Meeting held at boardroom conference room

‘Enterprise-grade’, ‘bespoke’, ‘scalable’: the agency words that don’t mean anything

Commissioning Desk favicon
9 Min Read

Agency homepages have a habit of reaching for the same dozen words. Bespoke. Scalable. Enterprise-grade. Data-driven. Full-service. They are built to sound like promises while committing the agency to nothing you could later hold against them. The way to read a pitch is to translate each adjective back into the plain thing it stands in for, then ask the one question that forces a real answer.

This isn’t about catching anyone lying. Most of these words are technically true, which is the problem. ‘Bespoke’ is true if a developer wrote a single line of custom CSS. ‘Enterprise-grade’ is true if one of your clients is large. The words survive on a homepage precisely because nobody can ever prove them wrong. That’s the test we’ll apply to all eight: not ‘is it true?’ but ‘what would have to be the case for it to be false?’ If the answer is ‘nothing’, the word is decoration.

Why agencies reach for words that can’t be wrong

An adjective shifts risk onto you. A number shifts it back onto them. ‘We build scalable sites’ is a claim no one can breach. ‘This architecture holds up to a stated concurrent-user ceiling, and re-engineering above that would be a separate, costed phase’ is a claim you can hold them to in eighteen months. Agencies reach for the first kind because it reassures you at the point of signing and costs them nothing afterwards.

Good agencies don’t mind being pulled off the adjectives. They tend to have the numbers ready, because they’ve had to answer for them before. The ones who get tetchy when you ask ‘scalable to what, exactly?’ are telling you something their homepage was written to hide. None of what follows is a reason to walk away from a pitch. It’s a reason to ask better questions in the room, the same way you would when you’re working out why two agencies quoted wildly different numbers for the same brief.

‘Bespoke’: custom-built, or a theme with the colours changed?

‘Bespoke’ is meant to signal that the work was built from scratch, for you, unlike anything else. In practice it covers everything from a genuinely custom application to a cheap off-the-shelf template with your logo dropped in. The word spans that entire range, which is exactly why it tells you nothing on its own.

The useful question is: bespoke at what level? The visual design, the templates, the underlying data model, or just the colour palette? Plenty of excellent work starts from a framework or an existing theme, and there’s nothing wrong with that. The issue is paying a bespoke price for a configuration job. Ask them to show you what the non-bespoke version of your project would look like and what it would cost. If they can’t draw that line, the word was doing marketing, not describing the work. (If you want the honest version of where custom genuinely earns its fee, discovery is usually where it shows up.)

‘Enterprise-grade’: graded by whom, against what?

There is no certification body for ‘enterprise-grade’. No standard, no audit, no badge. It’s self-awarded. On most homepages it means one of two things: they’ve worked with at least one large client, or the hosting comes with an uptime figure they read off a dashboard.

Every component of ‘enterprise-grade’ that’s real is a separate, checkable thing. An uptime SLA in writing, with a number and a penalty. Penetration testing, with a date and a report. Role-based access control. A disaster-recovery plan with a stated recovery time. Each of those you can verify before you sign. ‘Enterprise-grade’ you cannot. So don’t ask whether it’s enterprise-grade. Ask which of those specific things are included, and get the answers in the contract rather than the proposal.

‘Scalable’: scalable to what number?

‘Scalable’ almost always means the same thing: it’s on cloud hosting. Usually AWS , sometimes another big provider. Being on AWS is not the same as being built to scale. A badly architected site on AWS falls over under load exactly like a badly architected site anywhere. You just get a more impressive bill on the way down.

Scalable to what? That’s the whole question, and a real answer comes with a ceiling. ‘At what concurrent-user count does this need re-architecting, and what does that work cost?’ An honest agency will give you a range and an assumption set. ‘Infinitely scalable’ is a reliable tell that nobody has load-tested anything. Scale is a number with a price attached, not an adjective.

‘Data-driven’: which decision did the data actually change?

A favourite of marketing and creative agencies, ‘data-driven’ is meant to signal that decisions come from evidence rather than the creative director’s mood. Frequently it means Google Analytics is installed. Data existing and data driving anything are different states, and the homepage word quietly assumes the second when it can only prove the first.

