Cutting Back Office Admin With AI
Timesheet chasing, invoicing, month-end, reporting, and the non-billable hours that go with them... its all necessary work in a consulting firm. Without it, clients don't pay you. It's costly though – for most firms, admin as a % of revenue sits at about 15%. Its a whole lot of cost that otherwise could be sitting in your EBITDA. In the AI era, the best consulting firms will bring this figure into the single digits. Learn how in this episode of the podcast.
Back-office admin can eat 15% to 30% of a consulting firm's revenue, and almost all of it would otherwise be profit. On the latest Projectworks Podcast, CEO Mark Orttung and CCO Dominique Rennell argued that cutting back office admin costs is one of the easiest ways to lift a firm's profit, and one of the last places leaders think to look. The cost they mean is general and administrative expenses, the G&A that covers finance, software, and the work of running the firm behind the scenes. Most leaders barely track it. AI is what changes that, because the repetitive part of that work is exactly what it handles well, and you can cut it without hiring anyone new.
Dominique put it plainly. "It is part of running and growing a successful consultancy firm that we believe lots of leaders underestimate."
Why lean admin is where margin starts
Lean admin sits at the base of how Projectworks thinks about a maturing firm. It isn't client-facing, so it rarely gets attention, and that is why the opportunity is still there for the taking. Everyone works over sales and delivery. The back office gets left to run however it always has.
Mark's take is that this is backwards.
So it's, on the one hand, one of the easiest places to increase your bottom line. It's also one of the least paid attention to, I think.
Every point you take off G&A is profit you keep, and you don't have to win a better contract or lift utilization to get it. Both of those take quarters to move. Admin is the one lever a firm can pull on its own, this month, without anyone else's sign-off.
What goes into a consulting firm's G&A line?
G&A splits into two kinds of cost, and only one of them is worth pointing AI at.
Fixed costs you can't easily change
The first kind is fixed. Salaries, rent, software licenses, and the admin headcount that keeps the lights on. You can move it, but only by cutting people or reopening contracts, and most firms have squeezed it more than once already. The gains here are slow and they hurt.
Back-office work you can manage
The second kind is the repetitive work that sits on top of the fixed costs. Invoicing, month-end, timesheets, resourcing admin, the paperwork around hiring. This is where the hours stack up week after week, and where AI gets real traction, because the work follows rules and repeats. It used to be too fiddly to automate without a developer, and now it isn't. Picture the finance lead who gives up the last three days of every month to reconciliations, or the delivery manager who loses a Friday to chasing timesheets. That is the work to hand off first.
What is the benchmark for healthy G&A in professional services?
There isn't one number, and anyone who hands you one is selling something. The honest version is a range that tightens as you grow. A small firm might sit above 15% and be doing fine for its size. A bigger firm can get toward single digits, because scale spreads the fixed costs over more billable people. The picture below is a rough guide to how that target shifts with size, not a benchmark pulled from real data.
For what good looks like at scale, look at Mark's own firm. He took G&A at Nexient from about 10% when he joined to roughly 7% as it grew. Three points doesn't sound like much until you remember it all lands in profit. Most smaller firms won't get to 7% easily, and that's fine. It tells you which way to head, not where you should already be.
The cost climbs as you add people, which Dominique has watched happen.
When we polled the live audience during the recording, the answers ran from under 10% to over 30%, and plenty of people weren't sure at all. That last group is the one to worry about. You can't manage a cost you've never measured.
Where admin hides in every engagement
Admin is easy to underestimate because it never lands as one big line. It's smeared across the whole life of an engagement, a bit at every stage, and it grows as you take on more work. No single piece looks like much. Add it all up and it's a real chunk of the cost base, which is why you fix it in small pieces rather than one grand move.
Three ways Projectworks is bringing AI into the back office
Projectworks is putting AI on the back office in three ways, each at a different stage of release.
Proposals are first. The AI is built into the CRM and trained on the proposals a firm has already won, so teams bid on more work faster without the quality dropping off. This is in private beta.
