In the history of business, every system was designed for one kind of actor.
Advancements in artificial intelligence have introduced a second actor. Understanding what changed — and how an organization evolves to meet this new world — is the beginning of everything we do.
A new kind of actor
Before computers, information lived in cabinets, organized by drawer and folder and identified with labels: a system designed for human hands, human eyes, and the human brain. Computers changed the medium, not the design: files, folders, directories reflect the same cabinet, on a screen. Over time, other systems were adopted where machines were the audience, but wherever a person was the one working with the information, the basic cabinet design held. It held because it fit: a person read the information, made the judgment, did the work, filed the result, and the cabinet was shaped for that purpose.
That is no longer the whole picture. A new kind of actor now works inside businesses. Businesses have always had software and forms of automation, but this actor is different in kind: it reads, interprets, understands, generates, acts. Today these actors are mostly large language models, and they are neither lesser versions of a person nor greater ones. They are differently capable: new skills, new limitations.
Here is the point most businesses haven't absorbed: a system optimized for one actor isn't wrong — it's the correct approach for a past paradigm. When there was one variable, you optimized for it. There are now two, and optimizing a two-variable system is a different problem: you have to account for both, each within its own skills and limits, and for how they work together.
Humans
Skills
Judgment. Accountability. Intent — knowing why something was done. Context that was never written down anywhere.
Limitations
Bandwidth. One pair of eyes, one working day. Recall fades, attention wanders, consistency is expensive.
Models
Skills
Pattern recognition at extraordinary scale — that is fundamentally what they are, probabilistic pattern recognizers. They read everything, tirelessly, in parallel, and synthesize across more material than any person could hold.
Limitations
A bounded working memory — the context window. A model can only reason about what's in front of it at that moment: a brilliant analyst with a small desk. And because it's probabilistic, what isn't stated gets inferred — often well, not always right.
Neither actor's limitation is a flaw. Each is simply the shape of the actor. The question for a business isn't which one to favor — it's whether its systems let each work at full strength, and let each cover for the other.
What a one-actor system looks like in a two-actor world
None of these are tool problems. They're what a one-actor design looks like the day a new actor shows up.
Every new tool demos beautifully and underdelivers on your own material — because your material was never arranged for it.
The answer is in there — somewhere. Whether it's found depends on who remembers where.
People spend their scarcest resource — attention — on work the other actor is better at: searching, collating, cross-referencing, re-keying. Meanwhile the model gets asked to do what people are better at, and disappoints.
What the new actor produces has nowhere to land. It appears in a chat window, gets used once, and evaporates — effort that also doesn't compound.
And beneath all of it: years of proprietary data — the record of everything the business has learned that nobody else knows. It's the one asset competitors can't buy, and it's exactly the material the new actor is built to harness. A one-actor design leaves it sitting untapped.
What we do
We understand
The business first, always.
Nothing about the right answer can be derived from technology alone. It comes from how a business actually runs: what decisions get made, by whom, on what evidence; how work moves from started to done; where information really lives, as opposed to where it's supposed to live; what the business wants to achieve that it can't today. So the work begins by stepping into the business — its operations, its people, its data as it is.
This isn't a preliminary. The map of how your business works is precisely what the new actor doesn't have — and it's what every later stage encodes.
We design
The data layer: everything the business knows, arranged as a platform.
The data systems businesses run today still mimic the one that existed before computers. That design is not flawed — it reflects the way humans think and the way they work, and for the human actor it may be close to optimal. There are improvements to make, but the core design is sound.
What it lacks is a second way in. We design the data layer around the strengths and limitations of both actors, so that each operates at full strength: the way people interact with their information stays intact, and alongside it we add a new route in — a new interface, a new point of access, optimized from first principles for the other actor. Both read the same underlying information — documents, records, messages, numbers, structured or not — and each sees it, receives it, and acts on it in the form designed for them.
A second way in is necessary, not sufficient. Whether the layer delivers depends just as much on things no diagram shows: whether the data is any good, who actually owns the process it serves, what is allowed to see what, and whether the business is willing to change how the work itself gets done. We treat those as inputs to the design rather than obstacles discovered after it. That is why the work begins with understanding the business, not with its data.
This is not cleanup. It's the platform layer — the foundation and the utilities. Everything in the next stage is built on it, and is only as good as it.
We harness
Turning potential energy into kinetic energy.
The right data layer creates potential energy: everything the business knows, arranged and ready. Harnessing it means converting that potential into kinetic energy — the right tools on top, the right models, the right systems on the right workflows, designed for both actors: automation for work that shouldn't need a person; assistants and agents pointed at real business material, operating inside governed limits; analysis, reporting, and models the data can now actually support; connections that make separate systems behave like one.
And a well-built system doesn't just convert energy — it conserves it. In most businesses, work dissipates: an answer gets found, used once, and forgotten, and the same energy is spent finding it again. We design against that loss. What either actor produces feeds back into the platform, so today's work adds to tomorrow's instead of evaporating. That is what it means for effort to compound.
What gets built is chosen from the business's goals, not from a catalog — people where judgment matters, models where scale does.
We sustain
The field moves. The business moves with it.
Keeping the business at current best practices as they change — and they change quickly. Watching the landscape for technologies that would benefit this business, and saying so when they appear — and, just as often, when they're noise. Implementing them when they're worth it. And measuring that it all still works — on the business's own material, not a vendor's benchmark.
Our approach
First principles, every time.
Every business runs differently; the right answer is derived from yours, not templated from someone else's.
Vendor-agnostic, outcome-aligned.
We survey the best of what exists, give you the real options, and implement and manage the result for you. We have no software of our own to sell — so the only thing our work can optimize for is your outcome.
The simplest structure that works.
Complexity is a cost paid forever.
Deterministic where it should be deterministic.
If a process wants to be a rule, it should be a rule — exact, repeatable, boring. Models belong where judgment and pattern meet scale. Sometimes the right answer is not AI.
The two-actor era isn't coming. It's here.
The businesses that thrive in it won't be the ones that bought the most tools. They'll be the ones that arranged themselves for both actors — whose systems let people do what only people can do, and let models do what people never could. That arrangement doesn't happen by accident, and it doesn't come in a box. It's designed — from the business outward, starting with its data. That's the work we do.
