Services firms can use AI for drafting, meeting summaries, emails and research support. To assess its value, look beyond whether the team uses a tool: measure the work it changes, the checks it requires and the cost of running it.
A useful next step is to review a repeatable workflow, such as onboarding or reporting, and identify where the team repeats the same preparation or formatting work.
This guide explains how to assess a workflow, run a controlled pilot and distinguish time saved from financial returns.
Services firms were built on humans. That’s the whole reason we’re behind.
A software company can ship a feature once and serve a million customers. A creative agency can’t do that. Every brief is bespoke. Every architecture project has its own constraints. Every legal matter has its own facts. For decades, the answer to “how do we serve more clients” was “hire more people.” In services, capacity has historically meant headcount. The unit economics of a services business were humans per project, full stop.
AI can assist with repeatable drafting, summarising and formatting, alongside existing software and automation. Documenting the workflow helps you identify suitable steps and the checks required. Buying a tool alone does not establish that it improves the workflow or pays for itself.
Clear workflows also make measurement easier: you can compare like-for-like work before and after a change.
The bit a lot of founders run into here is the realisation that nobody on the team has actually documented the most-repeated workflow in the firm. Onboarding looks different on every client. Status updates depend on which senior is running the project. Invoicing is “however Sarah does it.” If you can’t write the workflow down, you definitely can’t standardise it, and you can’t hand any part of it to software.
The last time this happened, we called them laptops.
Earlier changes in office technology illustrate a practical point: a new tool changes results only through how people use it. The comparison is an analogy, not evidence for a particular AI adoption timetable or financial return.
For a services firm, the useful questions are concrete: which steps take time, what quality standard must they meet, and can a different process produce the same or better result at a lower total cost?
The result will depend on the firm’s work, systems, clients and review requirements. Test the change rather than assuming that early adoption guarantees an advantage.
Using AI and measuring its value are different tasks.
Here’s the distinction that does most of the work in this piece.
There’s ad-hoc AI, which is speeding up a one-off task. A senior uses ChatGPT to clean up a proposal draft. An associate runs a research summary through Claude. A junior gets help structuring a deck. All of this is fine, and it saves real minutes across a week.
Then there is workflow AI: AI assistance integrated into a repeatable process, with documented inputs, checks and approvals. For example, a reporting workflow might use AI to prepare a draft from verified project records. Measure the full process, including source checks and revisions, before claiming a saving.
Ad-hoc AI vs workflow AI
Two approaches, each requiring measurement
Ad-hoc AI
Speeding up one-off tasks
- What it touches
- Individual tasks
- Time saved
- Measure against baseline
- Review required
- Yes
- Workflow change
- None
- Financial return
- Assess net revenue and costs
Net effect
Task-level result
Include the checking and revision time.
Workflow AI
Assisting repeatable workflow steps
- What it touches
- Whole workflows
- Four-week capacity
- (saved hrs/wk) × 4
- Capacity equivalent
- saved hrs ÷ available hrs
- Review required
- Yes
- Financial return
- Assess net revenue and costs
Net effect
Workflow-level result
Include setup, integration and ongoing review.
Here’s how to size this for any firm, not just the example below.
Plug in your own numbers. The bigger your team and the more repetitive the workflow, the more there may be to gain; measure it.
For a hypothetical fourteen-person firm, assume weekly reporting takes 30 team-hours before a change and 5 afterwards, including checking and rework. Over a four-week planning period, that frees (30 − 5) × 4 = 100 hours. Using an assumed 160 available hours per person for that period, the capacity equivalent is 100 ÷ 160 = 0.625 FTE. These are invented inputs to explain the calculation, not measured results or standard working hours.
The team can use freed capacity for client work, internal improvements or other priorities. Extra revenue requires demand for additional work; a cash saving requires an actual expense reduction. Saved hours alone do not fund the next investment.
The point worth being clear on: this isn’t about replacing the person. The person is still there. They’re now spending those hours on the work the role was actually hired for: strategy, client conversations, judgment calls, the parts that needed a human to begin with. Standardising then automating gives the team back to itself, with the repetitive work taken off their desks.
Measure profit separately. Fewer hours charged to a project can improve its allocated margin while total salary costs stay unchanged. Business profit improves only if extra work generates revenue or actual costs fall, after allowing for AI subscriptions, setup, review and rework. See the Queensland Government guide to financial ratios for the distinction between profit and efficiency measures. Visory Insights provides monthly financial reporting to help you assess business performance.
Reclaimed hours aren’t even the biggest thing. Consistency is.
This is the part that tends to surprise founders the most.
When you standardise a workflow, you also remove variance from how the service actually gets delivered. The output stops depending on which senior was on the project that week. Brand identity work doesn’t drift in quality based on whether your A-player or your B-player drove the brief. Status updates land at the same level of detail every time, regardless of who pulled them together. The financial close looks the same whether your most experienced bookkeeper closed it or someone newer did.
Most professional services firms have always been held hostage by individual brilliance, even if nobody calls it that out loud. The senior who runs onboarding is the only one who really knows how onboarding goes. The architect who’s done forty council submissions is the only one who knows where the traps are. When that person goes on leave, gets sick, or eventually leaves the firm, the workflow degrades and the firm absorbs the cost.
