Here are two "what-keeps-you-up-at-night?" flashpoints builders contend with at 2023's midway point, even as sales exceed budget and demand conditions look solid:

• Capital finance runway in a Hobson's Choice lending and investment environment that's both more restrictive and more costly.
• Margin and margin-of-error erosion, where high-20s percent gross margins are doing battle as shock absorbers funding pace, leaving little or no wiggle room as the battle continues through the back half of 2023 and into 2024.

One good friend – a residential development executive strategist – characterizes the mid-year squeeze this way:

There is more frustration over underwriting new deals: Net price down, absorption pace stable, production cost – other than lumber -- relatively unchanged. And land not down – or even up.  Does not add up to much margin. Common refrain is to buy land, you have to pick the tallest midget in the room. Very little shows interesting margins."

Of course, this applies especially to many hundreds of privately capitalized operators whose firms complete and deliver 10 or more homes per year. They're in a separate universe altogether from that of the nation's 20-or-so enterprises capitalized with publicly funded equity and debt. Together, they account for nine of every 10 ground-up single-family homes in America.

What many principal owners, leaders, and strategists fret over – the ones tossing and turning at night even though demand up to now eclipses what they'd expected, and even though scarcity among listed-for-sale existing home may not let up anytime soon – is this: An indefinitely elongated period ahead of compressed margins, almost non-existent margin for error, and a drawn out battle for access to operating and land acquisition capital.

In this context – squeezed margins and negligible margins for error in homebuilders' execution – we spoke a few days ago with Salman Ahmad, co-founder and CEO of Mosaic, a supporting partner of The Builder's Daily. Our conversation explores the still evolving and largely untapped opportunity area among homebuilding operators and their construction and distribution partners to optimize the build cycle.

From pre-development and pre-construction through the start-to-completion construction cycle, it's a business of hand-offs, linkages, dependencies, reliabilities, probabilities, etc. They add up to what experts continue to assert could mean as much as 20% cost opportunity in each new house, running at 450 SKUs, 20,000 separate pieces, 700 or so lines of code, 25 or so crews for every 2,100 square feet.

We asked Salman for a high-level take on Mosaic's progress as a six-year-old technology and data-fueled full-stack general contracting platform, powering 150 team members across more than 100 new-home communities and more than 20,000 build-to-rent units.

Specifically, we wanted to learn where computational power and machine learning might figure ever more profoundly in home construction operations where repeatability and feedback loops allow for learning and improvement.

Here's a selection of Salman Ahmad's responses to questions and challenges we explored.

Q: How does Mosaic's role as a GC working in so many geographies, with so many projects, and teams, position the company to build value by optimizing the process?

Salman Ahmad:

The best way for Mosaic to provide the best services to our customers is through data. Having access to data allows us not only to improve, but also to detect things that can be hugely beneficial to our customers. The best way to have that data, get access to it, and learn from it is by being positioned at the cross section of the entire stack. You can see the purchasing. You can see the the field execution. You can see the contracting side of things as well. Being the GC is at the cross-section. That platform that allows you to get access to that data. That's why it's been valuable for us to be in the position that we are at."

This is nothing unique, by the way, for Mosaic in the construction industry. You see this in most segments of the economy right now. The most forward players in the realm in various industries are ones that are looking and using data. Data is the hottest commodity in the tech world today. It's also the case that A.I. allows us to deliver huge values in many other industries and construction is no exception in that regard."

Q: Will applied A.I. solutions to improve the build cycle involve adding "juice," or forward-thrust improvement? Or will those solutions focus more on removing friction from current processes and systems?

Salman Ahmad

I think the biggest potentials we have are on the reducing friction side of things. That's only if I have to force myself into a binary choice. Obviously, it's on a continuum. It's either lowering barriers or enabling things to be done that were not done before -- not because they were impossible, but because they were too time consuming. Now you have the ability to do them a lot faster."

A way I look at what we have now relative to before – that changing condition in the A.I. computer science world – is that we now have computers and software that can actually communicate and talk in human language. The value of that is at a stage where it can communicate and converse on general human interest topics, general information, general knowledge.

The next step is getting domain specific data that allows the A.I.s to talk and converse about vertical specific domains, domain specific information. That's why data is so necessary for A.I.s to function within the real estate world. Once you have that, developers and general contractors will be in a position to use software and technology to do things much more quickly than they were able to do before.

They'll be able to accomplish more of that during the finite timeframe of a project. That's the useful way to think about the change that's happening in the A.I. world. We've gotten computers to talk. By reducing the friction, you can add velocity, and doing more of something allows you to learn and become better as well."

Q: How do you look at integrating data and learning into play in construction operations?

Salman Ahmad

What formats are most conducive for A.I.s to use to learn to improve operations?  That's an engineering topic that is being that's being worked on. We believe that that's going to be necessary to train A.I. models and other future upcoming systems and technical innovations.

The one thing that I think is important that we can talk about right now is that data is necessary. It's the most important piece to unlocking the A.I. innovations in the construction real estate world.

And, that data needs to probably come from general contractors and for the same reasons that I mentioned before – the general contractor is across the entire process. GCs , in my view, will play a vital role in unlocking AI efficiencies in the development and real estate construction worlds. GCs will need to operate differently as companies to access this data.

It's going to be a steep learning curve, for general contractors. The payoff for the industry and its firms will enable a lot of things, initially by  reducing friction. Then, you're going to see ways that it adds juice as well."

Q: Does the eventual unification of data – and the issues it exposes – involve stepping back to resolve how to move forward?

Salman Ahmad

We have computers that solve a whole bunch of problems, but they also create new problems, which is why you have teams of people, like IT administrators, that you didn't have before. Any technology is like that. There's always a hidden cost for new technologies. Specifically in this space, it makes me think the more complicated that these ERP systems get, the more likely a user has to hire specific talent to upkeep them and maintain them now. You're not just hiring construction folks. You are hiring system administrators,  ERP specialists, and that sort of thing.

It's a classic phenomenon. It's like a technologies create a lot of efficiencies up until a point and then just through social machinations, they clearly start to create inefficiencies. And in some cases, inefficiencies that wipe out the actual original efficiencies that they, you know, they purported to have generated.

Q: How do you sustain the commitment and investment to get to the end game here?

Salman Ahmad

Our desired goal is to make things better. Assuming you want to think make things better, you need to make an investment. This is an investment in data to get A.I. and other systems to work.

The major players that need to make those investments are GCs; they do control the entire stack. There are efficiencies we've gotten irrespective of A.I. By investing in the engineering and R&D efforts that we've done in creating these unified data models for construction projects, we've achieved tremendous efficiencies just by keeping things up to date by using static analysis. These include techniques to identify issues before they actually become problems in the field. There's a tremendous amount of efficiencies you you gain. A.I. can generate a lot of the documentation automatically so you don't have to manually do them by hand. This not only creates time savings, it avoids a lot of clerical errors that would happen in due course.

But there are hidden costs. You need to be able to think abstractly around data while also being able to know how to build stuff and integrate that overlay of theoretical computer science, data information modeling, and meanwhile, actually execute on the sticks and bricks in the field. That's a very rare combination.

That's likely going to be the challenge this industry needs to to overcome to really unlock. For our part at Mosaic, we've been making a very strategic investment for the past several years and will continue to do so. We're investing not only in the data systems, but how we operate as a general contractor, and building best-in-class teams that are able to execute in the field.