CG Common Ground | 301 W Osborn
What we did and what it producedActive

The work, decision by decision

The work, in the order it happened

1. Zoning by right, before anything else. Known: the C-1 H-R zoning and the corner-lot condition. Unknown: whether the program needed a rezone or a variance, which adds months and a public hearing to the front of the schedule. I asked the narrow question, does this exact program qualify by right, rather than the usual one, which rezone do we file. The city's zoning confirmation came in April 2025 and a written plan-check recap in May 2025: multifamily by right, no rezone, no variance, no hearing. The largest schedule risk on the project was gone before any design work started.

2. The classification, then the program. With zoning settled I classified the scope, the Level 3 alteration and the B-to-R-2 change, and only then locked the program: six residences on each of the two lower floors and five on each of the two upper, from 871 to 2,101 SF each, over a 51-stall garage. The count moved through several iterations before it settled. The classification did not move.

3. Qualifying for self-certification, with the appeal filed early. The city's site-plan comments in October 2025 said the site plan could not be approved without a fire marshal appeal on access. Rather than let that surface at certification, we filed the fire-access appeal in December 2025, months ahead of the packet. The packet itself, 83 pages, was assembled with fill, page and assembly scripts rather than by hand, which is what let LÏEF produce it in days instead of weeks. The city's review window is the program's, not ours; the scripts changed how fast we could produce the packet, not how fast the city reviews it. Attachment B lists what the packet contains.

Self-certification in plain words

Phoenix lets a registered architect or engineer of record certify eligible plans directly. The permit issues within a few business days and the city performs no ordinary plan review first. At 25,000 SF and above the project draws an automatic audit, so on a project this size the file is checked after the fact, and any error surfaces at audit, inspection or occupancy, which is the most expensive place to find one. Two consequences follow. The packet has to be right before it reaches the stamp, not fixed in review. And any question that could stall the certification, an access appeal for instance, is worth settling early, because a late appeal stalls the one step the whole path depends on.

4. Modeling the building in house, with more checking than a normal schedule allows. Rather than hand pre-development to an outside firm, we built the Revit model on a dedicated workstation: fourteen specialist skills, for the shell, structure, MEP, envelope and QA, driving a 138-tool connection into Revit, and a model that reached 12,684 elements. The bet was small and capped: a modest direct spend to find out whether agentic tools could carry drafting, against a line item the project could not move until it was solved. If nothing worked, the money was gone and the project had not moved; if it worked, I would walk into the architect with working agentic draftspeople and a different conversation. It returned roughly 40 percent of the architectural work. The larger prize was the conversation. The offer to the architect of record was not "do it cheaper" but "be at the front of this, at a rate that reflects a smaller scope." That takes margin off the architect, since drafting is close to pure profit for a firm, and making it attractive rather than adversarial is the only version that leaves a working relationship behind.

Read the differential as roughly that much less than the outside quote: the team's own hours on the model and the review pipeline sit between the direct spend and the differential and are not itemized. Source: outside proposals on file; the direct spend per the principal, September 2026.
Read the differential as roughly that much less than the outside quote: the team's own hours on the model and the review pipeline sit between the direct spend and the differential and are not itemized. Source: outside proposals on file; the direct spend per the principal, September 2026.

5. Rebuilding the QA function we had cut, not just the drawings. An outside firm's fee buys drawings and a second party whose process catches errors before a set goes to permit. Cut the fee without replacing that function and you have not lowered the risk; you have deferred it to whoever inspects the stamped set. So the question was what that QA process actually checks for, so the same checks could be rebuilt rather than approximated.

The review pipeline, in four steps

  1. Export the model's geometry, its walls, dimensions and room boundaries, to a structured format.
  2. Have a set of checking agents compare each element against two references: the governing code sections for this occupancy and alteration type, reduced once by a person into fixed, checkable parameters, and the project's own design-criteria sheet of unit counts, ceiling heights and setback minimums.
  3. Write any element outside a stated tolerance from either reference to a findings list.
  4. A person clears every item on that list before the set moves forward.

The code half is the hard part. A code section is prose with conditional logic; exit separation is a fraction of the diagonal that changes with sprinkler status, and an elevator cab minimum is stated differently across tables. Each applicable section had to be read once and turned into a number before anything could be checked against it. What the agents add is consistency: the same parameters applied to every element in the export instead of a person re-checking each by eye. The pipeline's tolerance check flagged two errors, a wrong building setback among them, both ahead of the point a person doing an unassisted review would have had to catch them. An earlier version of the model turned out to carry geometry from an unrelated building's set, pulled in during an export, and that version was ruled out before it shipped. I credit that catch to the discipline of looking rather than to the pipeline checking itself: it was an uncontrolled single event, not a designed self-test with a known-bad input.