So ask for one example. Which decision on a recent account got reversed by data, and what was the data? ‘Driven’ implies the numbers occasionally win an argument the agency would otherwise have lost. If the analytics have never once changed the agency’s first instinct, they’re a dashboard, not a method. The platforms matter less than the habit; you can run a genuinely data-driven setup on HubSpot or Marketing Cloud and still never look at it.

‘Full-service’: in-house, or subcontracted and marked up?

‘Full-service’ and its cousin ‘end-to-end’ promise one team handling everything, no gaps, no finger-pointing. Sometimes that’s real. Often it means they do the parts they’re good at in-house and quietly subcontract the rest, with a margin on top.

There’s nothing wrong with subcontracting. There’s a lot wrong with paying full-service rates for it without knowing. Ask which disciplines are done in-house and which are farmed out, and ask who you’ll actually be talking to for each. ‘Full-service’ where the SEO is a freelancer the account manager has never met is a perfectly fine arrangement, as long as you know that’s what you’re buying and have priced it accordingly.

‘Future-proof’: proof against what, exactly?

Nobody can future-proof anything. The phrase promises a thing no agency can deliver, because the future’s defining feature is that you don’t know what it holds. What good teams actually do is build so that change is cheap, not impossible. That’s a real and valuable thing. ‘Future-proof’ is not the words for it.

Drop the word and ask the questions underneath it. What happens when you want something they didn’t anticipate? What does a change cost, six months in? Who owns the code, and could a different agency pick it up without a rebuild? Those answers tell you about your actual future. ‘Future-proof’ only tells you about their copywriter. Code ownership in particular is worth pinning down early, and it sits next to a handful of other clauses worth reading before you sign.

‘AI-powered’: powered by what, and what happens when it breaks?

The current favourite. On a growing share of homepages, ‘AI-powered’ means the product makes an API call to somebody else’s model, OpenAI’s or Anthropic’s or Google’s. That can be genuinely useful. It can also be a thin wrapper around a feature that didn’t need a model at all.

Three things are worth asking, and the answers separate the real builds from the stickers. What does the AI do that a conventional feature couldn’t? Whose model is it, and what’s your exposure when that provider changes its pricing or deprecates the version you’re built on? And what’s the fallback when the API is down? ‘AI-powered’ with no answer to ‘what model, and what happens when it breaks’ is a label stuck on the outside, not an engine under the bonnet.

‘Award-winning’: which award, what year, how many entrants?

‘Award-winning’ wants to signal peer-recognised excellence. The catch is that some design and digital awards charge a fee to enter, and not every badge tells you much about how competitive the process was. An award can mean real recognition from people who know the work, or it can mean an invoice was paid.

The question is small and it works: which award, what year, and how many entrants? A named, competitive, free-to-enter award from a credible body means something, and an agency that has one will name it before you finish asking. ‘Award-winning’ with nothing specified is the agency betting you won’t follow up.

The same trick runs through the smaller stuff once you start noticing it. ‘We’re a partner, not a vendor’ often translates to ‘we’d like a retainer’. ‘Results-driven’ tends to mean ‘we would like to be paid’. ‘Holistic’ frequently means the scope hasn’t been decided yet. ‘Industry-leading’ is self-declared, every time, by definition. And ‘seamless’, which is doing a lot of work on a lot of homepages, usually means the integration mostly holds together if nobody pushes it too hard.

The pattern is always the same. The word makes a claim that sounds like a commitment and turns out to be unfalsifiable the moment you press on it. The fix is never to demand the agency stop using the words. It’s to treat every adjective as a prompt for a question that has a number, a name, or a date in the answer. Will they put a figure next to ‘scalable’? A named award next to ‘award-winning’? A model and a fallback next to ‘AI-powered’? If they will, the claim was real and you’ve learned something. If they reach for another adjective instead, you’ve also learned something, and it’s the more useful of the two.

Share This Article
Published by the editorial team at Commissioning Desk, an independent publication covering digital project commissioning, agency selection, and technology decisions for non-technical buyers. Commissioning Desk is founded by Kasper Polanski and draws on input from agency practitioners, in-house digital leads, and the buyers who've sat on both sides of the table. Every article published under this byline is written and reviewed by practitioners with direct experience of the subject matter.