AI for Architects: Fee Proposals, Planning Packs, and the Drawing Line

Architecture practices sell judgement and design, but the working week fills up with everything else: fee proposals, planning documents, meeting minutes, RFI responses, tender queries, and the long email tail of every project stage. For a 3–50 person practice, AI's realistic value is almost entirely in that written layer — and the profession's own rules make the boundary unusually easy to draw: AI drafts documents; it does not do architecture.

The work AI does well in a practice

Fee proposals and bids. Most practices assemble proposals from the same ingredients every time — scope by RIBA stage, team, programme, fee basis, relevant projects. AI drafts from your precedents in your voice, tailored to the enquiry, in minutes rather than an evening. The judgement calls — scope boundaries, the fee itself, the risk flags on an awkward brief — stay yours. Faster, better-tailored proposals win work; this is the practice-sized version of the speed effect we see across every sector.

Planning and pre-app documents. First drafts of Design and Access statements, planning statements, and supporting documents assembled from your drawings' narrative, site notes, and precedent documents — with the local authority's validation checklist run against the pack before submission. The design argument is the architect's; the document production around it doesn't have to be. Heritage statements and the like deserve extra care: AI is fluent about places it has never seen, so anything site-descriptive gets checked line by line.

Project correspondence. Meeting minutes from a recording, issued the same day with actions assigned. RFI and tender query responses drafted from the specification and drawing set for an architect to verify — the verification matters, because an answer that contradicts your own spec is a claim waiting to happen. Chasing consultants, contractors, and clients for the information the programme is waiting on.

Practice admin. Timesheet nudges, resourcing summaries, CPD records, appointment document first-drafts from your standard terms. Ordinary SME office automation, no drawings involved.

The boundary the profession draws for you

Design decisions, regulation compliance judgements, and anything you sign as an ARB-registered professional stay human — your PII insurer, your regulator, and the Building Safety Act's competence expectations all point the same direction. AI-generated imagery needs care with client expectations (a seductive render of an unbuildable idea is a conversation you'll pay for later), and client-confidential material belongs in business-tier tools under the usual data rules, not personal chatbot accounts. None of this is a reason to abstain; it's the same AI drafts, a named human owns pattern the rest of this series applies to law, healthcare, and construction — the industry you spend your working life alongside.

What it costs

The standard ranges apply, and practices mostly live at the assistant tier (£15–£60 per user per month) for a long time before a connected build is worth it — the document wins need your precedents and templates more than they need system integration. The £5,000–£25,000 build enters when you want project correspondence wired to your job-running system and document management.

Where to start

Fee proposals: high value, weekly volume, zero regulatory exposure, instantly measurable in turnaround time and win rate. Minutes and correspondence second, planning document assembly third once you trust the review habit. The full-practice view — where the hours leak across jobs, and what to fix in what order — is what AI Mapping produces; the two-minute readiness assessment gives you a score straight away, no email needed.