6. Who checks the checker. The architect of record was certifying his own work under a compressed timeline, which is a single-reviewer process however qualified the reviewer. A project meeting in July 2026 established that no complete code study had ever actually been finished for the building, and that the base survey geometry conflicted across three sources. We produced an independent, owner-side code study the next day: 89 pages, every finding cited to a code section and a sheet, and the critical findings handed to the architect pre-cleared instead of asking for a blind re-read of the building. Four of them: the plan set's title block carried the address of a different parcel; the separation between the two exit stairs looked short of the code distance; the construction type was stated three inconsistent ways across the record; and an elevator cab depth fell short of the code minimum for an ambulance stretcher, flagged priority one.

Three layers, three different kinds of catch: geometry against the criteria, a person against the export, and an independent read of the existing building against the code. Source: the findings list and the code study.
Three layers, three different kinds of catch: geometry against the criteria, a person against the export, and an independent read of the existing building against the code. Source: the findings list and the code study.

7. The handoff, turned into a template. A 182-file, nine-section packet went to the architect of record in June 2026, with a working session the next day. It became a reusable ten-item template for the next handoff. The template travels; the file count does not.

8. The roof. The building carried one carrier's rooftop telecom lease. The buyer of that lease opened with an 85-year easement across the entire roof. That is a common move: whoever drafts first tries to buy more roof than they are renting, because roof rights are cheap to ask for and expensive to get back. Say yes and you have handed away every future rooftop use on the building, the mechanical equipment the conversion needs, amenity space, a second carrier, solar, for one payment today. I went back to what the carrier actually had and needed, the physical footprint of its existing equipment, cut the grant to that, and held it there through the back-and-forth on term and roof language. The deal closed in April 2026 on the narrower scope, with the proceeds paid at closing rather than deferred into a long rent stream and the rest of the roof retained. Outside counsel papered the easement assignment and the lender's non-disturbance agreement. The question underneath was whether one tenant's lease gets to claim every future use of the roof at today's price, and the answer was no.

Roof zoneBuyer's opening draftWhat closed
The carrier's existing equipment footprintIncludedConveyed as the easement
Mechanical and equipment zone for the conversionIncludedRetained
Amenity spaceIncludedRetained
Remaining roof, for a future carrier or solarIncludedRetained
PaymentA rent stream over the termPaid at closing

Attachment C draws the roof plan and the negotiation sequence.

9. The data room, re-audited every time the project moved. The investor data room went live in June 2026 with every link checked by hand, one at a time, the same day it was built. A completeness audit in July, against five source locations, found the room held a fraction of the relevant record and staged the rest, including the substance of a city comment letter that survived only in the civil engineer's own response memo. The audit also corrected three claims that had drifted from the record: there was never a second cell tower, the fire appeal had already been filed, and the building was already sprinklered under that appeal. A data room built once is a snapshot; this one reruns its audit each time the project moves.

Two audits a month apart: the launch check of every link, then the completeness sweep that found what the room was missing and staged it. Source: the launch audit and the completeness audit.
Two audits a month apart: the launch check of every link, then the completeness sweep that found what the room was missing and staged it. Source: the launch audit and the completeness audit.

What it produced

The program that came out of the sequence is a set of residences whose living area, footprint and gross building area are three different measures, and none is ever presented as another. The gross figure includes the subterranean garage; the footprint is the four residential floor plates; the living area is the conditioned space inside the residences. Outdoor patios sit outside the floor plate and are listed by residence in Attachment A.

The residential floors' efficiency is 89.5 percent: the recessed patios, 1,783 SF, plus the core and circulation, 4,449 SF by subtraction, are what sit between the residences and the floor plate. Source: the unit schedule; the architect's total-project figure.
The residential floors' efficiency is 89.5 percent: the recessed patios, 1,783 SF, plus the core and circulation, 4,449 SF by subtraction, are what sit between the residences and the floor plate. Source: the unit schedule; the architect's total-project figure.

What each part of the sequence closed: zoning by right with no rezone, variance or hearing; a classification locked ahead of the program; a self-certification packet the architect of record could certify; a model with two errors caught by its own tolerance check before a person had to find them; an independent code study handed over pre-cleared; a roof grant scoped to the footprint with the rest of the roof retained; and a data room that holds up on the hundredth click, not the first.

The building as a buyer or a partner meets it is on its own site, liefmidtownphx.com, and in the paper behind it. Attachment H is the compiled plan set, and Attachment I is the selections and materials book the site shows on its Look Book page. Attachment E is the brand it was all built in, and Attachment F is the first design direction, room by room. Attachment J is the December 2024 price list, the projections the homes would be listed against, drawn up before the program settled at the twenty-two homes.

