Golden Sky AI

How do I measure the time AI actually saves my firm?

Record each task's hours before AI goes in, measure the same tasks after with the review time counted, then track what the freed hours turn into.

Golden Sky AI12 min read

You measure the time AI saves by recording, before any AI goes in, the hours each person spends on each task and what those hours bill at, then measuring the same tasks once AI is helping, with the time spent checking its work counted in. The saving only counts once you can see what the freed hours turned into, such as more clients or more advisory work.

Why can't you just ask your team how much time AI saves?

You can ask, but what comes back is a guess, and guesses about AI tend to run high. In Financial Cents' survey of 486 people in accounting and bookkeeping, run in July and August 2026, the time saved came from each person's own estimate, and the biggest group, 37%, put it at 1 to 3 hours a week. Only 20% of the firms using AI had seen clear, measurable return, while 52% called it promising but hard to quantify (2026 survey). Thomson Reuters found that only 19% of tax firms measure the return on their AI tools at all (Thomson Reuters, April 16, 2026).

The feeling and the clock can also disagree. In a 2025 trial run by METR, experienced software developers took 19% longer on real tasks when AI was allowed, yet afterward believed it had sped them up by 20% (METR, July 10, 2025). That was software work on early-2025 tools, and METR now calls the result out of date, but it shows that someone can sincerely feel faster and be slower.

So we don't treat a guess as a measurement. The time AI saves gets measured in three places inside a bigger process: the hours each task takes today, from mapping the work, the hours each solution should free up, from costing it, and the hours it actually frees up, which the operating model shows once it's live. Each depends on the parts before it, so a firm that's serious about AI goes through the whole process in order.

To see where AI would free up the most hours in your own firm before you measure anything, take the ten-question quiz below.

What has to be in place before you can measure anything?

The firm has to be compatible with AI first. We check its AI layers, which are the business as a whole, every department, the data and tools those departments run on and the people in them, then fix whatever falls short in the tech stack, data quality or data security. For measuring, the weak spot is usually the time data, because plenty of firms time some work and not the rest, fixed-fee work often isn't timed at all and the task codes rarely match what people actually do. If that data isn't clean, complete, and queryable, any number you put on AI rests on wrong information. Nobody can measure work done on a free personal account either, and IRS guidance issued in June 2026 says client data should be handled using only secure, enterprise-approved AI.

Then we map the work. Through short intakes, each person tells us every task they handle on a repeat basis, how long it takes in a normal week and in busy season, what their time bills at, which client data and software it uses, and how they use and check AI on it today. That map is the baseline everything gets measured against, and it has to be taken before anything changes, because once AI is in, nobody remembers how long the old way took. It also picks up what AI now makes possible for clients that the firm couldn't offer before, which shows up as new revenue and gets measured on its own line. The foundations under all of this are the six layers a firm needs right before any AI goes in.

How do you work out what a task's saved hours are worth?

Every solution gets costed before anyone implements it, so you know the hours it should free up, what it costs to put in and its return in the first year. We take the total time someone spends on the task and multiply it by 0.7 to 0.85, because the person who owns it still has to read and correct what the AI produces. Most time-saved figures leave that review out, and the IRS won't let you skip it, since its June guidance says practitioners "cannot rely solely on AI" and must verify the facts, citations and calculations it gives them, as the Journal of Accountancy reported in June.

Here's the method on made-up numbers, so this isn't a client result. A senior on the client accounting team bills $150 an hour and spends 4.5 hours a week writing the monthly notes that explain each client's numbers. They're spending 4.5 hours on this task today, and at 0.7 to 0.85 a solution unlocks 3.1 to 3.8 hours of their time each week. Over 48 working weeks that comes to about 151 to 184 hours, or $22,680 to $27,540 of billable time a year, which we set against what the solution costs to implement and run. The same sum runs for every person and everything they do, so the solutions can be ranked.

Measuring afterward uses the same method. If those notes take 1.2 hours a week once the solution is live, including the senior's review, it has unlocked 3.3 hours a week, inside the range the costing predicted. If they take 2 hours, that's the cue to find out why.

We also check each solution's compliance, governance and data security against the rules of every regulator and professional body the firm answers to, and that check can change the numbers, because work on full returns runs on your own machine or in your own cloud account. There are no nasty surprises after implementation. You know what each solution costs, what it gives back and what the risks are before it goes in.

If you'd like a first estimate of where those hours sit in your own firm, the ten questions in the quiz below will give you one.

What should the hours AI saves turn into?

Most firms never get this far, and it's the bit a partner actually cares about. We always ask what the freed hours will become, because capacity that doesn't mean anything for the business isn't worth unlocking. When Financial Cents asked 330 owners and partners for their top priority, 43% said capacity, meaning doing more with the same team, well ahead of the 31% who chose service quality and client experience (same 2026 survey).

So before any solution goes in, the partners decide what each team's freed hours are for, such as taking on the clients you've had to say no to or moving people onto advisory work. That way the same people take on more billable work without the firm having to hire. Those decisions shape the roadmap, which puts the solutions in order of what they're worth, under the firm's own AI policy.

