All hail the tech adjacent – meaningful AI innovation outside of big tech

The really cool uses of AI I’m seeing aren’t coming from big tech who make the models, but from websites and software makers in specific verticals who are using those models to do amazing things. I call these as “tech adjacent,” because they are developers but lack expertise in data science. Amazingly they are creating real value while “high tech” is struggling to figure out why a consumer might want to pay for AI search when Google is free.

The hallmark of these innovators is that they start with their deep knowledge of a specific industry or a customer and use generally available tech to solve a problem. They work within industries vs. being hell bent on disrupting them. They are creative about business models.

This is going on in many industries, but I recently had chance to get up close with AI innovation in home improvement. I want to share what I've seen because I believe it's a road map for many other industries.

In September I had the chance to talk about AI at the HIRI Summit (Home Improvement Research Institute) here and here.) To prep I dug into what’s public, but at the conference I learned of lots more cool things under development. I’m a serial remodeler so I think home improvement is just inherently interesting, but I think the lessons are applicable in many industries.

Three takeaways:

·?????? Use insider knowledge to identify a REAL problem.

·?????? AI expertise not required. The models and hardware are generally available.

·?????? Complement vs Disrupt: Support the incumbents when they have the relationships with the end buyers.

Identify a REAL problem.

HIRI’s research has found that homeowner indecisions is the #1 problem for small contractors, bigger even getting paid quickly. Personally I know that it can be hard to commit to a big spend when you can’t picture the finished project. Commercial architects use AutoCad and special cameras to create virtual renderings and fly throughs to help their customers see the finished product. That is beyond the reach of small contractors because it’s expensive, complex and time consuming. The essence of a real problem to solve.

AI expertise not required

I know of at least three companies actively working on solutions: BuildXAct, Houzz Pro, and Hover. No coincidence they all have deep expertise and relationships in the industry, but none have expertise in AI.

The basic solution is to use a smartphone app to scan the tops and bottoms of the walls in a space and have the software transform it into a 3D model. I’ve tried the Houzz Pro app and it’s freaky good. If nothing else, it saves time and errors from hand measuring and transferring dimensions into design software. The renderings are easy to edit. On the Houzz Pro version, the user toggles between keeping and removing existing cabinets and furniture. Some systems have libraries of branded products and finishes for realistic finished projects. The prospective customer can walk through the space on any PC or phone, even seeing how the light will be at different times of day. ?

The AI models that power this are available to any developer via API or SDK. I’m skeptical of those who say you can build applications without code, but bullish on developers with no expertise in data science but a clear understanding of user need.

The systems work best with lidar, cameras that bounce light off of objects to measure the distance of things in the image. Real estate agents contract for lidar images to drive virtual walkthroughs on Zillow. Currently the only consumer smartphones with lidar are the iPhone Pros, the ones with the distinctive 3 camera lenses. Up until this point I have thought of those cameras as tech in search of an application; but from now on I’ll think of them as hardware available for innovation. I’m betting there are tons more applications in other verticals. ?

Note that Lowe’s is testing something similar, but without automated measurements and using Apple Vision Pro for a fully immersive walk through. ?I think the smartphone systems will do better in market because 1) they do the measuring and 2) I hear many customers are reluctant to don the headsets. But the fun part of a new field is that we will all get to see how things play out.

Complement vs disrupt

The tech start-up ethos is to disrupt markets, but in this case I think the smart money is looking to complement contractors. After all, they have the relationships with the end customers and it’ll be a long time until 3D printing can remodel my kitchen.

Looking at this I’m reminded of work I did for the Bill & Melinda Gates’ Foundation on K-12 personalized teaching tools. Those innovators treated teachers as some combination of obstacle and irrelevant. Not surprisingly, none of them have had much success getting into mainstream classrooms. Teachers are gatekeepers. So are contractors. Working with the gatekeeper is usually smarter than trying to blow up the gate.

Case in point, working through established relationships means these developers have access to contractors and allied companies to refine the offer and get trial. ?Since these systems can bring branded products (eg: cabinets, countertops, appliances, windows, furniture) into the plans these developers also have options for revenue streams beyond the contractors themselves.

TL/DR: The tech adjacent - developers with no AI expertise - are creating valuable applications because they are thinking deeply about their customers’ problems. The models and the hardware to realize their visions are largely available. I predict a fascinating next few years, with amazing stuff coming from firms you’ve likely never heard of.