain. The production labor leaves; the judgment stays. That’s what makes it a fit for the valley specifically — it’s sized to one workflow, not to a department, and the firm never has to grow into it.
What Actually Has to Fit
Size isn’t really the gate, which is why “10 to 50” is a description, not a rule. A smaller firm with a genuine recurring-work problem can be a fit; a larger one with the same dynamic can be too. What has to fit is the work: it needs to be recurring, it needs to have judgment criteria a system can be taught, and it needs to produce a reviewable artifact someone on your team signs off on. Open-ended, one-off, or undefined work — the kind where there’s no consistent output to operate — isn’t the right shape, and we’ll say so rather than force it.
That’s the honest boundary. Within it, the firm in the valley finally has an answer built for its size instead of scaled down from someone else’s.
Show us one recurring block of work your team is buried in, and we’ll tell you whether it fits what we operate.