For a creative agency, anything that returns hours is worth more than anything that returns a feature. The work that leaks hours is not the creative work itself. It is the same client question answered three times by three people, the brief that arrives without the information the work needs, and the status update that interrupts the work to report on the work. Most of that is answerable from a website, on the agency's own published terms, without the account manager getting involved.
A creative agency sells hours. That sentence does more work than it looks like. It is not just about how the invoice is built, it is about what the agency has to lose. A feature you do not use is a wasted licence. An hour that goes back to a designer, a strategist or an account manager is a unit of billable work the agency can now sell twice. When you read about custom AI agents for creative industries, the question that matters is which hours come back, and how many of them.
The hours leak in three specific places
It is easy to talk about agency life in generalities. The interesting work is to name the conversations that cost time, and to ask whether each one needs a human.
One: the same client question, answered three times by three people. Every agency knows this one. The client asks the project manager how billing works. The project manager asks the account manager. The account manager asks the finance lead. Three people, one question, and the answer is the same paragraph that already sits on the agency website under "How we work" or "Engagement". This is the single biggest leak, because it is a published answer to a recurring question.
Two: the brief that arrives without the information the work needs. A creative brief lands. It says "we need a campaign". It does not say who the audience is, what the budget is, what the deadline is, what success looks like, or what the previous campaign did. The account manager writes a polite email asking for those things. The client writes back asking what the agency needs. Another email. Another day. Meanwhile the work is not being scheduled, because nobody knows what the work is. A published intake form on the website, with the right questions in the right order, removes most of that ping-pong.
Three: the status update that interrupts the work to report on the work. The client wants to know how the shoot went. The strategist wants to know when the concepts are ready. The producer wants to know whether the client has signed off the storyboard. Each of those is a small, legitimate question, and each one pulls a person out of the work for ten minutes. Across an agency with fifteen projects running, the cumulative interruption is several hours a week for each person whose job it is to answer.
These are the three leaks that drain an agency. None of them is the creative work. Naming the creative work as the thing being automated is the mistake most pieces on this topic make, and it is the mistake that makes agency owners stop reading.
What a published answer on a website can absorb
The first leak is almost entirely answerable from a website, on the agency's own terms. The questions are recurring, the answers are stable, and the client is already on the website when they ask. They want to know how engagement works, how billing works, how revisions are handled, what the timeline looks like, what they need to provide, and who their day-to-day contact is. None of that needs the account manager, and almost all of it is already written down somewhere in the agency.
Pulling those answers onto a single page, or into a chat that points to those pages, returns the hours the agency has been giving away. If a question takes an account manager five minutes to answer by email, and it lands twenty times a month, that is just over an hour and a half returned every month, from one recurring question. Multiply by the ten or fifteen questions that recur, and the recovered time becomes meaningful.
The same is true of the second leak, but in a different way. The brief that arrives without the information the work needs is not a question with a single right answer. It is a process question: what does the agency need to know before it can start? The intake form on the website is the published answer. A prospective client opens it, fills in what they have, sees the fields they have left blank, and either comes back with the rest or accepts that the agency will need to ask. Either way, the agency does not spend its afternoon chasing the missing fields by email.
This is also where a professional AI design agent for creative agencies earns its keep. Not by designing anything, but by sitting at the top of the intake form, answering the questions a prospective client has about the form, and routing the submission to the right person once it is complete. The work the agent does is administrative. The work it returns is hours.
What needs the account manager
The third leak is different. The status update that interrupts the work to report on the work is not a question with a published answer. It depends on what happened today, on this project, for this client. A website cannot answer it. The account manager can, or the project manager can, or the producer can. The way to recover those hours is not to answer the question faster. It is to batch the questions, so the work is interrupted once an hour rather than ten times an hour.
Two changes help. First, publish the cadence: clients get a written update on Tuesday and a call on Thursday. That alone reduces the ad-hoc interruptions, because clients stop asking when they know when the next update is coming. Second, give the client a way to see the status without asking. A shared project board, updated by the producer, removes the need for most of the small questions. The board is the published answer. The chat that points to the board is the polite layer on top.
