AI Blueprint

Where AI fits in the architecture firm

ak-labz · Updated 2026-05-30

90-second scan

Architecture firms with 10 to 100 staff feel every operational pain of a bigger firm without the budget for a Big-4 implementation. Most have not bought into industry-specialized platforms like Synthesis or Newforma Konekt yet workflows are clearly outgrowing shared drives and tribal memory. AI agents finally have enough leverage to change that.

This blueprint maps where to look. Five universal capabilities every firm should have running: meeting capture, document intake, smart inbox, project knowledge, firm knowledge. Two are off-the-shelf; three need a custom build to fit your stack. Three architecture-specific opportunities worth a custom build: public RFP response, construction admin routing, project precedent search.

Where to start depends on what is loudest right now. Meeting capture if your principals are in 4-8 hours of design reviews, client meetings, and construction calls a day. CA routing if the lead architect is the bottleneck on every RFI. Project knowledge retrieval if associates interrupt a principal every time they need a precedent. Pick the loudest one. The rest is second-project work; it rewards the foundation being in place.

Section 1

Five universal capabilities every firm should have

Cross-functional AI capabilities every firm needs, whatever the industry. Foundation for the architecture-specific opportunities in section 2.

Meeting capture and action items

The pain

The principal who took the meeting is the only one who knows what was actually agreed. Minutes get written by an associate who wasn't deciding, or they don't get written at all. Decisions get re-litigated next week. Action items live in someone's inbox until they don't.

The AI solution

Records the call, identifies who agreed to what, drafts the recap email and action items for the principal to approve. Action items push into your project tracker. The principal hits send instead of writing.

Buy, build, or wait

Buy. Granola, Fathom, Fireflies, and Otter are all mature. Teams and Zoom have it built in. The recording, transcription, recap, and integrations into HubSpot, Slack, and Notion are settled buys. At firm scale, a thin custom layer on top routes action items by call type and links them to the right project. That layer is a small build, not a separate product.

Time back

~3-5 hrs / week per principal

Document intake

The pain

Invoices, change orders, insurance certificates, lien waivers, owner-issued letters. They land in inboxes, get printed, get re-typed into accounting, get forwarded to whoever needs to decide. Three different people read the same change order trying to figure out which project budget it hits.

The AI solution

AI reads the document, extracts the fields, routes it to the right person, flags the ones that need a human decision. Standard fields: vendor, amount, project number, dates. People stay in the judgment loop, not the data-entry loop.

Buy, build, or wait

Mostly build. Tipalti, Bill.com, and Stampli handle accounts payable well. Buy those. Project-side documents need a custom build: RFIs, change orders, COIs, lien waivers, owner letters. The routing into Newforma, your accounting tool, and the project folders is firm-specific. So are the exception rules.

Time back

~5-10 hrs / week firm-wide

Smart inbox and email triage

The pain

Senior staff get 100+ emails a day. Half are CC'd RFIs, submittal notifications, consultants chasing decisions. Most don't need an action. A few are real decisions. The signal-to-noise drowns the morning.

The AI solution

AI sorts incoming email into action, FYI, and decision buckets. Drafts first-pass replies for repeat patterns. Surfaces the three things that actually need your attention today. Inbox becomes a triage queue, not a slot machine.

Buy, build, or wait

Buy + build. Superhuman AI or Shortwave for individual inboxes. Buy those. The firm-wide rules and the shared inbox triage on info@ and projects@ need a custom build because the routing logic and the firm voice are yours.

Time back

~10 hrs / week per senior person

Project knowledge retrieval

The pain

An associate needs to know what was decided on this project two months ago. The answer is in an email thread, an RFI in Newforma, minutes in OneNote, a Bluebeam markup. There is no one place to ask. They guess, or they interrupt the PA at 4pm Friday.

The AI solution

AI search across this project's actual sources: email, RFIs, meeting minutes, drawings, markups. Ask in plain English, get the answer with a link to where it came from. Anyone on the team can self-serve instead of pulling the PA off CA work.

Buy, build, or wait

Build. The value is the integration to your actual project stack: Newforma, OneDrive, Bluebeam, email threads. No off-the-shelf product covers it because every firm's stack is different.

Time back

~30-60 min / day per architect

Firm knowledge retrieval

The pain

Marketing needs project sheets, past spec language, fee schedules, team bios. Every proposal, on a deadline. The 'library' lives in folders nobody curates and in principals' heads. Every proposal rebuilds the same wheel.

The AI solution

AI search across the firm-wide archive: past proposals, specs, project metadata in Vantagepoint, images in OpenAsset if you have it. Surfaces the right precedent fast. The archive compounds as people add to it. No library to maintain.

Buy, build, or wait

Build. Synthesis by Knowledge Architecture is the category-defining product, targeted at 70+ staff firms. Below that size, a custom build with hooks into Vantagepoint, SharePoint, and OpenAsset is faster, cheaper, and a better fit.

Time back

~5-15 hrs / week marketing + senior time

Section 2

Three architecture-specific opportunities

Workflows unique to architecture practice. Each is a candidate for a custom AI build that meaningfully changes the shape of the day. Each diagram below shows the workflow today vs with an AI agent in the loop.

2.1

Public RFP response

Public sector architecture work runs through Qualifications-Based Selection: a Request for Qualifications goes out, the agency shortlists 3-5 firms, only those firms get to submit a full RFP. The diagram below shows the post-shortlist flow. Marketing pulls past project sheets, a project architect writes the team narrative, and a principal writes the design vision. Marketing assembles. The last week is always a print-and-proof scramble. Two senior staff quietly use ChatGPT for first drafts.

