Let’s All Stop Pretending We’ve Never Done This Before

I’ve worked for digital design agencies for most of my career, and all of our projects generally have the same shape. We start out learning about our clients’ business and their needs, form hypotheses, make proposals, eventually align on a solution, and then we go build it. This is so standard, it has a (semi) famous diagram about it:

The (semi) famous double diamond

If you’re at all involved in designing digital tools, especially if you work at an agency, doing work for clients, you’ve seen this. It’s ubiquitous enough to get spoofed on the TV show Silicon Valley with the “Conjoined Triangles of Success.”

What I am going to tell you right now, is that you probably don’t need to keep doing projects like this. This is because this project shape assumes that we do not know what we’re doing, which we very much do.

Ok, with that thesis out of the way, why do projects look like this? There are reasons! The most obvious is that this model assumes that the people doing the work are coming into a new project for a business they don’t understand. This used to be the case. Back when I was starting out designing websites (we didn’t have “apps” yet), we genuinely didn’t know what we were doing. It didn’t, so much, matter, because NO ONE knew what they were doing, even the people who hired us. “One website, please!” with not-a-lot of sense why, aside from you needed to have one. (This is not dissimilar to the situation with AI right now.) We were making it up as we went along, and the projects were pretty simple, with simple goals, so everyone was fairly happy to have a website that had some brochure content on it and looked pretty nice.

As time went on, our work moved up the value chain, from awareness to more core business stuff. We could make sales, or offer a service online. Eventually, there were businesses that were online only, and getting it right, objectively doing something measurably useful, became really important. Along with this budgets, timelines, and numbers of stakeholders increased. Now, we needed to do two things: 1) since we were designers parachuting into a new business, we needed to learn all we could about their assets, problems, and goals, very quickly, and 2) we needed to convince our clients that we understand all that stuff, and now we’ve come up with a good solution to their needs.

This is the point where the double diamond shape really developed. Both of these goals were fulfilled by that left-side diamond. Over time, both contracts and agency teams took on a shape to match the diagram. The work had two phases. Design split into more specific disciplines and some designers became UX designers, became Researchers, became Strategists. Other disciplines got more specific too, and sometimes you’d have whole separate teams dedicated to the discover and define steps. They’d hand things off to an executional team to do the right side of the diagram to build things, while the strategic team moved on to the next new project.

It made sense, because we really were learning new businesses, new audiences, and new ways of behaving online all the time. One month I would be selling sunglasses. Another, I’d be working on an interactive digital pet cat. Or designing software for a digital sign in Times Square. This two phase approach of making sure you knew what you’re doing (“Build the right thing” in our lingo) and then building it (“Build the thing right”) worked pretty well for us. 

Note that the diamond part of the diagram has its own significance that I’m not going to get into here. For right now, we just need to understand that there are two parts:

Diamonds?! In this economy?!

Businesses liked this model. There was plenty of time for the agency team and the client team to get to know each other, and to make sure everyone was on the same page. Another lingo thing: we call this “alignment.” Plenty of time to get aligned. And you could prove you were getting aligned, and making progress towards getting to “Build the thing” with lots of presentations and documents and spreadsheets (“deliverables”). Clients like deliverables because they’re proof of work, and easy to plan around. We can say, “Let’s spend 10 hours planning stakeholder interviews, 10 hours doing them, 10 hours preparing a presentation, 1 hour presenting, and 4 hours on revisions. This will take two calendar weeks.” That’s a list of items that fits really well into a contract and a checklist. You can plan this out a year in advance, no worries. Everyone can track progress and agree when it’s all done. Little risk, and little possibility anyone gets yelled at. Agencies like this because creating all of this proof of work takes a lot of billable hours.

EXCEPT! This is a schedule, list, and definition-of-done anchored to a bunch of things that are not actually directly related to the success of the final outcome of the project. We assume that if we do these tasks, in this order, over this time frame, we will get something valuable out of it — important knowledge alongside alignment about its accuracy and relevance to our work — but we don’t actually KNOW this. The work exists at a level of abstraction from its actual usefulness. As the saying goes “what gets measured gets managed,” and over time, this process and these deliverables became the primary point of the activity, regardless of where they were actually helping with the quality of the end result. Sometimes, a lot of the time if you’re not careful, all of that work gets done and then goes into a file folder, never to be referenced or seen again.