The liefmidtownphx.com home page, captured whole and cut into five columns to read left to right: the building and the vision at the top, then the residences level by level with a card for each home, then the finishes every home includes, then a gallery that sets the bare concrete floors beside design studies of the finished rooms, and the contact form. Source: liefmidtownphx.com, captured 25 September 2026.
The liefmidtownphx.com home page, captured whole and cut into five columns to read left to right: the building and the vision at the top, then the residences level by level with a card for each home, then the finishes every home includes, then a gallery that sets the bare concrete floors beside design studies of the finished rooms, and the contact form. Source: liefmidtownphx.com, captured 25 September 2026.

What we kept, replaced and installed

Kept. The architect of record, for the stamp and the certification; the city's own self-certification program; Kanopi for the estimate; outside counsel on the easement and on the contracts. The stamp is what a license attaches to, and it stays valuable. Nothing that carries liability was cut.

Replaced. The outside pre-development scope, drawings plus the firm's own QA, with an in-house Revit model plus the review pipeline. The industry default, a full-service firm and the standard queue, was not wrong. It was slow and priced as one deliverable, and this deal could carry neither the fee nor the queue. The QA function was rebuilt on purpose, not skipped.

Installed. The classification-first sequence; the review pipeline; the link-by-link data-room audit and the re-audit rule; the handoff template. Each replaces a check the outside firm's process or the city's plan review would otherwise have performed with one the owner's team runs itself.

The fit for this method is narrow by design: an owner-side team willing to carry entitlement risk directly instead of routing it through a full-service firm, on an existing shell with by-right zoning. That posture is rarer than it sounds, which is the reason it is worth owning rather than listing as a caveat.

What it cost to hold the line, and what I would watch

You have to build the tool before you can use it, and that time lands before any of the benefit does. You take margin off a counterparty who was counting on it. The position only exists while the capability is ahead of the market, and on this one that will not be long. The pipeline has run one full cycle on one project; what transfers to the next model, the criteria-sheet format, the parameterized code set, the export script, and what has to be rebuilt each time is not yet documented, and until it is, this is a result with a replicable method rather than a proven portable asset.

Outside counsel also earned its place on the contracts. An architect's proposal carried a perpetual 2.5 percent match payment on investor distributions, buried in a profit-sharing note; it was caught before signature and that contract was never signed. Read every architect contract before it is executed.

The constraint that turned out to bind was never the drawings or the code. It was capital. The construction capital is being raised, and the entitlement and design work is complete and waiting on it. Every step above drove design and entitlement risk toward zero, and none of them could touch that variable. The lesson I took for the next project: run the raise beside the entitlement rather than behind it, or hold major entitlement spend to a lighter path until the raise clears, so a capital delay does not strand a finished design against an uncertain closing date.

What I would watch from here: the automatic audit that follows a self-certified permit, since that is where a packet error surfaces; the construction-type determination, which the code study found stated inconsistently across the record and which drives fire rating and allowable area downstream; and the fire-access and sprinkler path the appeal settled, because those are the items that reopen if the plan set changes.

Questions answered

Rooftop Telecom Lease Negotiation

Where did the buyer push hardest in the term or easement language, and how did you keep the negotiation anchored to what the carrier actually used?

They pushed hardest on the easement language. They wanted the entire roof. We couldn't give that away, because the building still has to function. Every mechanical system needs penetrations through that roof. They only cared about their cell tower. If the building had sat vacant forever, it wouldn't have mattered to them. So we kept every conversation tied to the existing cell lease area, the footprint they actually used, and we cut the scope back to the language we needed to make the build work.

Answered 2026-09-29.

What specific plans, records, or expertise did you rely on to define the existing cell-lease area and make the narrower easement workable in the final documents?

The existing lease was the anchor. The carrier already had a cell lease area on that roof, so we didn't have to invent a boundary. We just refused to let the easement grow past it. On our side, the test was what the conversion needs from the roof: the mechanical zone and every penetration for our systems, the amenity space, and room for a future carrier or solar. Anything outside the carrier's existing area stayed with the building. Outside counsel papered the easement assignment and the lender's non-disturbance agreement. My part was holding the scope. Theirs was making it hold in the documents.

Answered 2026-09-29.

How did keeping the roof outside the carrier’s existing footprint protect that committed conversion program, what specific design or entitlement risk would the whole-roof easement have created?

The building still has to function. Every mechanical system in the conversion needs penetrations through that roof, and a whole-roof easement would have handed that away, along with the amenity space and room for a second carrier or solar, all for one payment. The carrier only cared about its cell tower. If the building had sat vacant forever, it wouldn't have mattered to them. So we kept the grant to the existing cell lease area, the footprint they actually used, and held the rest of the roof for the build.

Answered 2026-09-29.

