In this Post:
Collaborative Hyper-Innovation
Five Key Maxims for Micro-Scale Innovation
(For an intro, check out ‘The ‘AI First’ Architecture Firm’)
In a previous post, under the heading of Building a Firm-wide, A.I. Ecosystem, I posited that the architecture firm of the future will be a hive of micro-scale innovation2, where individual architects are using Natural Language, Generative A.I. (NLGAI) to make their work easier and faster, through the rapid, constant and ubiquitous creation of ‘Mini-AIs.’ NLGAI enables non-coders to participate in this kind of innovation for the first time. Historically, we have been dependent on software companies like Autodesk and Adobe to think up tools, routines and shortcuts and then add them to the software platforms we already use. The last few decades have allowed for a kind of ‘bounded’ innovation through visual programming tools like Dynamo and Grasshopper. But these tools still require coding knowledge to exploit their full capabilities. That’s going to change, very rapidly.
Collaborative, Hyper-Innovation
That change might look like this example where three architects at Lang, Shelley & Associates (architects Adele, Beyonce and Cher) are working on three different projects (Projects A, B, & C, naturally). In that example, the three architects all innovate independently by creating Mini-A.I.’s to execute various tasks. The firm as a whole moves ahead by facilitating an ecosystem where those innovations can be easily shared.
Pointedly, this reflects how architects already collaborate. Indeed, they have been doing so for decades. Sharing little blocks, scripts, and techniques is a common feature of firm life. The difference here is that no employee needs to be a techno-whiz to participate. And for the techno-whizzes in your firm, their job is made exponentially easier, because tools like ChatGPT and Github Copilot can write the code for them. In the past, architects may have shared little CAD Blocks for tables, chairs, etc. Going forward, they’ll share Mini-A.I.’s that everyone can have a hand in creating, just like Adele, Beyonce and Cher. Our three architects are all working on their individual projects, which include the following steps:
Here, I’ve imagined the design process as a 10,000 step process, and Steps 3,443 through 3,454 in the middle of that process, but that’s completely arbitrary. I’ve also envisioned that all three projects have the same steps, in the same order, and that’s not how design works. However, it’s a useful structure for describing how innovation will work in the age of A.I.
In an age of NLGAI, architects, working on their own, will engage in a collective, stepwise process of innovation that rapidly automates steps using Mini-A.I.’s, and then aggregates those Mini-A.I.’s into ever-larger A.I.’s
In the case of Adele, Beyonce and Cher, their innovations were shared at the weekly Monday staff meeting. In truth, I did this because it’s a work practice with which we are all familiar (the Monday staff meeting, I mean). I thought it would seem more palatable to those still trying to wrap their head around A.I. and what it will do to architectural practice (which includes me). In all likelihood, it will not happen that way.
The employees of LSA will continue to generate various Mini-A.I.’s which automate the 10,000 steps of the design process, and subsequently combine and remix those steps into ever-larger A.I.’s. It is therefore necessary to have a central repository of Mini-A.I.’s that is freely accessible and transparent to all employees at the firm, so that every employee can share in the benefits of everyone else’s Mini-A.I.’s, and leverage them both in their own work, and to make more Mini-A.I.’s.
As this repository grows larger, it seems impractical to have employees search through it every time they want to get something done. It seems equally impractical to have everyone wait until the Monday staff meeting to hear about what their colleagues have created. The more logical solution is to have an A.I. that searches the repository and automatically surfaces appropriate tools. I described a particular model for this in Developing an Autonomous A.I. Workflow, also under the post Part 3: An A.I. Firm ‘Strategy’ that uses Microsoft’s Jarvis, utilizing Chat GPT and the AI model hub at HuggingFace to create unique solutions to any specified request. It would be simple enough to train an A.I. to perform a Jarvis-like function, survey the Mini-A.I.’s available and surface those that are appropriate to the task at hand. It’s not all that radical - computers have done this for a long time, albeit badly (e.g. remember Clippy? Who would constantly patronize us by saying ‘hmmmm, it looks like you’re trying to do X. Rot in hell, Clippy). The critical thing will be to have a central repository, and minimize the friction involved in getting Mini-A.I.’s in and out.