This two phase approach can also create disjunctions. Even if the deliverables are useful, when team Right Thing rolls off, and team Thing Right rolls in, there’s an opportunity to lose things. And shifting people and modes takes time. And now that you’ve got team Thing Right in the room, and team Thing Right is off doing something else, if you have a question, to ask about behavior or strategy or competitors or whatever, you don’t have the right people there to answer it. There’s inefficiency baked into this process. If you’ve got a lot of resources to throw at the problem, that’s not that big a deal, but what if you don’t? 

Don’t drop the ball!

This approach solidified during a period when there was, just, honestly, a lot of money to throw at things, and time to let it all play out. There’s a documentary about the band Steely Dan where one of the guitar players is talking about the luxury of making a rock album in the 70s, and he says something like, “It used to take me six weeks to find a comfortable chair at $900 per day.” I think about this a lot, because, during the early 2000s, when all of this was evolving, that was very much the vibe. Around 2005 or thereabouts, I was flown 500 miles, and put up in a pretty nice hotel, to spend four days making one powerpoint slide. I don’t recall how comfortable the chair was. Although I do still use a version of that slide. Hopefully whoever was paying for it knows how much I appreciate their contribution to my future success!

Tempus fugit, and today everyone is paying much closer attention to time and budget. When people are looking for efficiency, the big, mutli-month project where you’re not actually building anything yet is a prime candidate for the chopping block. This is OK! Really. This is fine, and not in a dog-drinking-coffee-with-the-room-on-fire kind of way. 

The reason it’s fine is that we do, in the year 2026, know what we’re doing, and it’s not necessary to pretend otherwise. It’s actually a detriment to our work to pretend otherwise.

What do I mean when I say that we know what we’re doing? I mean that after many years of talking to people, solving problems, testing, and looking at the performance of our solutions in the real world, we have a very detailed picture of how people go about solving problems and getting their needs met using digital tools. Many, MANY, of the basic tasks people perform online have relevance across a whole range of online activities regardless of industry or user type. Demographics don’t matter for usability nearly as much as we’d assumed back in the day, and we have a good grasp of where the real differences are and how they affect things. We have basic interactive patterns for many common things that users do that are so familiar that it would be ridiculous to tinker with them too much.

Even very specific use cases are well researched and documented. We know a lot about how people use ratings and reviews online to find goods and services. We also know a lot, specifically, about how people use ratings and reviews online to buy books, buy cars, hire a plumber, a financial advisor, or find a primary care physician. I can tell you specific differences in how a person on Medicaid who’s making a doctor’s appointment looks for care depending on whether they’re in the US southeast or northeast. We’ve… I’VE talked to, observed, and surveyed thousands of people. And we’ve got experience with specific approaches to designing tools for all of those situations. We’ve done it, tested it, observed the results, rinse, repeat, over and over again. These are, at a high level, at least, solved problems.

But the standard project model, the one that everyone sees in their heads, still, when they think about this stuff, still assumes we’re looking for basic answers, and that every new client business is sui generis, and agency teams are starting from scratch with each engagement. Agencies are happy to pretend we need to do the whole song and dance because there are a bunch of billable hours in there. But it really undermines our authority as experts to do this. And clients can feel how unnecessary it all is, which undermines us further. As an old job, we had some principles, and two of them were “Be Consultative” and “Build Strong Relationships.” It is hard to do those things when you’re pretending to learn things you already know, reading them bck to clients who definitely already know them, and billing 400 hours to do it.

So! Let’s just cut it?

No!

Well kind of yes. For probably 90% of projects, way more than you’d think, the way we’ve done this to date isn’t doing what we need it to do anymore, but the WAY it gets pared down is really important.

We know what we’re doing, but, unless we’ve worked with a client before, and have trust built up, they don’t know that we know what we’re doing. And we still need alignment on project goals, and the specific application of what we know to the problem at hand. If you just chop out the “right thing” phase, and don’t make any changes to the “thing right” phase, you might not, in fact, be building the right thing OR the thing right. 