This is the line that is worth drawing clearly. A published answer on a website can absorb the first leak and most of the second. The third leak is recovered by a process change, not by a chat agent. The mistake is to think that adding an AI to the third leak makes it better. It does not. It adds another interruption.
What changes the answer for a given agency
Three things change how much an agency recovers, and they are not the size of the agency or the sector it serves. They are volume, project shape, and opening hours.
Volume is the easy one. An agency that takes fifty enquiries a month from its website recovers more hours from a published answer than an agency that takes five. The maths is linear in the number of enquiries. If a question takes five minutes to answer by email, and it lands twenty times a month, that is an hour and forty minutes. If it lands fifty times a month, that is over four hours. The published answer on the website returns the same hours in both cases.
Project shape matters because of the brief. An agency that runs many small projects, with many different clients, lives inside the second leak. The intake form matters more than the engagement FAQ. An agency that runs a few large projects, with the same clients year after year, lives more inside the third leak. The shared project board matters more than the intake form. Most agencies are somewhere in between, and the work is to look at the past quarter and ask which leak cost more hours.
Opening hours is the one that most pieces on this topic miss. Many of the questions an agency gets arrive outside office hours. The prospective client is researching at nine in the evening. The existing client is checking on the campaign on a Sunday morning. The intake form and the engagement FAQ do not care what time it is. The account manager does. A website that answers at any hour returns hours the agency could not have returned by hiring an out-of-hours receptionist.
The mistake most agencies make, and what it costs them
The mistake is to treat the website as a brochure rather than a working member of the team. The website sits there, it has the engagement page, the work page, the contact page, and not much else. The account manager answers the same questions by email. The producer writes the same status update by hand. The hours leak out, and the agency hires another junior to handle them, and then another.
The cost is not the salary of the junior. The cost is the hours the senior people spend training the junior, answering the junior's questions, and reviewing the junior's work. Those are hours the agency cannot bill for, because the client did not ask for training, supervision or review. They are real hours, and they are the hours a working website returns.
A second mistake is to confuse the creative work with the administrative work. The creative work is what the agency sells. The administrative work is what the agency does in order to sell it. A chat agent on the website is not going to design the campaign, write the copy, or shoot the film. It is going to answer the questions about the campaign, the copy and the film that are already answered on the website. The agency that reads this and thinks "this will replace my designer" has read the wrong article. The agency that reads this and thinks "this will let my designer design" has read it correctly.
What the hours are worth, on Asyntai's published numbers
The Asyntai pricing page, checked September 2026, sets out four self-serve plans and one enterprise contract. The plans are priced by message, not by seat, which matters for an agency because the agency's volume is uneven. A campaign launch brings a spike. The week before Christmas is quiet. Per-seat pricing charges the same in both weeks. Per-message pricing follows the work.
| Plan | Price | Messages included | Approx. enquiries per month | Cost per enquiry |
|---|---|---|---|---|
| Free | $0 | 50 | 6 | $0.00 |
| Starter | $39 | 800 | 100 | $0.39 |
| Standard | $139 | 15,000 | 1,875 | $0.07 |
| Pro | $449 | 50,000 | 6,250 | $0.07 |
A "message" in Asyntai's pricing is one exchange between the visitor and the agent. A short enquiry that takes three messages to resolve counts as three messages. The "Approx. enquiries per month" column above assumes an average of eight messages per enquiry, which is this article's working assumption and not a published figure. Measure your own average before you use it. The "Cost per enquiry" column is the plan price divided by the enquiries that fit inside it.
The conversion from messages to a plan matters, because the agency owner reading this does not think in messages. The owner thinks in enquiries: how many prospective clients will land on the website this month, and how many of them will ask something the website can answer.
Two worked examples, using only the figures from the table above.
Whether that is worth it depends on what an enquiry costs the agency today. If an account manager spends five minutes answering an enquiry by email, plug the agency's own loaded cost per hour for that person into the five-minute figure to get the real cost of the email reply. Even on the cheapest Asyntai plan, the agency is paying a small fraction of that to have the same question answered from the website. The agency owner should use their own loaded cost and their own average handle time when they run this maths for their own agency.