Which specific roof-dependent element of the condominium plan made preserving the non-carrier area most consequential?

The mechanical systems. Every one of them needs penetrations through that roof, and the building has to function before anything else. That's why we couldn't give the roof away.

Answered 2026-09-29.

Agentic AI Production System for a Development Firm: 65% Cut in Production Time

How did that validation discipline carry into the six-week Revit production build, particularly the three-agent adversarial review, and what did the wrong-setback catch prevent before construction documents were issued?

Same rule on both sides: nothing grades its own homework. On the estimating side we held one bid out at a time, priced it as if unseen and published only that number, never the flattering in-sample one. On the architecture side the agents drafted a stalled as-built into a design-development set over six weeks, and I reviewed and signed off on the model before it moved forward. A three-agent adversarial review caught a wrong setback before any person did and before it reached construction documents, the kind of mistake that's cheap on a screen and expensive after concrete. The same model was later found to carry a footprint problem traced to another project's plan set, which is why the overlay rule and the blind pass exist, and why no self-caught error is treated as the last one.

Answered 2026-09-29.

Walk me through the human release gate you used before the design-development set moved forward, what did you personally verify, and what would have stopped the model from being issued?

The gate was a findings list, and I cleared every item on it before the set moved. The model's walls, dimensions and room boundaries were exported, and checking agents compared each element against two references: the code sections for this occupancy and alteration type, reduced once by a person into fixed numbers, and the project's own criteria sheet of unit counts, ceiling heights and setback minimums. Anything outside tolerance went on the list, and an open item meant the set didn't move. The stamp still belonged to the architect of record, so we also handed him an independent 89-page code study with the critical findings pre-cleared, instead of asking him for a blind re-read of the building.

Answered 2026-09-29.

We have the wrong-setback example already. What was the second material design error the review system caught?

Plenty: wall height, deck width, floor slab thickness, cell tower placement. It was less about what it caught than about doing the Revit drafting ourselves instead of paying a drafter, which is a big cost and time savings. A human drafter can't work 24 hours a day.

Answered 2026-09-30.

What concrete baseline did you use to measure the cost and schedule savings from bringing the drafting in-house?

Actual architectural bids, about five different ones. Line by line, we were able to cut out a great portion of those bids based on the work we did agentically.

Answered 2026-10-03.

In-House Multi-Trade Labor Force: Recruiting, Training, and Deploying 200+ Tradespeople

What did the A/B/C grading actually measure, and how did you determine that a crew was ready for more demanding work?

The grade was per trade, not per person. Most tradespeople are an A at one thing and a B or C at others, and it comes from how they learned. A framer might have started as a drywaller or a painter before framing became his trade. So we graded each guy's primary skill and his second and third trades, and that told us where else he could work.

Answered 2026-09-29.

Walk me through one project where that trade-by-trade profile allowed you to staff or rebalance the work differently, and what concrete impact it had on the project.

Midtown. My other managing partner is John Armstrong of Armstrong Construction Group, and Armstrong brings its own in-house labor force, the same crews behind the Ritz-Carlton Residences in Scottsdale. That's why LÏEF works: the labor is in-house through Armstrong.

Answered 2026-09-29.

Lïef Midtown Adaptive Reuse

What decisions did you personally make to turn those three influences into a cohesive Japandi-Sonoran residential concept, and what did that work unlock for the project?

I chose it because it serves the market and the luxury buyer, and there's nothing like it in Phoenix that I know of. It's close to downtown, which fits that buyer too. The look is what we wanted: natural materials, simple lines, no clutter.

Answered 2026-09-29.

What evidence did you use to validate that this positioning would support the project’s pricing and buyer demand, such as comparable properties, broker input, buyer feedback, or another form of market analysis?

The positioning wasn't only my read. The architect, the interior design team and our broker all confirmed it, and the broker has done as much adaptive-reuse and condo work as anyone in Phoenix.

Answered 2026-09-29.

The uploaded project materials also reference an 89-page independent code study with 18 critical findings. Which finding or code issue most changed the design, budget, schedule, or permitting strategy: and how did you resolve it?

The pool location and what we could do on the roof. We decided to wait and resubmit the pool later in the process, after approval of the main shell. On the roof, we decided to forgo a roof deck. It was too complex, and the FCC radiation requirements made it too difficult.

Answered 2026-10-03.

The result, in short

Zoning came back by right, with no rezone, variance or hearing. The classification locked ahead of the unit program, and the self-certification packet was assembled for the architect of record to certify. The in-house model returned two errors caught by its own tolerance check before a person had to find them, and an independent code study was handed over pre-cleared. The roof grant closed scoped to the carrier's existing footprint, with the rest of the roof retained. The entitlement and design work is complete; construction capital is being raised.

A slice of the project list

A few related projects.