How to scale an IT consultancy without hiring more staff

An IT consultancy adds billable capacity without hiring by giving the non-billable admin around each engagement to an AI worker: the statement of work drafted from discovery notes, the status report, the timesheet chase. The ratio that matters is the share of each consultant's week that a client pays for, and because the firm's product is its people's hours, every hour spent on that admin is an hour the firm bought and could not sell. The estimate, the price, the scope decisions and the code stay with your people.

Why is utilisation the number, and why does it drift?

Harvest sells its scheduling product on balancing workloads to improve utilisation and on seeing when you can take on more work and when you need to staff up. Staffing up is the usual answer, either with someone non-billable to absorb the admin, a project manager or a delivery manager, or with another consultant, whose week will also be part admin, and either way headcount grows ahead of margin.

Utilisation drifts for a reason that has nothing to do with how good the consultants are. Each new engagement adds a fixed packet of work that is not the engagement: a statement of work to write, a weekly or sprint report to assemble, a support address to watch once the system is live, hours to log and get approved before the invoice can go, and a proposal to follow up while the current sprint is on, and five engagements bring five of each. The consultants do it because there is nobody else, and late, so month-end invoices can leave late too.

What work grows with every engagement?

Look at a time and materials job from the client's yes to the first invoice. Someone turns discovery notes into a statement of work: the work, the people, the timeframe and the cost. In a small firm that may be a director, starting from an old one. Once work starts, someone assembles the sprint report from Jira, adds hours against budget, and sends it. When the client asks for more, someone turns an email thread into a change request, so the extra days are agreed before they are worked. And at month end, someone chases the timesheets. In Harvest a time entry is unsubmitted, submitted or approved, and a firm that invoices on approved hours cannot bill a week nobody has submitted. Harvest's own reminders go a few hours before the deadline and again a day later, and still the hours arrive late, for the reason Harvest itself gives: people forget timesheets when they are busy on meaningful work.

Every item on that list is done by someone whose hours are the product, or by someone hired so that it is not. The IT consultancies and software houses page names the tasks.

What stays with your people?

Your people keep the estimate and the price, which are the firm's judgement about its own capacity and risk, and the decision to send a proposal or a statement of work. They decide whether a support request is a bug under the agreement or chargeable new work, which is commercial, not clerical. They approve hours and make every correction to them; the agent never enters or changes time. The code, the repositories, the pipelines and every client environment stay with them, and the agent neither needs nor gets them. So does any conversation with the client about scope, price or terms.

What can an AI worker take on?

An AI worker works to a written specification, answers to a named person and waits for approval wherever you require it. Each of the two roles here is written up with you first, and works through the route the maker of your time system or work tracker publishes, confirmed with you at the start.

The first in a consultancy is timesheet chasing, because on time and materials nothing is billed until the hours are in. It would hold the calendar of cut-offs and invoice runs, find the missing and part-filled weeks in the time system, remind each person at the intervals you choose, query entries that break your rules, and tell the delivery manager who has still not submitted and for how long.

The second is quote and proposal preparation, aimed at statements of work. It would build the draft from the discovery notes, in your template, with your rate card and standard terms, so the director's time goes on the estimate and not on the document.

Measure the week before changing it

For one month, have every consultant log non-billable time against three codes: writing documents for sale, reporting to clients, and chasing or being chased. The total is what you are trying to recover, and the number to look at again in six months. Then start with timesheets, because the rules are already written down in your invoice run. Add statements of work when a director is ready to edit a draft rather than start from the last one.

Firms outside IT will find the same ratio worked through in how to scale a business without hiring more staff; what an AI worker is, and why it is managed rather than installed, is in our explainer on AI staff, and the homepage carries our starting price. A software house may ask why it should not build this itself. It could, and it would be one more internal project competing with client work for the same people, which is the ratio this article is about.

In an IT consultancy or a software house the worker's share would be the back-office chasing and drafting, never the engineering, the estimate or the price. Talk to our agent about the timesheet chase before you staff up for the next engagement.