Essay

Productized services: the data on selling work, not hours, as AI speeds up the hour

Published · Updated 16 min read

Productized services are services sold the way products are sold: a fixed scope, a fixed price, a fixed timeline and a named result, offered in the same form to many buyers. They are the opposite of billing by the hour, where the buyer pays for time and the seller's income rises only with the hours worked. The hourly model is younger than most people think. It was popularised among US lawyers in the 1950s, after an American Bar Association pamphlet, "The 1958 Lawyer and His 1938 Dollar," steered the profession away from fixed fees and toward units of time, and it later spread to accountants.[1]

This essay brings together three bodies of evidence. The first is experimental and field data on how much generative AI speeds up skilled work, from three peer-reviewed or widely cited studies. The second is official US data on what an hour of professional work costs and what clients have paid for professional services since 2019. The third is our own analysis of that public data, which shows that prices for advertising agency services have fallen well behind inflation while prices for other professional services have not. We close with a framework, the Hours-to-Product Ladder, and a six-step method to package a service.

Idea in brief

The problem. In controlled studies, AI cut the time skilled people needed for a task by 40% in professional writing and 55.8% in a coding task.[2][4] If you bill by the hour, every one of those minutes comes off your invoice.

Why it happens. Hourly billing ties income to time spent, not to what the client gets. Any gain in speed, from experience or from tools, flows to the client as a smaller bill.

The gap (our data). Between September 2019 and September 2026, prices received by US advertising agencies rose 9.4%, against 31.1% for consumer prices. Prices at law offices rose 50.4% over the same period.[5][6]

The answer. Move repeatable work up the Hours-to-Product Ladder: from time, to fixed projects, to named packages, and, where the result is large and measurable, to value pricing. Keep the hour for work nobody can define in advance.

Key findings

The essay in 2 minutes
  1. Finding 140%

    AI cuts task time sharply in controlled tests.

    Professionals given ChatGPT finished occupation-specific writing tasks in 40% less time, with output quality rated 18% higher.[2] Developers given an AI pair programmer finished a coding task 55.8% faster.[4]

  2. Finding 215%

    In real work, the gain is smaller but real.

    Across 5,172 customer support agents at one firm, an AI assistant raised issues resolved per hour by 15% on average and cut average chat time by 8.5%.[3]

  3. Finding 336%

    The least experienced gain most.

    The lowest-skilled fifth of support agents resolved 36% more issues per hour with AI. The most skilled fifth saw no increase.[3]

  4. Finding 4−17%

    Agency prices lagged inflation.

    The producer price index for US advertising agencies rose 9.4% from September 2019 to September 2026. Consumer prices rose 31.1%, so agency prices fell by about 17% in real terms.[5][6]

  5. Finding 550.4%

    Other professions held price.

    Over the same period, prices at law offices rose 50.4%, at accounting firms 23.8% and in management consulting 16.2%.[5]

  6. Finding 6$80.19

    An hour of marketing work has a clear market rate.

    In May 2025 the median US wage was $80.19 an hour for marketing managers, $37.87 for market research analysts and marketing specialists, and $30.27 for graphic designers.[7]

Chapter 1 · 1 min

What productized services are

A productized service takes a job a firm does often and sells it like a product. The buyer sees what they get, what it costs and when it arrives before anyone talks about hours. Common examples include a website audit, a monthly bookkeeping package, an equipment inspection program or a market-entry research pack.

Terms used in this essay

TermMeaning
Productized serviceA service with a fixed scope, fixed price, fixed timeline and named result, sold in the same form to many buyers.
Billable hourA unit of time charged to a client at an agreed rate. Popularised in US law in the 1950s and later adopted by accountants.[1]
Fixed fee pricingOne price for a defined piece of work, whatever the hours it takes.
Value pricingA price set from what the result is worth to the buyer. See the companion essay on price versus value.
Scope creepWork added to a fixed-fee project without a change in price. The main risk of fixed fees.
Effective hourly rateFee received divided by the hours worked. It rises when a fixed-fee job gets faster.