Today

Senior hours: 30-60 per RFP
Sprint length: 2 weeks per RFP
Fire drill: Last week always slips

With an AI agent in the loop

Senior hours: 5-10 per RFP
Sprint length: Steady pace
What the agent does: Drafts narrative from past projects

The agent does the heavy drafting using your past projects and design language. The principal and marketing edit instead of writing from scratch. Replaces the off-the-books ChatGPT use with something firm-sanctioned.

2.2

Construction admin routing

When something on the job site does not match the drawings, the contractor sends a question to the firm. A wall conflicts with a pipe. A part does not fit. The lead architect reads every one, decides which engineer should answer, chases the response, and sends the final reply.

Today

Turnaround: 5+ days
Lead architect time: ~10 hrs / week
Bottleneck: One person reads + routes everything

With an AI agent in the loop

Turnaround: 1-2 days
Lead architect time: ~1 hr / week
What the agent does: Sorts, routes, drafts the answer

The agent does the reading, sorting, and drafting in seconds. The lead architect stays in the loop as the judgment checkpoint, not the queue.

2.3

Project knowledge retrieval

An architect needs a precedent: a wall section from a past project, a detail, a similar problem. They ask in Teams chat, search SharePoint, or interrupt a principal who has it memorized. SharePoint search is bad. The library initiative two years ago stalled. Partly because curation is unfunded labor, but the deeper reason is that architecture knowledge is tacit: knowing-in-action that resists being captured in a wiki. This happens 5-15 times a day across the firm.

Today

Time lost: 1-3 hrs / week / person
Firm-wide: 50-150 staff hrs / week
Library: Stalled, nobody curates

With an AI agent in the loop

Time per query: Under a minute
Library to maintain: None
What the agent does: Searches the archive in plain English

Your archive becomes a discoverable asset instead of tribal memory. Your house vocabulary surfaces in search: strategic daylighting, meaningful materials, gathering spaces. No curation overhead.

Section 3

What to avoid

Three pitches vendors and consultants push as the responsible AI default: one model for everything, Zapier as the AI brain, Notion AI as the firm's knowledge base. None of them hold up at this size.

3.1

Locking the whole firm into one AI vendor: OpenAI, Claude, or Gemini

All three frontier vendors will pitch you on standardizing the firm on their model. IT shops and consultants will tell you to just pick one and move on. Do not. AI is a portfolio of specialized tools, not a single tool. A kitchen does not run on one knife. A contractor does not show up with one wrench. Different models are specialists at different jobs right now: Claude tends to lead on code generation and long-document reasoning; GPT leads on broad chat and structured output; Gemini leads on multimodal work like images, video, and very long documents. Locking the firm to one means using a worse tool for every job outside its strength. And the leaderboards shift every few months on top of that.

AI gateways like OpenRouter and Vercel AI Gateway route per task with no commitment overhead. Your builder picks the right model for each job and swaps when a better specialist emerges. The volume discount the vendor dangles almost never matters at SMB usage levels. You do not need to pick. You should not.

3.2

Critical AI workflows on Zapier-style tools: Zapier, n8n, Make, Workato

The pitch: build your AI workflows in Zapier, n8n, Make, or Workato. Visual, fast, no engineers required. The trap is that no-code platforms ditch every engineering practice that keeps software reliable. No version control. No automated tests. No staging environment. No alerts when something breaks. When the 50-step Zap fails at 9am Monday and the person who built it is on PTO, the firm finds out from a customer complaint.

"No code" does not remove the complexity. It moves it into a drag-and-drop UI, where 50 conditional branches are harder to read and harder to maintain than 50 lines of code. Use these platforms for two-step linear automations in low-risk workflows that are okay to fail. Never as the brain.

3.3

Single-tool AI as your firm knowledge answer: Notion AI, Slack AI, custom GPTs

Each tool indexes ONE source. Notion AI indexes Notion. Slack AI indexes Slack. A custom GPT indexes whatever PDFs you upload to it. But your firm's knowledge lives across 5+ systems: your CRM, your accounting tool, your file server, email, your industry-specific app, Slack history. Think of single-tool AI as a spotlight in a dark room. It illuminates one corner clearly; the rest stays dark.

The illusion of completeness is the dangerous part. When you forget the rest of the room exists, you act on partial information. A junior staffer asks Notion AI for "our policy on rush fees," gets a confident wrong answer because the real policy was in a Salesforce custom field, and ships the wrong reply to a $200K customer. The bespoke build in capability 4 exists exactly because of this gap.

Where to start

Pick one project. Ship it before the next deadline crunch.

All three candidates sit on the build side of the buy-build-wait line. Timing depends on your project mix: build during schematic or CD phases when the team can absorb new tools, not mid-CA when the lead architect is buried.

Pick the one that gives back the most expensive hour in the firm. The other two reward the foundation being in place.

Three candidates

Meeting capture
Your principals are in 4-8 hours of design reviews, client meetings, pin-ups, and construction calls a day during active phases. Recap labor falls on the partner who took the call.
CA routing
Your lead architect is the queue for every RFI from every active project. Turnaround is 5+ days. The triage is a bottleneck, not a service.
Project knowledge retrieval
Associates ask "did we do this before?" multiple times a day. The library initiative two years ago stalled.

Next step

If you want help getting a project going, we can shortcut from either end. The free diagnostic walks your firm in 10 minutes and ranks the top three. The 30-minute call is direct: tell us about your firm and we tell you what we would actually build.

Free diagnostic

Tailored to your firm. 10 minutes.

Run the diagnostic

Discovery call

Work it through together.

Written by

K. Iwasaki

is a co-founder at ak-labz, where he builds AI software for operators of 10-100 staff businesses.