Next is Kea, an AI team-mate built into Projectworks and named after the New Zealand alpine parrot. Kea takes the resourcing data a firm already has and turns it into staffing answers and risk signals, so the weekly resourcing meeting runs on live numbers instead of memory. The first release is advisory. It's built to help you understand and decide rather than act on its own, and it's still in development.
Third is the Projectworks MCP. MCP is a standard way for AI tools to plug into other software, a universal adapter between the AI you already use, like Claude or ChatGPT, and the systems your firm runs on. It brings your Projectworks data into those tools so you can act on it there. Right now it covers resourcing and timesheets, with expenses to follow. If you want to see any of this working, book a demo of the AI features.
Mark made the bigger point on a past episode about why connected data is what makes any of this pay off.
Having all of this data together and well-structured really becomes powerful as you start to put a system of action or an AI set of agents on top of it.
Three ways to put this to work today
The episode ended on three examples you can try now (watch the recording above for the video walkthroughs). One of them isn't a Projectworks feature at all - the contract review runs on an independent Claude skill. The other two, timesheets and resourcing, run on the Projectworks MCP, which is a real product capability.
Reviewing a contract in minutes, not a day
Mark told a story most firm leaders will know too well. One master services agreement ran to about a hundred pages, and every negotiation meeting turned into a three-hour slog because there were so many sections to get through. A general-purpose Claude skill can read a contract that size, pull out the terms that matter, and flag anything odd, which turns a day of reading into a first pass you can act on in minutes. Mark's phrase for the payoff was blunt: "Hopefully this gets you, effectively an agentic contract admin."
Drafting timesheets from your task app
The second example pairs the Projectworks MCP with a task management app like Asana or Jira. It reads the tasks a consultant ticked off during the week and drafts their timesheet against the right budget lines, so the tedious part of remembering and typing it all out mostly goes away. Timesheets are one of the most dependable hiding places for admin cost, because every consultant pays that small tax every week and it adds up fast across a firm. You claw back hours the firm was writing off, and the entries come out more accurate, because they follow what was done rather than what someone half-remembers on a Friday.
Turning meeting notes into a resourcing plan
The third connects the MCP to a meeting transcript. In the demo, the team dropped in notes from a made-up project, the "Mission to Mars" run-through, and asked the AI to turn them into a resourcing plan. Because MCP resourcing reads and writes, it can check who's free and update the plan itself, instead of only reporting back. If you want more examples of how AI can help with meeting admin, check out 18 ways to automate post-meeting admin with AI.
Start where it counts
No single move takes a firm from 15% or 30% G&A down to single digits. It's the pile-up of smaller wins, made where the admin collects. So find the work that's both repetitive and expensive, and start chipping at it one area at a time. Contract review, timesheets, and resourcing are three good places to start, because the effort is low and the tasks keep coming back.
Firms on Projectworks grow by 20% on average in their first year, and cutting admin is part of how.
Listen to moments from similar podcasts
In a small firm, delivery quality holds because everyone works side by side and the founder touches every project. Then the firm grows, the founder runs out of hours, and work gets delivered by people who never learned your delivery method firsthand. Project margins turn inconsistent, and it's rarely the same reason twice. Most AI you've used so far waits for you to ask it something. Agents are different: they work in the background and tell you what needs attention before you'd have caught it yourself. This episode is about what that shift means for delivery, and we'll introduce Kea, the AI agent we're building inside Projectworks, and where we're taking it.
Client acquisition cost is the number no one measures but everyone should. Most firm leaders can quote utilization and margin, then go quiet when you ask what it costs to win a client. We'll work through how to measure it, what healthy looks like, and the leadership question underneath it: who you hire, and what an AI-forward exec team needs to be.
Every proposal, pitch, and scoping call is overhead you carry before a client says yes, and the faster you grow the more it stacks up. In the AI era that's changing, with cost per proposal dropping by as much as 75% at firms using the right tools and workflows. We'll dig into what those leading firms are running, and how to keep this cost from blowing out as you scale.