Where Asyntai fits, and where it does not
Asyntai is a chat agent that sits on the agency's website and answers visitor questions from the agency's own published material. It is installed with a script tag, or through Code Injection on Squarespace, which matters for the agencies that have not touched their CMS in three years. Per-tab permissions and per-website scoping matter for the agency group that runs several brands under one roof. The append-only audit log matters for the agency that has to show a client who said what, and when. Configurable chat retention with auto-delete matters for the agency that handles brief contents it cannot keep.
Asyntai answers the recurring questions from the agency's published material, and routes everything else to the person who owns it. The campaign, the storyboard and the colour palette stay with the people the client is paying for. The agency that buys it so its design team stops answering the same five questions by email every Monday morning has bought it for the right reason.
For agencies whose clients include regulated industries, the PII redaction and the separate output classifier (not a prompt rule) are worth naming. A client in financial services will not send its customers to a chat that might surface someone else's account number. A client in healthcare will not use a chat that stores the conversation forever. These are the configuration choices the agency makes once, on behalf of all its clients, and then forgets about.
A short list of questions to ask before you buy anything
- Which of the three leaks is biggest for us? Look at last month's emails and count. The leak that costs the most hours is the leak the website should be doing the most work on.
- How many website enquiries do we get in a typical month? Multiply by the average handle time. That number, in minutes, is what the agency is currently spending on questions a website could answer.
- What is the agency's loaded cost per hour for the person who answers? The maths in the section above only works if the agency's own numbers go into it. Use them.
- What does the intake form ask for, in the order it asks for it? If the order does not match how a prospective client thinks, the form is part of the second leak, and the website is doing less work than it could.
- What is the published status cadence, and where does the client see it? If the answer to either is "we just email them", the third leak is wide open, and no chat agent on the website is going to close it.
- What are the agency's opening hours, and what proportion of enquiries arrive outside them? That proportion is the hours the agency is leaving on the table, every week, for a website that does not sleep.
What to do this week
Three things, in order, none of which require a purchase decision.
First, pull the last month's emails and tag each one as leak one, leak two or leak three. Count them. The pattern is usually obvious by the end of the second day, and it tells the agency which leak to attack first.
Second, write down the answers to the five questions that arrive most often. Put them on a single page on the website, under a heading a client would actually search for. Do not write them for other designers. Write them for the client's procurement lead, who has never met the agency and is reading at ten in the evening.
Third, write down the fields the agency needs on every brief, in the order they are needed. Put them on the intake page. Make the page send the submission to the account manager's inbox, with the fields labelled, so the account manager does not have to read the email to find them.
Once those three are in place, look at what is left. The leftover is the work that genuinely needs the account manager, the producer and the designer. That work is what the agency is for. The work that has just been moved to the website is what the website is for.
The point, in one sentence
A creative agency earns its money on the hours it bills for. The hours it does not bill for are the ones that decide whether the year was a good one. The work that returns those hours is not the creative work. It is the small, recurring, published-answer work that the website can do at any hour, in any volume, for a price that does not depend on how many people the agency has hired this quarter. That is the work a chat agent is for, and it is the only work a chat agent should be bought to do.
Asyntai installs on a creative agency website with a single script tag and answers from the agency's own published material. The Free plan covers 50 messages a month, the Starter plan covers 800 messages for $39, and the Standard plan covers 15,000 messages for $139. Agencies that need more than 50,000 messages a month go on a signed contract. See the full plan details on the Asyntai pricing page.
Sources. Asyntai plan prices, message allowances and tier labels: Asyntai's own pricing page, checked September 2026. The conversion of messages to enquiries (eight messages per enquiry) is this article's working assumption, not a published figure; measure your own. Loaded cost per hour for an account manager, average handle time per enquiry, and current email volume: figures the agency should supply for itself; no industry figure has been quoted in this article.