Why we publish prices, and what 20 agency price lists show

Our audit is $2,500. Builds are $3,000 to $15,000, fixed price, paid half up front and half at handover. Care retainers are $1,500, $2,500 or $4,000 a month. The $500 written diagnostic is credited in full against the next step, and the audit comes off any build started within 90 days. All of that sits on our pricing page where competitors can read it, and people ask why we publish prices when so many agencies make you book a call to hear one. This is the argument, the evidence we collected in August 2026, and the best version of the counterargument.

Hidden pricing moves the cost to you

When an agency hides its prices, you pay the cost in time. To learn one number you fill in a form, wait for an email, sit through a 45-minute discovery call that is mostly questions about your budget, and then wait for a proposal. Compare four or five agencies, as you should, and shopping carefully has cost you a working day.

The friction is deliberate. It exists so a trained person can hear your situation before naming a figure, judge what you might tolerate, and anchor there. "Book a call to discuss pricing" means the price depends on who is asking. If the number were fixed it would not need a conversation.

There is a second cost, quieter than the first. A single quote arrives without context, so the first proposal you receive becomes the anchor for every proposal after it, and you spend the rest of your shortlist judging agencies against a number you had no way to evaluate when you got it.

What 20 price lists showed

In August 2026 we opened 20 AI automation agency websites in the US, the UK and a few elsewhere, and recorded every price each one published: diagnostics, builds, retainers, day rates, hourly rates, and what each fee covered. The line by line version lives in our comparison of AI automation agency pricing for 2026. What follows are the four patterns that changed how we think about our own page.

The first correction is aimed at us. We have described published pricing as rare, and the sweep says otherwise. Twenty sites took an afternoon to find. Among the agencies chasing the same buyer we chase, a price list is now ordinary. What stayed rare was an agency willing to explain its numbers: which band a project falls into and why, what pushes a build from $6,000 to $14,000, what the retainer stops covering. A figure parked on a page above a contact form gives you a number and keeps the reasoning behind it private, which leaves the buyer holding the same question they typed into a search box in the first place.

Transparency is contracting at the top of the market

The result that stayed with us came from the pages that had emptied out since we last looked.

2V Automation used to publish $5,000 to $20,000 per project, with retainers starting at $1,000 a month. Those numbers were readable on the site through 2024 and 2025. In 2026 they are gone.

AY Automate published more than almost anyone and then published nothing. Its sheet ran discovery from $4,500, builds from $8,000 to $35,000, retainers between $500 and $3,000 a month, and fractional engagements at $6,500, $9,500 and $15,000 a month depending on days per week. That is about as complete as agency pricing gets in public. It came down during 2026.

Both agencies moved upmarket, and the economics of that move are the strongest argument against everything else in this post. Enterprise buyers negotiate, so a published floor becomes a ceiling the moment procurement quotes your own page back at you. A single $200,000 programme can outweigh a year of small fixed-price work, and it is exactly the deal a public range talks you out of before anyone picks up the phone.

The honest reading is that we publish partly because we are small. Our sales cost is the expense that public pricing cuts hardest, and at our size that saving is worth more than the negotiating room we give away. The agencies that removed their prices were doing the arithmetic their new position rewards, and we would probably reach the same answer in their seat.

If we are still publishing this page at thirty people, ask us again.

Where paid diagnostics cluster

Diagnostics were the most consistent part of the sweep. The prices land on a short list of anchors, and the same anchors repeat across currencies and countries.

Diagnostic	Published price
CloudNSite Current State Assessment	$999
The AI Consultancy audit	£495
Axivon Revenue Leakage Review	£495
The Automation Agency Process Audit	£1,500
Palavir audit	$2,500
Justin McKelvey AI Readiness Assessment	$2,500
The AI Consultancy Readiness Sprint	£3,500
Our written diagnostic	$500
Our AI Opportunity Audit	$2,500

The clustering tells you something the individual numbers do not. Under about $1,000, a diagnostic is a structured conversation plus a document, priced to be an easy yes. From $2,500 upward the agency has committed to days inside your systems, and you should hold it to that: access to the tools, a count of the volumes, a workflow map a new engineer could read. That gap is the subject of what a paid AI opportunity audit contains, and it is why we sell the $500 diagnostic and the $2,500 audit as two products instead of blurring them into one.

Four grades of credit-back promise

Nearly every paid diagnostic in the sweep says the fee comes off the build. The phrase is doing different work in different places, and the differences are worth the two minutes it takes to check before you pay.

• Credited in full, with a deadline. Justin McKelvey credits the full $2,500 against work started within 90 days, and Axivon credits its £495 Revenue Leakage Review in full. Our $500 diagnostic works this way, and our $2,500 audit comes off any build started within the same window.
• Half credited. The AI Consultancy returns 50% of its £3,500 Readiness Sprint against work inside 90 days, which is a genuine discount and a smaller one than the words "credited against your build" suggest at a glance.
• Credited only if a build follows. The Automation Agency's £1,500 Process Audit comes off the build when there is a build. If the finding is that you should not build anything, the credit disappears along with the project, which puts a quiet thumb on the scale of the recommendation you paid for.
• Not credited at all. The fee is the fee, which is a defensible way to sell advice as long as the page says so plainly.

Four materially different offers, one shared sentence. Read the deadline, and check whether the credit survives a recommendation to stop.