The basics of setting up software and a repository will be relatively straightforward. The harder part will be building a cultural ecosystem around their use. Architects are pretty innovative people to begin with, we just have to organize the firm in such a way that it maximally supports this particular kind of innovation. Smart business leaders have always known that innovation doesn’t magically start when you say ‘OK, everyone innovate!’ One has to create a culture of innovation and an ecosystem in which it can flourish. There’s tomes and tomes about how to create a culture of innovation at your firm, and most of it you probably already know. Here, I’ll only offer maxims specifically around NLGAI.
Five Maxims:
Demystify NLGAI Technology.
Include Everyone.
Build an Incentive System Around Innovation.
Reconsider KPI’s.
Innovate the workflow, not just the work.
Demystify NLGAI Technology. I also wrote in Developing an Autonomous A.I. Workflow that one of the most promising dimensions of NLGAI is that it opens the door for everyone to be a coder, without learning to code. But it seems inevitable that different architects will embrace this at different speeds. Employees may not know that the NLGAI revolution, for all its dizzying complexity, has actually made certain things a lot simpler, like ‘Step3,447: Run a price check on all possible substitutions.’ Sure, you could have automated that prior to NLGAI, but it would have taken significant competency in Python or Java, and probably taken you just as long to code it as it would have to just do it the old-fashioned way. NLGAI has new competencies that were previously unavailable in any of the familiar ‘programmable’ software platforms like Dynamo and Grasshopper, like being able to have a Mini-A.I. that can automatically draft & send custom emails to individual suppliers of all possible substitute products in your BIM model, and return a filtered list of any substitutions that would constitute an added cost.
Everyone at your firm now has a tool where they can begin to automate tasks outside the bounds of any particular software platform, because of the way that NLGAI facilitates interoperability between different platforms. This will require thought-leadership at multiple levels, and a process by which everyone at the firm is introduced to, and made comfortable with, what NLGAI can do. In an ‘AI-First’ firm, architects are not only familiar with this, but fully embrace it. Their first question is always: how do we use A.I. to solve this problem?
Include Everyone. At every firm, there are architects who are more ‘techie’ than others. People who either have a greater facility with technology, or a greater enthusiasm for it, or both. It usually falls on these actors to inspire, convince, and cajole the firm into accepting any new software or technological approach. This creates all sorts of unhelpful asymmetries. Sometimes, the people that need technology the most know the least about it. Other times, someone might have a brilliant idea about how to get something done, but have no idea how to frame that potential workflow in a technological way, or even that a technology exists which could help. The promise of NLGAI is that it makes everyone a designer, not just a user, of technology. Sure, you won’t be redesigning proprietary software programs like Revit. But you can have a hand in designing the meta-technology around which the firm is built: A.I.
The greater the number of people involved in the co-creation of practice, the greater the potential for finding newer, better ways to do the same old things. Do not make the design and implementation of A.I. something that just the ‘techies’ work on and then roll out to everyone else. That kind of hierarchy is no longer necessary. As we saw with Adele, Beyonce and Cher, every architect now has the ability to design automated processes for themselves, and each other.
