Product
What We Chose Not to Build
Most product decisions are about what to add. The decisions that defined MAX are about what was taken away, and about the price one must be willing to pay to hold a refusal over time.
There is, in the design of any serious product, a particular kind of decision that almost no one talks about in public. These are the decisions of refusal. Not a feature one chooses to add, but a feature everyone expects, that seems obvious, that sells well, and that one deliberately chooses not to build. These decisions, by definition, do not show up in the finished product. They leave a void no one notices, because there is nothing to notice. And yet, they are often what defines the product more strongly than anything actually built.
For MAX, the decision of refusal that shaped everything can be summed up in one sentence. We chose not to build an interface. No MAX application to open. No MAX platform to learn. No new place to go. This decision seems anecdotal when read, and it is, in reality, the heaviest choice we made. Everything that follows in MAX, the architecture, the position vis-à-vis firms, the economic model, and even confidentiality, flows from that initial refusal.
The pull in the opposite direction
To understand why this refusal was difficult, one has to see how strong the pull in the opposite direction is when building a legal AI product. The path of least resistance, and it is the one almost all our competitors took, is to build a visible application. A platform. A new place that lawyers open, log into, learn and adopt.
There are excellent commercial reasons to do this, and they are not bad ones. A visible application is easy to demonstrate, because there is something to show. It is easy to sell, because there is something to point at during the pitch. It is easy to put a logo on, easy to market, easy to make feel like a product. It has a name the user remembers, an icon they see in their dock, an interface they can criticise or praise. Everything the enterprise software industry has learned to do in forty years pulls in this direction, and pulling in that direction is a default choice it takes conscious effort to undo.
If we had built an interface, MAX would have been easier to present, easier to sell, easier to grow in revenue over the first twelve months. There is a cost to having refused, and we pay it every day, in the difficulty of explaining a product that cannot be shown in a demo. The few competitors who took the same direction know this cost as well, and most of them have, at one point or another, been tempted to relent and add a visible application to their offering. But that cost is voluntary, because the decision was deliberate, and because what it costs in the short term, it gives back, multiplied, in the long term.
The easiest thing to build is a place for people to go. The hardest thing to build is a product they never have to go to.
The other way, and what it requires
MAX went the other way. There is no MAX interface to open, no MAX application to learn, no new destination, no extra tab the lawyer has to remember to go to. MAX lives inside the tools lawyers already use, and it operates from within them. The user does not visit MAX. MAX arrives where the user already is, does its work, and leaves without having asked for attention. The product, in the most literal sense, asks for nothing.
This sounds like less. It is, in fact, considerably more, and considerably harder. Building a product that the user does not have to notice is not a matter of hiding an interface. It is a matter of building everything that lets the work happen without one: the orchestration that coordinates what would otherwise be separate tools, the memory that carries the matter from one interaction to the next, the methodology that encodes how good legal work is actually done, and the governance that keeps all of it accountable. None of these things shows. All are indispensable for the product to function.
Invisibility, in an enterprise software product, is not the absence of engineering. It is the result of a great deal of engineering that has accepted not to show itself. For a user not to have to learn MAX, MAX has to have learned the user. For a lawyer not to have to wonder where their data goes, it has to go nowhere. For a firm not to have to organise the deployment of a new platform, MAX has to lodge inside the existing stack without disturbing it. Each of these conditions has a significant engineering cost, and it is that cost that makes invisibility possible.
Removing the need for the user to notice the system is architecture, not decoration. It costs more, not less.
The decision had a second consequence, which we had not fully anticipated
When we decided that MAX would have no separate platform, we also, without fully realising it at first, decided something about confidentiality. Because there is no MAX destination for the work to travel to, the data does not have to leave the firm's environment to be useful. What makes MAX invisible to the user and what makes MAX safe for the firm turn out to be the same decision, seen from two angles.
This equivalence was not the initial motive for the refusal. We did not, when laying down the decision, have in mind to solve the question of confidentiality by this route. It was as we progressed in the building that we realised the two properties flowed from each other. No interface means no destination. No destination means no transit. No transit means no exposure. And no exposure means a confidentiality that does not need to be promised on top of the product, because it falls naturally from its architecture. We added confidentiality as a feature on top of the product, in the communication; but in truth, it was already there, built into the initial decision of refusal.
Invisibility and confidentiality are not two features. They are one architectural decision, seen from two sides.
Why this decision endures
The decision to build MAX this way was not a marketing instinct, and it was not a stroke of boldness either. It was the consequence of thirty years of watching what makes enterprise software endure and what makes it disappear. The pattern is consistent and a little brutal. Tools that demand attention end up competing for it, and attention is the scarcest resource in any professional's day. They win it for a while, then lose it to the next tool that demands it more loudly, more visibly, with a shinier interface. Layers that disappear into the work, by contrast, are never in that competition at all. They simply become part of how the work is done, and they outlast the cycles that produced them.
This pattern shows up in the history of enterprise software, wave after wave. The tools that have disappeared are almost always those that demanded the most attention during their heyday. The layers that have lasted are those that lodged beneath the uses, beneath the interfaces, beneath the attention. The semantic layer of Business Objects, which inspired our choices, is invisible to 99 percent of those who benefit from it every day. That is precisely what explains why it still exists. A layer one does not see is a layer one does not replace, because there is nothing visible one could want to replace.
This is the test we held ourselves to. Not whether MAX would be impressive on the day it launched, but whether it would still be there, quietly load-bearing, long after the excitement of any particular model or interface had faded. Building for that test imposes renunciations that would have been simpler not to make. Giving up the interface. Giving up the dazzling demo. Giving up the engagement metric one could have presented to investors. Each of these renunciations has an immediate cost. None has a lasting cost.
The way to pass that test was not to add the most. It was to need the least. And we made that bet, knowing that it made MAX harder to tell as a story, because we were convinced it made it harder to replace.
Good design is not what was added. It is what no longer needs to be there.
What this refusal means for the firm that chooses MAX
So far, we have spoken of the refusal from the point of view of the one carrying it, that is to say us. There is a second perspective, more important still, which is that of the firm that chooses MAX. For that firm, the refusal translates into concrete properties no product built otherwise can offer, and they are worth enumerating.
First, the firm has no platform deployment to organise. There is no data migration, no user training on a new interface, no transition period during which two tools coexist and compete. MAX lodges inside the existing environment and operates from within. The adoption curve that usually weighs on team productivity for several months simply does not exist.
Second, the firm does not depend on the commercial longevity of a third-party platform. When a vendor whose platform has become central to the work changes its economic model, is acquired, or shuts down, the consequences for the client firm are severe. MAX does not create this dependence, because it does not own a platform whose disappearance could carry away part of the work.
Third, and this is perhaps the deepest point, the firm does not pay, in its daily work, the attentional cost of an additional tool that demands to be learned, opened, and managed. This economy is invisible in the first weeks, and it becomes considerable over time. A lawyer whose work environment accumulates eight interfaces spends a non-negligible share of their energy moving from one to the other. MAX removes this line from the ledger without asking anything in return.
The cost of a tool is not measured by its subscription. It is measured by what it takes from the day of the person who uses it.
These three properties, taken together, sketch an asymmetry of cost that is not visible at the moment of the buying decision, because the buying decision looks at features and the demo. It becomes visible over time, in the three or four years that follow, when the firm that chose MAX realises it kept its independence from a vendor, saved a costly deployment, and preserved its teams' attention for legal work. The initial refusal, seen from this angle, is no longer a renunciation on our part. It is a structural gift to the firm that benefits from it.
