Use more of a system the firm already owns.
AI operations and implementation for architecture firms
Turn scattered AI experiments into one working system.
Clearworks helps architecture and design firms define the work, choose one bounded improvement, build or connect it, preserve human judgment, and run it long enough to learn whether it actually helps.
Los Angeles basedAIA LA TAP co-founder25 years in technology and operations
StartReal work
BuildWorking layer
KeepHuman judgment
The adoption gap
Using AI is not the same as changing the work.
AIA’s August 2026 practice data shows a profession moving quickly into AI without the same maturity in training or measurable results. That is the gap Clearworks is built to help close.
48%of responding firm leaders said their firms were actively engaging with AI.
7%reported having a formal AI training program.
22%of firms using AI reported a substantial productivity increase so far.
Source: AIA/Deltek Architecture Billings Index, August 2026. These are survey results, not claims about every firm.
The operating loop
One job. One reviewer. One accepted result.
The useful first system is small enough to inspect and important enough to matter. It follows the real work from trigger to evidence—not a demo floating beside the firm.
- TriggerThe event that starts the work.
- SourceThe files, systems, and facts allowed to inform it.
- AssistanceWhat AI or automation retrieves, structures, compares, or drafts.
- Human gateThe named person who checks and decides.
- Accepted recordWhere the approved result becomes usable state.
- MeasureWhat tells the team to keep, revise, or stop.
Review loopIf the reviewer rejects, corrects, or asks for missing evidence, the work returns to assistance or its sources and runs through the gate again.
Nothing moves past the human gate merely because a model produced it.
Before selecting a product
The right answer is not always “buy AI.”
We define the job first, then choose the least complicated path that can responsibly improve it.
Carry information between existing tools without rebuilding everything.
Choose a focused product when it already fits the work and ownership model.
Create a small working layer when the firm’s process is the differentiator.
Wait when the source, owner, or review rule is not ready.
Keep human work human when automation would add risk or erase judgment.
Evidence, with labels
We separate what was observed from what might be possible.
No invented savings. No universal accuracy claims. No pretending a prototype is already a production system.
Studio PCH found the operating friction before choosing the fix.
Observed30% rework
Observed20 weekly hours
Identified$150–200K
StatusOpportunity
The audit documented rework and manual finance effort, then identified a potential annual opportunity. “Identified” is not the same as realized savings.
Read the case evidence →Useful first-workflow shapes
Start where information keeps being rebuilt.
These are example workflow shapes, not claims that every firm has the same problem or that every example should be automated.
Project and reporting visibility
Assemble current state from approved project records, then route the draft to the accountable reviewer.
Meeting and decision follow-through
Carry decisions, owners, and unresolved questions into the system where the team actually works.
Accepted precedent retrieval
Find approved standards and examples without treating every old file as current truth.
Proposal evidence retrieval
Find relevant project experience and approved language while preserving final authorship.
Finance and staffing visibility
Reduce reconstruction by joining the right inputs into a reviewable operating view.
First-pass comparison
Surface differences or missing evidence for a qualified person to evaluate—not an autonomous approval.
Three ways to begin
Match the engagement to how much you already know.
The goal is not to sell the largest engagement. It is to create enough clarity and evidence to make the next decision responsibly.
Blueprint
Define one recurring job, its source records, human review point, accepted destination, and the first fix worth testing.
Full Busywork Audit
Map how work moves across the firm, identify the highest-value and most buildable opportunities, and leave with concrete next steps—not a tool list.
Implementation Partnership
Build, connect, train, govern, and improve the working systems after priorities and ownership are clear.
The working boundary
Clearworks builds the working layer. Your firm keeps the decisions.
We make ownership explicit before launch so assistance does not quietly become authority.
Before a workflow goes live
- name the source records and permitted access;
- name the owner and human reviewer;
- define the accepted destination and exception path;
- confirm provider, retention, security, and specialist requirements;
- agree on the measure and review date.
The firm retains
Professional judgment, design decisions, authorship, verification, approval, sealing, contractual decisions, accounting, legal obligations, and business risk. Managed IT, security operations, legal counsel, and compliance remain with the specialists who own them.
Operating boundary updated September 28, 2026.
Questions before the first conversation
You do not need to know the solution yet.
Josh Weiss · Founder, ClearworksTwenty-five years in technology and operations. Founder of LA Creative Technologies and co-founder of AIA LA TAP.
Where should an architecture firm start with AI?
Start with one recurring job, not a shopping list. Map the sources, handoffs, decision, reviewer, accepted record, and measure before choosing whether to configure, buy, connect, build, defer, or leave the work alone.
Do we need an AI strategy first?
Not a long one. A bounded workflow creates the evidence a useful strategy needs: what the firm is trying to improve, what information is involved, where judgment stays human, and what the team can actually own.
Do we need to replace our current software?
Usually not. Clearworks starts with the systems and records already in use. The right decision may be to configure what you have, connect two systems, buy a focused product, build a small working layer, defer the work, or leave it alone.
Who reviews the AI output?
A named person with the right professional or business authority. AI can retrieve, structure, compare, or draft. Your firm retains authorship, verification, approval, sealing, contractual decisions, and professional judgment.
What happens after the first workflow?
The team reviews the evidence. Keep and improve what works, revise what is useful but incomplete, or stop what does not earn its place. A successful first workflow can become the beginning of an Implementation Partnership.
Bring one recurring job.
No AI shopping list required. Show us the report, handoff, decision, or task that keeps coming back. We will help you define the work and decide what deserves to happen next.
Bring one workflow →