Agents for the Brief & Programme

The first and least controversial value of an agent at the briefing stage is pure structuring: taking unstructured, scattered inputs and organising them into a coherent, reviewable form. This is knowledge work with a clear shape - inputs in, structured document out - and it is exactly what agents are good at, so it is a sound place to delegate boldly and verify carefully.

Concretely, feed an agent the raw material of a project - meeting notes, an email thread, a client's wish list, a rough budget, a set of reference images described - and it can do several genuinely useful things. It can extract and organise requirements, pulling the stated needs out of the mess and grouping them into a structured list rather than a jumble. It can draft a space programme: a room-by-room schedule of the spaces implied by the brief, with a first pass at areas, capacities and key requirements for each. It can draft room data sheets - the detailed per-space requirements (area, occupancy, servicing, finishes, environmental needs) that later stages depend on. It can draft adjacency and zoning logic - which spaces should be near which, which kept apart, how the whole might zone - as a matrix or diagram to react to. And it can produce the brief document itself in a clear, consistent, professional format, saving hours of writing and formatting.

The pipeline in the figure captures the healthy flow: messy inputs go in, the agent structures and drafts, and out comes a draft brief with the areas, the adjacencies, the open questions and - crucially - an explicit list of the assumptions it had to make. That last item is the difference between a useful draft and a dangerous one. Because so much of a brief must be inferred from incomplete inputs, an agent instructed to flag its assumptions and gaps turns a pile of guesses into a checklist you can resolve with the client, rather than a smooth document that hides its own uncertainty. For a small studio, this structuring can turn a daunting, procrastinated task into a fast first draft - a real productivity gain at a stage that sets up everything downstream. The gain is only safe, though, once you treat that draft as exactly that: a draft to interrogate, not a brief to adopt.

Beyond structuring what is there, an agent can help with something more active and more valuable: interrogating the brief to find what is missing, contradictory or unexamined. A good brief is not just an organised list of what the client said - it is the result of good questioning, because clients often do not know, cannot articulate, or have not reconciled what they actually need. Traditionally the designer supplies that questioning through experience and dialogue; an agent can augment it, and for a less experienced designer it can be a genuine aid to asking better questions.

Ask an agent to interrogate a draft brief and it can do useful work. It can spot gaps - requirements a project of this type usually has that this brief has not mentioned (accessibility, storage, services, circulation, future flexibility, statutory needs). It can surface contradictions - a wish list that wants more than the stated budget or site can hold, two requirements that fight each other, an area schedule that does not add up to the plot. It can question vagueness - turning 'a nice big kitchen' into the questions that pin it down (for how many, doing what, open or closed, what budget). And it can propose the questions to take back to the client, which is often the most useful output of all: a structured list of the decisions and clarifications the brief still needs.

This is the agent at its best in early design - not deciding, but pressure-testing, in the manner of a thorough colleague who reads the brief and asks 'but what about...'. Two disciplines keep it honest. First, the agent's gap-finding is drawn from typical patterns, so it will miss the specific and occasionally invent needs that do not apply here - use it to prompt your thinking, not to define the requirements, and bring your own knowledge of this client and this project to what it surfaces. Second, and more importantly, the answers to its questions come from the client, not the agent. When the agent flags that the brief has no clear budget or does not say how many people the space serves, the resolution is a conversation with the client, not the agent's plausible guess. The agent's job is to ask the sharp questions; the client's is to answer them; yours is to run that dialogue and hold the requirements to what is real. Used that way, an agent makes your briefing more thorough. Used to fill the gaps with its own inventions, it manufactures a confident brief for a project that does not exist.

The space programme is where agentic briefing meets hard, checkable fact - and where the verification duty bites hardest, because the programme is made of numbers that become commitments. Areas, occupancies, quantities, cost benchmarks, statutory minimums: get these wrong and you carry the error, unnoticed, into every later stage. An agent will draft all of them fluently and confidently, and some of them will be wrong.

Start with areas. An agent can draft a schedule of accommodation with a floor area for every space, and it is genuinely useful to have that first pass - but every number in it is a proposal to verify, not a measurement. The agent's areas come from typical figures and its own inference, which may be reasonable, outdated, drawn from a different context (a US or European norm applied to an Indian project), or simply wrong. Check each area against real requirements - the actual furniture and use it must hold, the real occupancy, the applicable standard - and check the totals against the site's buildable area and the budget. Any area or quantity that touches a statutory limit - a minimum room size, a required number of sanitary fixtures, an accessibility clearance, an occupancy load, a fire-egress width, an FSI or coverage limit - must be verified against the actual code, here NBC/SP 7 and the local byelaws, because an agent's memory of a code clause is exactly the kind of confident, checkable, sometimes-fabricated fact you can never take on trust (Module 3.1, 4.4).

Adjacency and zoning logic, shown in the figure, are lower-stakes but still need your eye. An agent can draft a sensible-looking adjacency matrix - reception near waiting, consult near lab, stores tucked away - and it is a useful starting logic. But adjacency is about how these particular people will actually use these particular spaces, about workflow and privacy and flow and delight, which the agent knows only generically. Treat its matrix as a first hypothesis to test against real use, not a decided plan. The healthy pattern across the whole programme is the same one that runs through this module: let the agent draft the structure and the numbers fast, then verify every number that becomes a commitment, verify every code-touching figure against the actual standard yourself, and bring your judgement to the relational logic the numbers cannot capture. The agent gives you a programme to check; making it correct and right is your job, and your responsibility.

There is a deeper risk at the briefing stage than a wrong number, and it is easy to miss precisely because the agent's output is so smooth: the risk that the brief quietly stops being the client's and becomes the agent's. A brief is a statement of what a specific client needs and wants, shaped by their life, their constraints, their aspirations and the things they have not yet articulated. An agent does not know any of that; when it fills a gap, it fills it with what is typical, and a brief made of typical is a brief for a generic project, not this one.

This matters because the brief is the foundation everything else is built on, and errors here are the most expensive of all - a mistaken or generic requirement, unnoticed, propagates through concept, design and documentation and is discovered late and at cost. So the discipline is to keep the client firmly in the loop and the requirements firmly theirs. Use the agent to structure, to draft, to interrogate and to surface questions - and then take the gaps and questions back to the client and let their answers, not the agent's guesses, define the brief. The agent's assumptions list (from the first section) is the tool for this: every assumption it had to make is a question for the client, and a brief is only agreed when those assumptions have been confirmed or corrected by the people whose project it is. A confident draft with unexamined assumptions baked in is worse than an obviously incomplete one, because it invites you to skip the conversation that matters.

And the professional judgement stays yours throughout. Deciding what the brief should be - reconciling the client's wants with the budget, the site, the code and what is genuinely good for them, sometimes telling them what they need rather than only what they asked for - is an act of professional judgement and care, not a structuring task. The agent can lay out the options and tensions with real clarity; the resolution is yours, made with the client, and you answer for it. Signed off, the brief is a commitment - to the client, and often to a budget and a programme - and commitments rest on a responsible human, never on a tool. Use the agent to make the brief clearer, more thorough and faster to produce; keep it, always, the client's brief and your professional responsibility.