Hourly, fixed fee and value pricing compared

Hourly pricingFixed fee pricingValue pricing
What the client buysYour timeA defined piece of workA result and what it is worth to them
Who carries the risk of overrunsThe clientYouYou, mostly
Reward for getting fasterLower payHigher marginHigher margin
Easy to compare with rivalsVerySomewhatHard
Best forOpen-ended, unclear workRepeatable work you know wellWork with a large, measurable result
Main dangerIncome ceiling, timesheet disputesScope creepOverreaching on a value you cannot show

Chapter 2 · 1 min

Where the billable hour came from

The billable hour was a deliberate choice, made for one profession in one era. A 2013 ABA Journal report, summarising a New York Times Magazine story, quotes the story's account of how it began: "The notion of charging by units of time was popularized in the 1950s, when the American Bar Association was becoming alarmed that the income of lawyers was falling precipitously behind that of doctors (and, worse, dentists)."[1]

According to that account, the ABA's pamphlet "suggested that the industry should eschew fixed-rate fees and replicate the profitable efficiencies of mass-production manufacturing," with the hour as the unit. It was then "adopted by countless other service professionals, including accountants."[1] The same report described a group of accountants who had dropped the billable hour in favour of fees based on the value of their advice to specific kinds of clients, such as owners preparing to sell a business. Their complaints were the pressure to work long days and the incentive to favour long projects over short tasks that need special knowledge.[1]

The hour was a pricing tool built to raise lawyers' incomes in the 1950s. It is not a law of nature, and a firm is free to choose something else.

Chapter 3 · 2 min

The evidence, part 1: AI makes skilled work faster

Three studies give the clearest numbers on how much generative AI speeds up skilled work. They differ in method, task and setting, which is useful: together they show a range rather than a single point.

Figure 1. How much less time a task took with AI

Peng et al. 2023 (95 people)
55.8%
Noy and Zhang 2023, Science (453 people)
40.0%
Brynjolfsson, Li and Raymond 2024 (5,172 people)
8.5%

Reduction in time per task (%), 0 to 60

Tasks: write an HTTP server in JavaScript; occupation-specific professional writing; customer support chats.

Reduction in time to complete a task for people with access to an AI tool, compared with people without it. Sources: Peng, Kalliamvakou, Cihon and Demirer, arXiv 2302.06590; Noy and Zhang, Science, 2023; Brynjolfsson, Li and Raymond, arXiv 2304.11771 (v2, 2024), Table 3.

In the coding experiment, 95 professional developers recruited online were asked to build a small web server. Among those who finished, the group with an AI pair programmer took 71.17 minutes on average against 160.89 minutes for the control group, a 55.8% reduction. The authors report a wide 95% confidence interval, from 21% to 89%.[4]

In the writing experiment, 453 college-educated professionals were given writing tasks typical of their occupations, and half were randomly given access to ChatGPT. Average time taken fell by 40% and output quality, as rated by evaluators, rose by 18%.[2] These were short, self-contained tasks in an online experiment, not full client projects.

The field study is the most realistic of the three and shows the smallest time effect. Average chat duration fell by 3.7 minutes from a baseline of 43 minutes, or 8.5%.[3] But time per chat is only one part of the story, because agents with AI could also handle more chats at once.

Figure 2. What changed for support agents with an AI assistant

Resolutions per hour
+15.2%
Chats handled per hour
+15.4%
Average chat duration
-8.5%

Change with AI assistant (%), -10 to +20

Baselines: 1.97 resolutions per hour; 2.4 chats handled per hour; 43-minute average chat.

Estimated effect of access to an AI assistant, with agent, location, tenure and month fixed effects. The paper reports chats per hour as an increase of 0.37 on a baseline of 2.4, "roughly 15%." Source: Brynjolfsson, Li and Raymond, "Generative AI at Work," arXiv 2304.11771 v2, Section 4.1.

Output per hour rose about 15%, and the share of chats resolved rose by a small 1.3 percentage points from a high baseline of 82%. In other words, the speed did not come at the cost of quality on average.[3]

Who gains most

Figure 3. The least skilled agents gained the most

36%
Lowest skill quintile
15%
All agents (average)
0%
Highest skill quintile (no increase found)

Change in resolutions per hour (%), 0 to 40

Highest skill quintile: no increase found (0%).

Change in issues resolved per hour after AI access, by agents' skill before the tool arrived. Source: Brynjolfsson, Li and Raymond, arXiv 2304.11771 v2, Section 4.2.1 and abstract.

The gains were uneven. Agents in the lowest skill quintile resolved 36% more issues per hour, while the most skilled saw no increase in productivity, and small declines in some quality measures.[3] The authors also found that agents with two months of tenure and AI access performed as well as agents without it who had more than six months of tenure.[3]

For anyone who sells skilled time, this matters twice. AI shortens tasks, and it narrows the gap between experienced and less experienced staff. Under hourly billing, both effects push the invoice down. Under a fixed price, the first raises margin and the second lets a firm train people faster on a repeatable offer.

Limits of the AI evidence

Chapter 4 · 1 min

The evidence, part 2: what the hour costs, and what happens to it

A worked example shows the arithmetic. This case is made up, with round numbers, but it is typical of a specialist who gets faster over time.