Standardising the workflow, then automating the pattern-matching parts, breaks that dependency. The process no longer lives in one person’s head. Your best people are still doing the strategy, the judgment, and the client conversations. They’re just not also being the single point of failure on every operational detail. That’s a different kind of business, and it’s a more valuable one when you eventually go to sell it or bring on a partner.
Clients buy outcomes. Be open about the method where it matters.
A lot of services founders worry about this one. If I use AI, will my clients feel cheated.
Think about what people actually buy. When someone buys your product or service, they are paying for the outcome. How good does the burger taste. How many leads did I get. How much time did I get back.
Translated to services: the client cares whether the brand identity landed, whether the building got approved, whether the strategy actually worked, whether the case got resolved, whether the books are clean. They care whether you delivered, on time, at the quality they expected. That does not make the method a secret. Be open about AI use where their information, your contract or your professional rules are involved; for tax and BAS work, get the client’s permission (for example in the engagement letter) before client information goes into an AI tool.
It also helps to be clear about what AI is and isn’t doing here. Think about a Michelin-starred chef. Nobody walks into the dining room and tells her she can’t use the sharpest knives on the planet because that’s “cheating” and the knives are doing the work for her. The knives don’t make the food. The chef does. The knives let her do what she does, faster and with less fatigue at the back end of a fourteen-hour service. Your senior strategist still has the strategic instinct. Your architect still has the design judgment. Your lawyer still has the read on the case. The taste, the strategy, the way you deliver, all of that is still yours. AI is just a sharper knife.
There is a real way to lose this: using AI to do worse work faster. The way to win it is using AI to free senior people up to do better work, more often. The competitive question worth asking is whether your outcomes are getting better, faster, and at lower cost than the firm down the street. How AI-forward you look from the outside doesn’t actually matter.
Standardise first. Automate second. The order matters.
This is where most firms get the sequence wrong. They buy AI tools, hand them to the team, and wonder why nothing changes. The tools aren’t the problem. The work they’re pointed at hasn’t been standardised.
Here’s the order that works.
The standardise-then-automate ladder
Five steps. The order is the whole point.
-
Find the workflow that runs through every client
Not the special-snowflake projects. The repeatable spine: onboarding, kickoff, status updates, invoicing, wrap. Pick the one costing the most senior-person hours.
Standardise -
Document it as it actually runs today
Six to twelve discrete steps. Inputs and outputs for each one. The part most firms skip, and the part that decides whether AI adds value or just adds noise.
Standardise -
Separate pattern-matching from judgment
Pattern-matching: drafting, summarising, formatting, comparing, classifying. Judgment: strategy calls, client conversations, sign-off, creative direction.
Standardise -
Apply AI only to the pattern-matching steps
Keep humans on the judgment steps and have a person review every AI output before it goes anywhere. The senior is still in the workflow; they’re just not doing the formatting bit anymore.
Automate -
Measure cycle time before and after
Set a target before you start. If it doesn’t, find out why (standardisation, tool fit or review effort) before going further.
Automate
Step 1. Find the workflow that runs through every client.
Look at the repeatable spine of the firm: onboarding, project kickoff, status updates, draft review cycles, invoicing, post-project wrap. Skip the special-snowflake projects for now. Pick the workflow that costs the most senior-person hours, every week.
Step 2. Document the workflow as it actually runs today, not as you wish it ran.
Six to twelve discrete steps. Inputs and outputs for each one. This is the part most firms skip, and it’s the part that decides whether AI adds value or just adds noise.
Step 3. Identify which steps are pattern-matching and which need human judgment.
Pattern-matching looks like drafting, summarising, formatting, comparing, classifying, extracting. Judgment looks like strategy calls, client conversations, sign-off, creative direction, anything where a senior is weighing context the system can’t see.
Step 4. Apply AI only to the pattern-matching steps. Keep humans on the judgment steps.
This is where the not-replacing-people principle becomes operational rather than a slogan. The senior is still in the workflow. The senior just isn’t doing the formatting bit anymore.
Three guardrails make this safe. A named person reviews and signs off every AI output before it reaches a client, a supplier or the ATO. Exceptions (anything unusual, incomplete or outside the documented steps) go to a person rather than through the automated path. And you decide up front what data the AI tool may see: check the tool’s terms, keep client confidential information inside tools your contracts and privacy obligations allow, and don’t paste it into consumer apps. Where tax or BAS work is involved, the Tax Practitioners Board’s guidance on AI is clear that registered agents remain responsible for the work, must review AI output, and should obtain client permission before entering client information into AI tools. Lawyers have their own rules on generative AI, such as the NSW Supreme Court’s Practice Note SC Gen 23.
Step 5. Measure cycle time before and after.
Record the time for a complete cycle, including review and rework, and set targets based on the workflow’s cost and quality requirements. Compare results after the pilot. Investigate errors, extra review effort or poor economics before expanding it.
Work through one workflow at a time. Expand when the measured results justify the cost and the team can maintain quality; there is no fixed number of workflows after which financial returns begin.
