Blog
AI Made You Faster. Should You Charge Clients Less?
August 26, 2026 · 13 min read time
TL;DR: AI does not create extra hours. Bill hourly clients only for the time you actually spend on their work. If a client is buying a fixed result, reserved capacity, or another agreed unit, price that unit clearly and track your internal time so you can tell whether AI improved your margin.
AI has created a strange pricing problem for freelancers.
A task that used to take a day might now take an afternoon. A developer can ask one agent to investigate a bug while another drafts tests. A designer can generate several directions while reviewing a different client's work. A consultant can have research running in the background while preparing a workshop.
Sometimes, two client projects are moving forward during the same hour.
So what exactly are you charging for?
Is it the elapsed time? Your attention? The AI tool's runtime? The outcome? And if two projects move forward between 10:00 and 11:00, can both clients be billed for an hour?
The clean answer starts with the agreement you made. Hourly billing, fixed-price work, retainers, and outcome-based fees sell different things. AI has not made those differences disappear. It has made them harder to ignore.
AI-assisted work is no longer a niche edge case
Australian business use of AI rose from 1% in 2021-22 to 12% in 2024-25, according to the Australian Bureau of Statistics. In professional, scientific, and technical services, the figure was 24%.[1]
Freelance marketplaces are seeing the shift too. Upwork reported that skills explicitly referring to AI grew 109% year over year in 2025.[2] Its July 2026 research found that freelancers performing AI work on the platform earned 34% more per hour than freelancers who were not incorporating AI.[3]
But the same research found an important split. Earnings for more complex AI work increased 45% year over year, while generative AI and creative production recorded strong growth in contract starts but a 13% fall in earnings per contract.[3] Faster execution alone was not where the strongest value appeared. Expertise, judgement, integration, and accountability mattered.
That is the pricing tension in one paragraph.
AI can make some work faster. It can also make basic execution easier to compare and cheaper to buy. The freelancer's job is to understand which part of the service the client is actually paying for.
Do not assume AI made you faster
Before changing how you charge, check whether the speed increase is real.
The 2026 Stanford AI Index found that AI productivity gains vary considerably by task. Studies cited in the report found gains of 14% to 15% in customer support, 26% in software development, and 50% in marketing output. The same review found smaller gains in work requiring deeper reasoning.[4]
In a separate randomised trial, METR found that experienced open-source developers took 19% longer with early-2025 AI tools, even though they believed AI had made them faster.[5]
In a February 2026 update, METR said developers were probably seeing greater benefits from newer tools, but selection effects made its follow-up data too unreliable to quantify the change confidently.[6]
AI can save time on one part of a job while creating new work elsewhere:
writing clearer instructions and acceptance criteria
checking sources and generated claims
reviewing code, designs, or documents
correcting plausible but wrong output
integrating the result into a larger project
taking responsibility when the finished work reaches the client
That means "AI was involved" is not a useful unit of measurement.
Track the work first. Compare similar projects. Then decide whether the tool reduced delivery time, increased output, improved quality, or simply moved effort from creation to supervision.
If you do not have a baseline yet, a 30-day time-tracking sample can give you a practical one.
The three clocks behind parallel work
When several projects move at once, it helps to separate three different clocks.
1. Wall-clock time
This is ordinary elapsed time. From 10:00 to 11:00, one hour passed.
2. Human attention time
This is the time you spent briefing, thinking, writing, reviewing, correcting, communicating, and making decisions for a client. Your attention can switch quickly, but it does not become two full hours merely because two tools were running.
3. Tool or processing time
This is the time an AI agent, render, data job, test suite, or automation ran. It may be commercially relevant, but it is not automatically the same thing as an hour of your labour.
The billing model should make clear which clock, unit, or result the client is buying.
If the invoice says "one hour of consulting", the natural meaning is one hour of consulting work. If the agreement instead charges per processed document, completed workflow, generated asset, compute unit, or project milestone, the elapsed time is not the unit being sold.
Confusion appears when a freelancer sells one unit but invoices as though they sold another.
Can you bill two clients during the same hour?
If both clients are paying for your time by the hour, the safest default is no: one human hour should add up to one hour of billed labour across those hourly clients.
Suppose an AI agent works on Client A's project from 10:00 to 11:00. During that hour, you spend:
10 minutes briefing and starting the task for Client A
35 minutes actively working on Client B's project
15 minutes reviewing and correcting the result for Client A
Your human attention record is 25 minutes for Client A and 35 minutes for Client B. The agent may have run for 35 minutes in the background, but that does not turn the hour into 95 minutes of your labour.
If the Client A agreement separately defines a processing fee, per-run charge, or other unit, that charge can be recorded as the agreed service or expense. It should not be relabelled as 50 minutes of human work.
Regulated professions can have stricter rules. For example, the American Bar Association's formal guidance says that lawyers billing hourly for AI-assisted work must bill their actual time. Where a firm separately charges for an AI tool developed in-house, the guidance also says that amount must not duplicate other charges to that client or others.[7] That opinion governs a particular profession and is not a universal rule for every freelancer, but it illustrates why the basis of the charge needs to be explicit.
Your own contract, profession, jurisdiction, client policies, and confidentiality obligations may add different requirements. When the arrangement is unclear, agree on the billing method before the parallel work begins rather than trying to explain it after the invoice arrives.
Fixed-price work changes the calculation
Fixed-price work does not sell a running clock. It sells an agreed scope, milestone, deliverable, or result for an agreed price.