Year oneYear five, billed hourlyYear five, fixed fee
Hours to finish the job1044
Rate or price$100 an hour$150 an hour$1,500 per job
Earned per job$1,000$600$1,500
Effective hourly rate$100$150$375

Even after a 50% rate rise, the hourly version earns less per job than year one did. Now add a tool that cuts the remaining time by 40%, the average saving in the writing experiment.[2] The four-hour job becomes about 2.4 hours. Billed hourly at $150, it earns $360. Sold as a $1,500 package, the effective hourly rate rises to $625. The client pays the same for the result in both cases. The only question is who keeps the time saved.

If you want to know whether your offer is clear enough to sell as a package, and whether it supports the price you want, the Brave New audit looks at exactly that.

Chapter 5 · 2 min

Original research: what clients pay for professional services, 2019 to 2026

We wanted to see whether the market has already started to separate professions that price on time from those that do not, and where marketing work sits. So we pulled two sets of official US data on 3 November 2026: producer price indexes for four professional services from FRED, and May 2025 wage estimates for fourteen marketing, creative and professional occupations from the Bureau of Labor Statistics' public data API. The method is at the end of this essay.

Finding 1: advertising agency prices fell behind inflation

A producer price index tracks the prices firms in an industry receive for their services, holding the service constant. We rebased four of them, and the consumer price index, to September 2019.

Figure 4. What clients paid for professional services, 2019 to 2026

Offices of lawyersOffices of certified public accountantsManagement consulting servicesAdvertising agenciesConsumer prices (CPI-U)
100120140Sep 2019Sep 2020Sep 2021Sep 2022Sep 2023Sep 2024Sep 2025Sep 2026150.4123.8116.2109.4131.1

Index, Sep 2019 = 100

SeriesSep 2019Sep 2020Sep 2021Sep 2022Sep 2023Sep 2024Sep 2025Sep 2026
Offices of lawyers100.0105.2108.5114.3120.7129.7139.9150.4
Consumer prices (CPI-U)100.0101.4106.8115.6119.8122.7126.4131.1
Offices of certified public accountants100.098.3101.3106.8115.0117.2120.6123.8
Management consulting services100.098.1115.3116.5122.6122.7122.9116.2
Advertising agencies100.097.697.9105.6106.3107.4108.2109.4

Index, September 2019 = 100. Producer Price Index by industry: offices of lawyers (PCU541110541110), offices of certified public accountants (PCU541211541211), management consulting services (PCU541610541610), advertising agencies (PCU541810541810); Consumer Price Index for All Urban Consumers (CPIAUCSL). Source: Brave New Media analysis of US Bureau of Labor Statistics data via FRED, downloaded 3 November 2026.

Over seven years, prices received by advertising agencies rose 9.4%. Consumer prices rose 31.1%. In real terms, the price of agency services fell by about 17% (109.4 divided by 131.1). Law office prices rose 50.4%, well ahead of inflation. Accounting rose 23.8% and management consulting 16.2%, both behind inflation. Management consulting is the only series that fell in the most recent year, from 122.9 to 116.2.

We cannot say from this data why agency prices lagged. The index does not record how agencies bill, and many factors move it: competition, in-house teams, media commissions and demand. What it does show is that in the profession most of our readers work in, the going price for a constant service has lost ground to inflation for seven years. A firm that sells time in that market is selling a unit whose real price is falling, while tools make each unit shorter.

Finding 2: the hour has a public price tag

Clients can look up what an hour of skilled work costs an employer. The Bureau of Labor Statistics publishes it for every occupation.

Figure 5. Median hourly wage by occupation, May 2025

Marketing managers
$80.19
Lawyers
$76.76
Public relations managers
$70.63
Advertising and promotions managers
$64.26
Web and digital interface designers
$50.00
Management analysts
$48.97
Technical writers
$43.46
Accountants and auditors
$40.23
Market research analysts and marketing specialists
$37.87
Editors
$37.46
Writers and authors
$36.98
Public relations specialists
$35.94
Advertising sales agents
$31.16
Graphic designers
$30.27

Median hourly wage ($), 0 to 90

National employment and wage estimates for employees, May 2025. Source: US Bureau of Labor Statistics, Occupational Employment and Wage Statistics, retrieved through the BLS Public Data API (series OEUN0000000000000 plus occupation code and data type 01, 03 and 08).

These are employee wages, not billing rates, and firms bill above them to cover overhead and profit. But they set the reference point a client has in mind when a proposal shows hours. A buyer who sees "20 hours of design" can compare it with a designer's wage in seconds. A buyer who sees "brand starter kit, two weeks, $4,800" compares it with the result.