Watch your realization rate as well, which UK firms call the recovery rate. You set each job's fee from a budget of hours, and when the job runs over, the extra chargeable time gets written off. So measure the hours on each job against its budget before and after a solution goes in. If AI gets the work done on budget or quicker, you stop writing that time off and recover every hour you bill, which lifts revenue and gross profit, and the hours won back go to the next client with the team you already have.

On tax work, measuring properly also protects the bill. The IRS guidance warns that charging for time AI meant you didn't spend could raise concerns about "an unconscionable fee," and says firms should "fairly credit to the client's account any cost reductions" (JofA, June 26, 2026). The AICPA has said that language doesn't appear to allow for value pricing, and on November 9 the Journal of Accountancy reported that the director of the IRS Office of Professional Responsibility, speaking in a personal capacity, expects more guidance on AI and fees "in the next couple of months" (JofA, Nov. 9, 2026). Nobody knows yet where the IRS will land on billing, but a firm that has measured its hours can show what changed and why. When a client, your insurer or a regulator asks how you use AI, you've got a clear answer ready.

If write-offs are what worry your partners most, we've set out how AI can lift a CPA firm's realization rate by keeping each job within its budget.

How do you keep measuring once AI is live?

Ahead of any implementation, the firm needs an operating model, a running view of where AI is doing work in the firm, for which people, and the hours each team has unlocked against its costing. From then on, that's where the measuring happens.

The solutions then go in one at a time, in roadmap order, each with an enablement pack for everyone using it that covers how they record the task now. Every solution has a named owner, because we still need experts to be able to approve things, and the owner's review time goes into the numbers each month. Once a solution is live, every output is logged, queryable and observable, so you can see how often it runs, how much the owner corrects and whether the hours are holding up.

That matters because a solution that unlocked 3.3 hours a week in the spring can drift when Claude or ChatGPT updates the model underneath it, so we keep every solution up to date whenever a model or a rule changes. The IRS guidance also expects AI accuracy monitoring and a documented record of how the firm uses and checks AI, and a monthly measure of each solution gives you that record. It's the point an accountant who runs a firm on AI made to us, that running AI means managing it.

How much work is it to measure this across a whole firm?

It's a lot, and it's better you hear that now. Someone has to sit down with every person in the firm, from the partners to the admin team, and get every task they do, its hours in a normal month and in busy season, their rate and whether AI touches it yet. The data and tools in each department need checking too, with the time records fixed wherever they're incomplete or coded in a way nobody can query. Every solution needs its costing, with the review time in it, and a check against Circular 230, section 7216, the FTC's Safeguards Rule, your state board and anything else the firm answers to.

Then the roadmap has to be put in order, the operating model set up before anything goes live, and each solution measured against its costing every month once it's running. The partners who could do all that are usually the ones whose hours bill the most. When we do it, every solution is built on the real tasks and processes your firm, your departments and your people do day to day, so it fits exactly how you work, and each of your people gives a few minutes to their intake.

If you want us to do it for you, book a free call and we'll find you at least $50,000 a year in new business that AI can unlock. It's compliant, data-secure and has governance built in.

What can you check this week?

  • Pull a month of time entries from your practice management software and see how much of it is coded clearly enough to tell which task it was.
  • Find out which AI tools your people use with client data, personal and free accounts included, because work on an account the firm can't see can't be measured.
  • Get the partners to agree on what extra capacity in each team is for.

Before you go pulling time records, the quiz below will tell you in ten questions where AI is likely to be worth the most to your firm.

Common questions

Can we measure AI on work we bill as a fixed fee?

Yes, and it's often where the saving matters most, because on a fixed fee every hour AI frees up goes to the firm's margin or its capacity. Nobody timed that work before, so the intake asks people how long each task takes, and the team times it for a few weeks to check.

Should we trust the time-saved figures that AI vendors publish?

Use them as a reason to ask questions. A vendor's figure comes from other firms' work, usually without the review time, and the number worth putting in front of your partners is one built from your own people's tasks and rates.

Does this mean tracking every minute our people work?

No, it doesn't. You measure the tasks that have a solution against them, the same way they were recorded in the baseline, and the operating model shows totals by team, so nobody needs to feel watched.

We've written this as general information, and it isn't legal or tax advice, so talk to your own attorney about how the IRS and FTC rules apply to your firm.

Work with us to get a real measure of the capacity AI can unlock in your firm

At Golden Sky AI, we help companies in regulated industries implement AI that frees up their team's capacity, so they can take on more business without having to hire. Every solution is checked for compliance, governance and data security before we implement it, and you own the intelligence, meaning any AI agent could plug into the solutions we implement, as they're robust and still deliver the same quality of work. That means you're never held hostage by one AI provider.

We'll find you $50,000 a year in new business AI can unlock in 72 hours, or you don't pay.

Sources

Field notes

Get the field notes.

What is working with AI inside professional service firms, written from the buyer’s side.

Sent when there is something worth reading, not on a schedule.

One email when there’s something worth reading. Unsubscribe any time. See our privacy notice.

Find out what AI would earn your firm, risk-free.

Book a call and we'll show you where AI can unlock $50,000 a year in new business for your firm.

Book a call

See where you rank