Even with all this knowledge, these things aren’t really one-size-fits-all. There’s always something specific to a need, audience, client, or project that benefits from custom tailoring to get to the right solution. Lots of places will sell a template, but even with something off the shelf, you’ve still got to make the solution fit inside the container. At best, that works kind of OK, at worst it’s a Procrustean bed situation, and in all cases, there’s work involved. 

I’ve seen “just use AI” proposed as a solution, but that doesn’t really solve anything. AI is a pretty efficient way to aggregate knowledge on a particular topic, but it’s not very good at applying a specific solution to a specific problem. It IS pretty good at appearing to do this, but you end up with general recommendations dressed up in specific language, a kind of platitude-in-expert-clothing. Sometimes the idea is that the product owner will do discovery, and then just tell the executional team what they need to know. But that creates all the problems of a disjointed two phase process, with the added problem of having no one to prod and stress test the product owner’s findings. This kind of work goes better when it’s collaborative, so the project can’t over index on anyone’s personal idiosyncrasies or strategic myopathy. Besides, whatever you do to set up the work, new questions always arise during the executional phase anyway.

The high level solution should be pretty obvious: hire experienced people who know what they’re doing. You can keep the pre-work to a minimum, focused on alignment rather than discovery, have a team that can answer strategic questions around when those questions arise, from beginning to end, and dramatically shrink the timeline of delivery.

To actually do this, we need a mental shift: collaboration over presentation, recommendation over observation, problem solving over deliverables. Instead of anchoring to presentations and deliverables, we need to think about collaboration and answering questions. If both clients and their consultants know most of the basics, and a lot of the specifics, you can focus on filling in the gaps in knowledge and crafting a truly bespoke solution. Instead of doing 10 hours of interviews over three weeks, then spending another week pulling together a presentation, get everyone together for a three hour workshop. You don’t need to present the findings, because everyone participated in finding them! Instead of trying to game out all possible questions ahead of time, address them when they arise. Focus on what you need to move forward. For a consultant with experience and depth, a couple hours of prep and an hour-long working session with key stakeholders is often enough to get unstuck. No need to hold up execution! And if you’re working together to solve problems, everyone has a lot more trust and ownership of the solutions you come up with.

As a practical matter, you need to do some things. First, you want a stable team over the whole life of the project. And they need to be experts in whatever they’re doing. Second, you need client and consultant teams to work together, collaboratively, to identify and fill gaps while aligning on project goals as quickly as possible. Push as much of this as possible back into the scoping work that happens before a contract is even signed! And speaking of contracts, we shouldn’t commit to specific deliverables that aren’t directly tied to project delivery. For the vast majority of projects, we do not need a formal competitive review, personas, or journey maps. Work of that sort should be conducted only as-needed, be based on prior knowledge, and worked into the ongoing delivery schedule. Projects still need milestones but those should be tied to actual feature design (or development, depending on how agile you’re being). Do what you need to get done to move forward, then measure project momentum by progress made towards completion. 

I’ve been pushing this approach for a few years, and the difference is dramatic. Fivish years ago, for a redesign and replatforming of a health system website, I used a two phase approach that took about a year. Three years ago, the same basic project took half that amount of time. Last year, I worked on a very similar project for a legal client, and the timeline was about the same, but the team was half the size. In each case, the quality of the work increased with the total number of hours spent shrank. I don’t think we’ve reached the end of this evolution, especially as AI use gets out of its silly/stunt era and starts delivering real efficiency. I have no idea where it ends. And there are hard questions about how you develop junior talent when a small, senior team is ideal. But right now, it’s irresponsible to keep doing things the way we have been for the past 20 years.

Why Cheaping Out on Content Is a Terrible Idea

First off, I promise this isn’t about AI. I mean, it does involve AI; I’m going to say “AI” a whole bunch. But it’s not about how AI Changed The Game and Everything Is Different Now, so you can read it as a little break from wall-to-wall OMG AI discourse. 

This is about why so many organizations seem super excited to cheap out on content, why this is a bad idea, and why I think the new paradigm in web search (ok, fine, AI search) is actually helping create a world where good content is very valuable.