Limits of our analysis

Chapter 6 · 1 min

A framework: the Hours-to-Product Ladder

The evidence points to an order of work. Each rung moves more of the gain from speed to the seller, and asks more of the seller in return: a tighter scope, a clearer result and better proof.

RungWhat the client buysWho keeps the time AI savesEvidenceWhat it asks of you
5. Value-priced outcomeA result priced on its worthYouA 1% price rise lifts operating profit 8% at an average S&P 1500 company (see price versus value)Proof of the result in the buyer's money
4. Productized packageA named offer with fixed scope, price and timelineYouFixed fees reward speed; AI cut writing task time 40%[2]A repeatable process and a written scope
3. Fixed-fee projectA defined piece of work, quoted each timeYou, if the quote holdsSame as above, with more quoting effortGood estimating
2. Capped hoursTime, with a ceilingMostly the clientThe client gains the speed; you carry overrunsTracking and change control
1. HourlyTimeThe client15% more output per hour in the support study shows up as a smaller bill per issue[3]Nothing new, and that is the trap

Most firms we see run all five at once without choosing. The point of the ladder is to choose on purpose: keep the hour for work nobody can define, and move everything you have done twenty times up to rung 4.

Chapter 7 · 1 min

How to package your work into productized services

Step 1: Find the job you repeat

Look at the last year of work. Which request comes up again and again, and which one do you deliver well without thinking hard? Start there, not with your most custom work. If the same kind of job keeps getting shorter, it is a strong candidate.

Step 2: Name the buyer and the result

"Brand refresh" is vague. "A new logo, colour palette and one-page brand guide for a regional food distributor, ready to print" is a product. The narrower the buyer, the easier this gets, which is why market segmentation matters even for a small firm.

Step 3: Fix the scope in writing

List what is included and what is not: deliverables, rounds of revisions, meetings and what the client must provide. Scope creep is the main risk of fixed fees, and this list is the defence.

Tip: Write the "not included" list as carefully as the "included" one. Most disputes start in the gap between them.

Step 4: Fix the timeline

Give the package a set length: five working days, two weeks, one month. A timeline makes it feel like a product, lets the buyer plan, and gives you a rhythm to improve.

Step 5: Price on the result

Start with the floor: the time the job takes now, times the rate you need, plus a buffer. Then estimate the ceiling: what the result is worth to the buyer. Price closer to the ceiling and do not show hours. Buyers weigh losses heavily, which is loss aversion, so frame the package around what the buyer stops losing.

Step 6: Give it a name and a page

A named package is easier to remember, recommend and buy. Put the buyer, result, scope, timeline and price on one page, and make the unique selling proposition obvious. Clear wording matters as much as the number, which is the subject of the essay on how clarity commands price.

Which to do first

ActionLadder rungEffortLikely impactWhen
Log hours on your last ten jobs of the same type1 to 3LowShows which jobs are shrinkingThis week
Write scope, timeline and result for one job4LowHighThis month
Quote that job as a fixed package to new inquiries4LowHighNext five inquiries
Stop showing hours in proposals for that job3 to 4LowMediumNext proposal
Build a short template and checklist using AI tools4MediumHigh margin gainThis quarter
Collect results in the client's money for one package5MediumOpens value pricingOngoing

Chapter 8 · 1 min

What this means for leaders

For service firms that sell into manufacturing and supply chain, the case is strong. Engineering, inspection, logistics and marketing providers still often quote by the day or the hour. A named, fixed package is easier for a procurement team to approve, easier to compare against a rival's open-ended quote, and much easier to take into a second region or country, because the scope travels with it.

The data adds urgency. AI shortens skilled tasks by anywhere from 8.5% to 55.8% in the studies above, and it narrows the gap between new and experienced staff.[2][3][4] In advertising, the real price of a constant service has fallen for seven years.[5][6] A firm that keeps selling hours in that setting hands its productivity gains to clients and competes on a shrinking unit.

A package changes what growth looks like. Once an offer has a fixed shape, a firm can build AI flows around it: intake forms that fill the brief, templates that draft the first version, checks that keep quality steady, and outreach that takes the same offer to new buyer segments and new markets. You can scale a package. You cannot scale an open-ended timesheet. That is the work we do with clients who want to grow their customer base without growing headcount at the same rate.

If you want an AI flow that takes your first package to more buyers and new markets, start with the Brave New audit.