If Client A agrees to pay $2,000 for a defined deliverable and Client B agrees to pay $1,500 for another, completing parts of both projects in parallel does not automatically reduce either agreed price. The efficiency gain can belong to the freelancer, just as the risk of underestimating the work often does.
That does not mean every fixed fee remains reasonable forever. Competition changes, tools change, and clients become more aware of what work involves. A sustainable price still needs to cover costs, reflect the value delivered, and make sense in the market.
Australian Government small-business guidance recognises hourly, cost-plus, value, and other pricing strategies, and notes that businesses can combine more than one. Its central advice is that the price should cover costs and reflect the service's full value.[8]
This is where AI can reward a well-run freelance business. If you can deliver the agreed result faster without reducing quality, missing scope, exposing client information, or overloading your review capacity, your effective hourly return can improve.
You should still track your time internally.
Without that record, you cannot tell whether the $2,000 project took 8 hours or 28, whether parallel work improved your margin, or whether the AI tool merely hid extra review and correction time.
What about retainers, day rates, and reserved capacity?
Retainers are where parallel work can become ambiguous.
One retainer may buy a set of monthly outcomes. Another may buy up to 20 hours. Another may reserve your availability on particular days. Those are not equivalent promises.
If the client is paying for exclusive or reserved capacity, quietly reselling the same capacity may conflict with the agreement even when the work is light. If the retainer covers defined outputs or a response time, parallel work may be completely normal as long as those commitments are met.
Write down what the client is reserving:
a number of hours
specific days or availability windows
a response time
a defined set of deliverables
access to a fixed amount of support
The word "retainer" is not enough. The unit underneath it matters.
Day rates need the same clarity. A day rate can mean exclusive access to your working day, or it can simply be the price for a defined day-sized service. Do not assume that the label answers the concurrency question.
A hybrid model often fits AI-assisted work best
Many freelancers do not need to choose between hourly and fixed pricing for every part of a project.
A hybrid structure can separate the predictable work from the unpredictable edges:
a fixed fee for the agreed deliverable or milestone
an hourly rate for meetings, support, and out-of-scope changes
a defined price per repeated output or automated workflow
separately agreed pass-through costs for unusual client-specific processing
That gives the client price certainty around the main result while protecting the freelancer when the scope changes. It also makes parallel work easier to explain. The project fee pays for the deliverable; the hourly component pays for actual additional attention; any processing charge pays for the clearly named unit.
The same rule applies to small client requests and revisions: confirm what sits outside the fixed scope before the extra work begins.
Avoid vague "AI fees" added at invoice time. If a tool cost or automated service will be charged separately, explain the basis beforehand. In Australia, pricing information must be clear, accurate, and not misleading.[9]
How to track parallel projects without double-tracking time
You do not need two live labour timers. You need records that match the commercial model.
For hourly work:
Run one timer for the client receiving your active attention.
Switch when your attention moves to another client's task.
Include briefing, review, correction, and client communication where those activities are billable under the agreement.
Record automated runtime separately only when it is useful operationally or is a separately agreed billing unit.
For fixed-price work:
Track your internal time against the correct project even though the client will receive a fixed invoice line.
Invoice the agreed milestone, deliverable, or amount rather than converting the work back into invented hours.
Review the effective hourly return after delivery: project revenue divided by the time you actually invested.
For hybrid work:
Keep fixed-scope delivery separate from hourly extras.
Confirm out-of-scope work before starting it.
Use invoice descriptions that make the distinction obvious to the client.
You do not need to account for every second. You need a timely record you can explain. A short note such as "briefed agent and reviewed export" is more useful than a perfect-looking hour with no explanation.
Where DoubleTime fits
DoubleTime keeps one live timer active so your labour record follows your current attention rather than creating overlapping human hours. When you move to another task, the active record moves with you.
That is only one part of the billing picture.
You can keep work organised by client and project, capture the time you invest in fixed-price work for internal analysis, and create invoices from tracked work or manual line items. That means a fixed project fee can remain a fixed invoice line without losing the time record you need to check its profitability.
The goal is not to force every service into hourly billing.
It is to keep the work, the pricing decision, and the invoice connected clearly enough that you know what happened and the client knows what they are paying for.
A simple rule for the AI era
Invoice according to the agreement. Track time to understand the business.
If you sell hours, bill the hours you actually worked and allocate them honestly across clients.
If you sell a defined result, charge the agreed price and let safe efficiency improve your margin.
If you sell availability, processing, usage, or repeated outputs, name the unit and set the rule before the work starts.
AI can let more projects move at once. It does not make double-billing honest.
Sources
[1] Australian Bureau of Statistics - Characteristics of Australian Business, 2024-25
View the ABS release
[2] Upwork Research Institute - In-Demand Skills 2026
View the Upwork report
[3] Upwork Research Institute - Future Workforce Index 2026
View the Upwork report
[4] Stanford Institute for Human-Centered Artificial Intelligence - 2026 AI Index Report,
Economy
View the Stanford AI Index chapter
[5] METR - Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer
Productivity
Read the METR study
[6] METR - We are Changing our Developer Productivity Experiment Design
Read the METR follow-up
[7] American Bar Association - Formal Opinion 512: Generative Artificial Intelligence Tools
Read Formal Opinion 512
[8] business.gov.au - Choose a pricing strategy
Read the Australian Government guide
[9] Australian Competition and Consumer Commission - Price displays
Read the ACCC pricing guidance