Build an Incentive System Around Innovation. At first, no firm principal will have to incentivize their staff to ‘innovate’ new Mini-A.I.’s. People will do so naturally, because people naturally gravitate towards systems that make their lives easier. Rapidly, people will start to realize that they are automating away their own jobs, which will chill innovation unless there is a meaningful reward structure in place. Moreover, architects should be compensated for bringing greater efficiency to the firm. That’s just the right thing to do. In the past, when innovation was periodic and incidental, principals could make ad hoc judgments about particular incidences of innovation, and reward employees who were augmenting the firm’s overall capability, usually through some kind of bonus structure. However, that likely won’t work when you have diffused, interlocking innovations that are happening in a fast-paced ecosystem. And besides, how much bonus would you have to give me to make me want to be unemployed? I’ll be releasing a post on this in the near future around compensation systems, but honestly, I’m still working on the math. In the meantime, a consideration of KPI’s should catalyze some thinking around incentives:
Reconsider KPI’s. In an ‘AI-First’ firm, traditional KPI’s will need to be re-evaluated. To foster an ecosystem of innovation, you want people to experiment. And experiments inevitably fail. As designers, we’re accustomed to that, within the bounds of a design process. Every sketch that ends up in the wastebasket could be looked at in some light as a ‘failure’ - an experiment that didn’t work out. Of course, we know that’s not true - it’s part and parcel of the design process. An ‘AI-First’ firm should use the same logic to design itself.
Let’s suppose I’ve tasked one of my employees, Dido, with performing a field inspection, and I expect it to take 8 hours of working time. Instead of performing the field visit, Dido spends 8 hours trying to program a drone to fly around the site using a LIDAR-enabled camera, so that the construction progress can be mapped directly to the BIM model and errors can be detected. But the experiment fails!
Now, the field inspection isn’t done, and Dido has ‘wasted’ the whole day. Through a conservative lens, this might be seen as insubordinate, a waste of the client’s resources, a delay to construction, or all three. Dido should get a reprimand!
Except, she shouldn’t. Those kinds of experiments are invariably going to be necessary in moving a firm to maximal A.I. deployment, and some will fail. Living and dying by utilization rates is ultimately a short-term strategy. Maybe, if given support and encouragement, and a few days to work it out, Dido will succeed in her experiment. It cost 32 hours of time, but now it saves thousands of hours out into the future.
This will be a challenging adjustment in the near term; I’ve never met a client that wanted to support an architect’s R&D process. How any particular firm deals with that conundrum probably depends on their client base, project types, their staff, their culture, etc. I would say for now that building an ‘AI First’ firm necessarily depends on giving everyone latitude to innovate their own work. So much so, that maybe the first KPI of an ‘AI First’ firm should be ‘how much did you innovate this month?’
Innovate the workflow, not just the work. It’s a gross generalization, but I’ve often found that architects are a lot more daring in their designs than in their work. The drafting board (metaphorically speaking) is the place to unleash creativity! The accounting procedures, not so much.
In building the ‘AI First’ firm, we have to siphon off a little bit of that creativity and apply it to the firm itself. There are a lot of good examples out there of firms who have worked creatively not only on their designs, but on how those designs get done. Ultimately, we should return to the original question: ‘How would you organize a firm, if A.I. were all you knew?’
Answering that question will be impossible in the abstract. None of us really know ‘what A.I. can do.’ Building out an ecosystem that facilitates micro-scale innovation is the first step to answering it, though.
Stay tuned for the next installment, Part 3 of 5 in optimizing for an ‘AI-First’ firm: “Demonstrable Social Value” or click the subscribe to get notified instantly.
2 I use the term ‘micro-scale’ here to refer to innovation that’s happening at the scale of a single individual, and to distinguish it from innovations that are manifested collectively. Although Micro-Scale Innovation happens at the scale of an individual, it is simultaneously and necessarily part of an A.I.-driven ecosystem that compounds innovation at the firm level. And perhaps at the scale of the profession, if LSA is feeling generous. It should also be considered distinct from ‘radical’ innovation. “Radical Innovation” and “Incremental Innovation” and “Small Scale Innovation” are MBA-ish words that have slippery meanings, IMHO. Most ‘radical’ innovations had a lot of incremental innovations leading up to them . . . It’s just more fun and glamorous to record your own innovations as ‘radical’ and forget the giant shoulders you stood on when you did them.