Chapter 9 · 1 min

Common mistakes

Chapter 10 · 1 min

Methodology

Published studies. Figures 1 to 3 report numbers stated in the text of each paper. Figure 1 uses each study's headline time measure: completion time for those who finished (Peng et al.), average time taken (Noy and Zhang, as published in Science), and average chat duration (Brynjolfsson, Li and Raymond, preferred specification). Figure 2's chats-per-hour figure is calculated from the reported 0.37 increase on a 2.4 baseline.

Price indexes. On 3 November 2026 we downloaded monthly series PCU541110541110, PCU541211541211, PCU541610541610, PCU541810541810 and CPIAUCSL as CSV files from FRED, Federal Reserve Bank of St. Louis, using a script that identified itself as "BraveNewResearch/1.0." We took the September value of each year from 2019 to 2026 and divided by the September 2019 value. The real change in advertising agency prices is the agency index divided by the CPI index for September 2026.

Wages. We requested national OEWS series for fourteen occupations from the BLS Public Data API (version 1, no key) on 3 November 2026: employment (data type 01), mean hourly wage (03) and median hourly wage (08). All values are May 2025 estimates, the latest published.

Limits. Industry price indexes and occupation wages are averages across many firms and people. They do not show any one firm's prices, billing model or margins. Our link between AI time savings and hourly invoices is arithmetic, not a measured outcome.

The data behind every figure in this essay is published alongside it as CSV files.

Chapter 11 · 2 min

Frequently asked questions

What are productized services?

Productized services are services packaged like products: a fixed scope, a fixed price, a fixed timeline and a clearly named result, sold in the same form to many buyers. Examples include a website audit, a monthly bookkeeping package, an equipment inspection program or a brand starter kit.

Is hourly or project pricing better?

Project pricing is usually better for work you have done many times, because you keep the gain from getting faster and the client gets a clear price. Hourly pricing still suits open-ended work that neither side can define. Many firms use fixed packages for repeatable jobs and an hourly rate for the rest.

How much faster does AI make professional work?

It depends on the task. In published studies, AI cut task time by 55.8% in a coding experiment, by 40% in professional writing tasks, and by 8.5% per chat for customer support agents at one firm, where output per hour rose 15%. The least experienced workers gained the most.

Will AI make hourly billing obsolete?

Not on its own, but it makes the hourly model harder to defend. When a tool cuts task time, an hourly invoice shrinks while the client gets the same result. That gives firms a strong reason to move repeatable work to fixed packages and keep the hour for work that cannot be defined in advance.

How do you set a fixed fee for a service?

Start with the floor: the hours the job takes, times the rate you need, plus a buffer for surprises. Then estimate what the result is worth to the buyer. Set the price between the two, closer to the value, and write a tight scope with a set number of revisions so the fixed fee stays fixed.

Have advertising agency prices kept up with inflation?

No. According to the US producer price index for advertising agencies, prices rose 9.4% from September 2019 to September 2026, while consumer prices rose 31.1%. That is a fall of about 17% in real terms. Law office prices rose 50.4% over the same period.

Your next step

This week, pick the one job your clients ask for most. Look up how long the last five took. Then write its scope, timeline and result on a single page, put a fixed price on it, and offer it to your next inquiry instead of an hourly quote.

If you want a team to turn that package into an AI flow that reaches more buyers in more markets, start with the Brave New audit.

Brave New Media is a marketing agency that builds AI flows to help companies scale their ideas, grow their customer base and expand into new markets, with experience across supply chain and manufacturing.

See also

References

1. Weiss, D. C. "Accounting 'rogue outliers' seek to change billable hour." ABA Journal, 31 July 2013. abajournal.com

2. Noy, S. and Zhang, W. "Experimental evidence on the productivity effects of generative artificial intelligence." Science, July 2023. scale.stanford.edu

3. Brynjolfsson, E., Li, D. and Raymond, L. "Generative AI at Work." arXiv 2304.11771, version 2, 6 November 2024. arxiv.org

4. Peng, S., Kalliamvakou, E., Cihon, P. and Demirer, M. "The Impact of AI on Developer Productivity: Evidence from GitHub Copilot." arXiv 2302.06590, 2023. arxiv.org

5. US Bureau of Labor Statistics. Producer Price Index by industry: offices of lawyers, offices of certified public accountants, management consulting services, advertising agencies, via FRED. fred.stlouisfed.org

6. US Bureau of Labor Statistics. Consumer Price Index for All Urban Consumers, series CPIAUCSL, via FRED. fred.stlouisfed.org

7. US Bureau of Labor Statistics. Occupational Employment and Wage Statistics, May 2025 national estimates. bls.gov

8. Brave New Media. Analysis of US professional services prices and occupational wages, 3 November 2026. Data published with this essay.