What this actually does to the numbers.
Track three outcomes separately; none follows automatically from the others.
Three outcomes to measure
Assess each outcome separately.
-
Capacity
Measure hours saved after checking and rework. Salary costs may stay unchanged.
-
Profit
Count extra revenue or actual cost reductions, less implementation and ongoing costs.
-
Cash
Track billing lag, collection time and cash receipts separately.
Measure capacity. Compare total preparation, checking and rework hours before and after the change. Freed salaried hours create capacity; they do not reduce payroll expense by themselves.
Measure profit separately. Fewer hours charged to a project can improve its allocated margin while total salary costs stay unchanged. Business profit improves only if extra work generates revenue or actual costs fall, after allowing for AI subscriptions, setup, review and rework. See the Queensland Government guide to financial ratios for the distinction between profit and efficiency measures.
Measure billing and collection separately. Earlier invoicing can shorten the time from completing work to receiving cash, if payment timing otherwise stays the same. It does not by itself shorten the invoice-to-payment interval or guarantee lower debtor days. Track billing lag, payment terms, overdue balances and actual cash receipts.
Then there are the second-order effects most founders don’t price in until they show up. The same standardise-then-automate move applied to non-client workflows (HR, finance, internal sales and marketing ops, recruiting) reclaims another tranche of hours that was never on a billable timesheet but was real cost all the same. And once a few workflows are running this way, you may find software you used to need can come off the bill entirely. A project management tool that was holding together a non-standard process. A reporting tool whose output now gets produced upstream. An automation platform whose use case got absorbed into the workflow itself. Check your software subscriptions after each workflow you standardise; those savings can go some way to covering the cost of the AI tools.
Standardisation makes the process clearer and easier to assess. Keep changes that improve quality or economics against the targets you set, and report capacity, profit and cash outcomes separately.
A quick word on what happens if you don’t move on this.
We said we’d cover the risks. Here they are.
Clients may start expecting faster turnaround as a baseline. Firms that haven’t moved may start to feel slow without quite knowing why.
Senior people may leave for firms where AI has cleared the repetitive work off their desks, and recruitment costs may rise.
Smaller, AI-native firms may start competing at price points that used to be impossible. The squeeze would come from below.
Organisational change takes sustained work. The time required depends on the workflows, systems and people involved; use measured progress rather than a fixed timetable for returns.
What we’d do Monday morning
Pick the workflow that costs you the most senior hours. Write down the actual steps. Mark the pattern-matching ones. Standardise them this month. Automate the pattern-matching parts next month. Measure the cycle time. Move to the next workflow.
Repeat the pilot-and-measure approach when the results justify it. Do not assume that a fixed number of workflows or a year of adoption will produce a particular financial return.
If you’re running a professional services firm and you’ve never mapped your most-repeated workflow end to end, start there. Forget the new AI tools and the next consultant pitch for a minute. The unglamorous part is what pays: writing down how the firm actually runs, so you can decide which steps deserve a human and which ones don’t.
Book a discovery call and we’ll walk through which workflow in your firm is costing the most, and where standardising it would show up first in growth, profit, or cash.
Frequently asked questions
Do I need to standardise every workflow before I touch AI?
No. Start with the highest-cost repeatable workflow, the one eating the most senior hours. Get that one right, then move to the next. The order matters: workflows that run through every client (onboarding, status reports, project handover, invoicing) before workflows that only run occasionally (RFP responses, partner reviews). To see where your own numbers stand before you start, get your free financial health check.
Won’t my clients feel cheated if they find out I used AI?
Clients buy outcomes, not methods. They care whether the brand identity landed, whether the strategy actually worked, whether the books are clean, on time, at the quality they expected. The risk isn’t using AI; it’s using it to do worse work faster, or putting their information somewhere they didn’t agree to. Be open about AI use where their information, your contract or your professional rules are involved; for tax and BAS work, get the client’s permission (for example in the engagement letter) before client information goes into an AI tool. Keep a person accountable for everything that goes out, and use AI to free senior people up to do better work, more often.
What’s the difference between “ad-hoc AI” and “workflow AI”?
Ad-hoc AI supports an individual task, such as preparing a draft. Workflow AI integrates assistance into a repeatable process with defined inputs and approvals. Either approach can help or add cost. Measure total effort and output quality, then assess whether any saved time produces additional revenue or lower actual expenses.
How long before this shows up in the financials?
There is no verified timetable for these outcomes. Measure hours saved after review and rework, then check whether they produce extra revenue or lower actual costs after implementation costs. Track billing lag and invoice collection separately: earlier invoicing alone does not mean customers pay invoices faster.
Should I hire an “AI lead” to run this?
Usually no. The person who already runs the workflow is the right person to standardise it. They know where the friction lives, where the exceptions hide, and where the judgment calls happen. Bring in outside help for the automation step (a contractor or specialist who can wire AI into the pattern-matching steps cleanly), but the standardisation work has to come from inside the team. Firms that try to outsource the standardisation phase usually end up with a beautiful workflow document that nobody actually follows.
About this article
Written by Jordan Bevans. Published . General information for Australian businesses, not advice for your situation.