The spread is narrower than the internet claims

The line that gets repeated in every thread about agency pricing is that quotes for the same AI work vary by a hundred times. Across 20 published price lists that did not survive contact with the data. Between the cheapest and the most expensive published price for a comparable production build, the ratio came out somewhere between seven and seventeen, which is still a wide enough spread to make shopping difficult and nothing like the hundredfold gap the folklore promises.

Band	Published range
Entry, one automation	£350, or from $3,000 to start
Mid-market production build	$3,000 to $15,000
Assessment only, Blue Orange Digital	$25,000 to $50,000
Enterprise programme, Aloa	$50,000 to $300,000 and above
Production build, Winder.AI	£120,000 to £400,000

The hundredfold illusion comes from comparing across bands. A £350 single automation and a £400,000 production programme never compete for the same purchase order, and quoting them in one sentence makes the market look lawless when it is merely segmented. Blue Orange Digital charging $25,000 to $50,000 for an assessment alone shows where a band boundary sits: that is more than most mid-market agencies charge for the finished system.

Inside a band the spread is real and it survives for a specific reason. Scope is vague and quotes float free of each other. The freelancer at $900 and the consultancy at $15,000 may well be describing different things, and you cannot tell, because neither has written down what is included.

Day rates, retainers and the salary they compete with

Day rates were the tightest cluster in the sweep: Axivon at £550, MQLFlow at £800, The AI Consultancy between £950 and £1,500, and Winder.AI publishing hourly instead at £150 to £400. A day rate is a fair way to sell advisory time and a poor way to buy a build, for the reasons in fixed price against hourly for AI work.

Retainers spread wider than any other category, from $200 a month to $8,500. MQLFlow starts at £200, Podlevskikh at $300, Zaps at $500, ours at $1,500 with tiers at $2,500 and $4,000, and The AI Consultancy's fractional Chief AI Officer runs £3,000, £5,000 and £8,500 a month by days committed. A $200 retainer and an $8,500 retainer are different products wearing the same word, so ask about coverage before price: what response time you get at 9am on a Monday, and whether anyone is watching the system when nothing is broken. That list is what a care retainer should include.

Every one of those numbers is competing with a salary, and the salary is the anchor most buyers forget to draw. A US automation engineer on a $150,000 to $200,000 base costs roughly $13,500 to $15,000 a month once payroll taxes, benefits, equipment and tooling are counted, and that is before the two or three months of recruiting. Our top care tier is $4,000 a month and our fractional engagements run $6,000 to $12,000, both under that line by design. The comparison, including where hiring wins outright, is in agency against freelancer against in-house and in what a fractional AI engineer costs.

A price is a scope

This is the part that gets missed. The hard thing about publishing a price has little to do with courage. You cannot publish a price without first defining what the price buys. Our $2,500 audit had to be pinned to deliverables: workflow map, ranked opportunities with the math, a 90-day sequence, fixed quotes. The $3,000 to $15,000 build range had to be pinned to what sits at each end and what counts as a change rather than an inclusion, which is the same discipline behind quoting fixed instead of hourly.

Publishing settles questions that a hidden price leaves open until they turn into an argument: what happens when the data is worse than described, and how many rounds of revision a fixed price survives.

That definition work is the same discipline that decides whether a project succeeds. In the rescues we have scoped, vague scope killed more projects than bad technology did, and it sits behind most of the reasons AI projects get abandoned. An agency that will not commit to public numbers has unfinished scoping homework, and it will finish that homework on your project, at your expense.

The case for hiding prices

Agencies that hide pricing give one reason: every project is different. This is half true. Businesses differ, integrations differ, and a quote for complex custom work does require the specifics, which is why our exact quotes come after the audit. There, the case for a call first is fair, and the 2V Automation and AY Automate story is the stronger version of that argument because it comes with evidence.

What the argument skips is that ranges are always possible. "Builds like this run $3,000 to $15,000" is a sentence any agency could publish. Every project being different explains the absence of your exact price. It leaves the absence of any price unexplained. Restaurants do not know what you will order either, and they still print a menu.

What a published price does not promise

A published price is a starting position, and we will scope up when the work turns out bigger than the page assumed. If your CRM is three CRMs held together by a shared spreadsheet, or a compliance review adds six weeks and two more stakeholders, the number moves. That is true of us and of all 20 agencies in the sweep.

Rescue work sits furthest from any published figure. When someone arrives because an automation stopped working and the person who built it is unreachable, the first job is finding out what exists at all, which is why the repair or rebuild decision gets priced after a look rather than before one.

What publishing buys you is the ability to leave before the first call.

Ask for a number before the call

Our position is plain: you should know roughly what something costs before spending an hour of your life talking about it. So ask every agency for a number, or a range, before agreeing to a meeting, alongside the other questions worth asking before you hire anyone. An agency that gives you one has done its scoping work and respects your time. An agency that insists a call must come first is telling you the price is negotiable, and it will be negotiated against you.

Published prices do not make an agency good. We could publish accurate numbers and still do mediocre work, and you should judge the work separately. Hidden prices guarantee one thing only: the shopping cost lands on you and the anchor gets set by them.

Everything we charge, including what each retainer tier covers and the 30-day bug warranty on builds, is on the pricing page, and the AI Opportunity Audit is the $2,500 line item on it. Read both before you talk to us. That was the point of writing them down.