The urge to cheap out on content comes from the incentives of the current online ecosystem. The ability of anyone to make money on the internet is heavily dependent on a small number of platforms, mostly Google and Meta (G&M from here on out), who both gatekeep content and run massive platforms to serve targeted advertisements. More or less, if you want to make money online, you’re probably reliant on one of these companies, either giving you money for hosting ads on your site, or by directing customers to your site so they can buy whatever you’re selling. G&M are essentially toll roads making money while providing the easiest paths for customers to reach online businesses. All of this ad and search activity takes place in competitive marketplaces with abundant data, which lowers both costs and revenue, putting pressure on businesses to cut costs everywhere. None of this is inherently sinister or bad! G&M provide services that many people find useful. But some time in the early 2000s, some businesses realized that if you can adequately game the toll keepers, you can do away with most of the products/services/useful information that were originally the whole point, and make money off of traffic and ads by, basically, tricking people into clicking your link, then giving them the runaround to keep them on your site long enough to rack up ad views. This is super efficient, getting something for close to nothing, and so has led to a situation where more and more of the internet has stopped trying to do much of anything other than game Google and Meta to achieve the cheapest possible click. 

In this world, it’s easy for content to stick out as a problem. Everything has to be optimized around the needs of the system, which is to say, simplifying somewhat, SEO. SEO manages to be both simple and obscure. In order to get seen, sites need content and page metadata to use the keywords people are searching for. Ads are served based on the same keywords. Reciprocal links are important, as is traffic. The exact mix of these factors is constantly changing, to make it harder to cheat. Content has evolved to be simple, keyword heavy, and easy to change. Why spend a long time writing the perfect article, or an amazing product description? All that matters is getting a ping on a trending search term so your link can be one of the top results. Content exhibiting human messiness can be a problem. Spending any more time than absolutely necessary producing it starts to seem like a real drag.

It doesn’t help matters that many tech and business and business-tech/tech-business people are suspicious of writing. It’s either too easy and everyone can do it, or it’s too hard, it takes too long, and it doesn’t matter anyway. It’s hard to plan and measure content: how much time does it take to make good content? What effect does good content have? What does “good” even mean in this context? How many of these writers even went to MIT? Humbug!

What gets measured gets managed, and to admit that our ability to measure is extremely skewed by the nuts and bolts of Google’s online metrics is to call the paradigm into question. Anything that’s hard to measure must not be important, CAN’T be important, otherwise the whole thing collapses. I mean, it really doesn’t, but try arguing with an MBA about it.

So when we’re looking to cut costs, content seems like a good place to do it. Outsource it, or, even cheaper, just let AI do a good-enough job and cut all the hassle. The first AI use case I ever heard tech and business people get excited about was, “How about we fire all our writers?!” But AI didn’t cause this; they ALREADY wanted to fire all the writers.

This is all a massive mistake, because the quality of online content, especially writing, is becoming more important than ever before.

“Quality,” here, isn’t vague. It means three pretty specific things: content will need to do something for the user, not just trick them into a click, it will need to be clear, useful, and well organized, and it will need to be distinctive.

Even before AI search, Google and Meta were working hard to keep people on their platforms as much as possible. You’re not the customer of G or M. G&M’s customers are the businesses selling ads, and the best way to maximize ad views is to keep users right where they are, instead of sending them to a web page where they might not see any ads at all! Even a straightforward, non-AI Google search from four or five years ago surfaced information formatted to remove the need to really go anywhere. AI accelerated this evolution, and now a significant chunk of web traffic is looking at content parsed into a summary that is experienced without ever leaving the search or chat context. It seems likely this situation will continue developing apace. In a world with no clicks, there’s no point to clickbait. Content that merely exists to bring someone to a page and keep them there long enough to trigger an ad view is no longer content worth anything at all.

And AI enabled search is different in what it surfaces as well. AI search responds to structured, well organized content that responds directly to the sorts of questions a person might ask when searching. AI search rewards content that is in active communication with the audience, and with its context. Not just adding words to a page’s metadata, but constructing one side of a three way conversation with both the user and related content. 

It also needs to communicate effectively. If you’re selling a product, for example, the content needs to describe that product. I’m always shocked by how many online stores fail at this basic task. There are understandable reasons for that, but if you want an AI search to pick up your product, and present it in a way that makes people want to buy it, the description you give needs to be both accurate and helpful. It’s the same for businesses that aim to inform or entertain. They need to talk about themselves in ways that make search recommend them, much like a person would recommend a movie they liked, if that friend also had encyclopedic knowledge of its cast, crew, production details, reviews, and streaming availability.

Finally, content needs to be distinctive. Marketing guru Byron Sharp has written on the importance of distinctive brand assets, and it’s pretty well accepted that “brand identity” is important. Companies spend eye watering amounts of money on branding. But, how do you be distinctive online if many users never even make it to your website? You can’t rely on visual design or innovative interaction design. All you have is what you say about yourself. Your voice needs to be authentic and strong enough to carry, no matter where, or how, it’s heard. 

If you’ve recently spent any time on a corporate web site, news site, or, god help you, LinkedIn, the situation is DIRE. Corporate content, AI infused or not, is stripped of all personality, and, sometimes, even specific details that could prevent repurposing and reuse. Mid to low tier online news, even from many formerly reputable sites, is constructed entirely as clickbait. I’m not even going to try to get into the whole deal with online recipes. And professionals trying to keep up a posting schedule to maximize their views on LinkedIn are creating personal content in the style of the most annoying marketing SPAM, bland and unmemorable, everyone speaking in the same, interchangeable voice, and chopped into the shortest possible sentences.

Each on its own line.

Because that’s the way to add some drama.

I don’t relish being the bearer of this message.

But it’s the truth.

I owe you that much.

ANYWAY! Sorry. The point is, none of this stuff is distinctive. It’s also, for the most part, not useful, valuable, or interesting. And people don’t like it! Research finds that clickbaitheadlines and low quality content damage credibility, and that people are distrustful of AI generated news. On Amazon, a flood of AI books is driving reader backlash and getting worse metrics than human written works. 

Meanwhile, over on Substack, regular people with expertise, insight, and/or an entertaining point of view are making actual money ($450 million in subscriber payments in 2025!) selling subscriptions for access to their writing. The New York Times, which has leaned into becoming a first-stop destination for paying subscribers, is making money, while the Washington Post, which has followed the typical Silicon-Valley-disruptor-efficiency playbook, is operating at a loss. And Reddit, where users upvote funny and insightful posts and comments while aggressively punishing anything that seems like SPAM, is growing at a rapid pace. We’re rapidly heading towards a reality, where anything with a point of view, that genuinely gives something valuable to readers, will shine out like a beacon for weary users AND drive measurable results. Arguably, we’re already there.

Actual people value trustworthiness and usefulness. Companies thrive on having distinct voices and identities. The needs of both people and businesses are completely mismatched with what our measurement/revenue framework has been rewarding. The paradigm is collapsing, because it’s stopped providing what anyone actually needs.

Search models will be tweaked to deprecate low quality content, and to reward useful and engaging content. It won’t be enough to trick someone into clicking a link, they’ll need to actually get what they’re looking for. The whole intermediated model that’s allowed people to make cheap money getting in between a user and their goal is rapidly changing. It’s possible, I think likely, that we’re about to accidentally create a world where the quality of content is both measurable and rewarded. 

In the very near future, content will need to be structured, useful, descriptive, engaging, responsive to context, part of a conversation, authentic, and distinctive enough to convey identities and get messages across, wherever someone encounters it, even if it’s filtered through a Google summary. Simply put, it’ll need to be good. Like, really good.

Think about a time someone said something so perfect that you adopted it into your own vocabulary, a little piece of that person emerging whenever you talk about that subject. People my age pepper their speech with old ad slogans, Simpsons dialog, and Big Lebowski and Spinal Tap references. Young people use “brain rot” slang and memes. None of that is high art or great literature, but it’s all categorically different from the weird, sanded down, aerodyne mush that’s taken over so much online content. It’s all strong enough to survive years, decades, of blending and remixing, identity intact, because it’s all, in its own way, good writing; distinct, specific, memorable, and full of life. That’s what your content needs to be.