Innovate or Die. Grow or wilt. That’s the feeling I have in the AI space today. Recessionary unemployment from 2008 peaked in 2010 and is now basically over. Now the world is all about growth.

Our consulting business has been taking on many brief AI consulting engagements this past year, to help companies get the best and newest stuff into their products and services. We can’t keep doing this same dog and pony show forever. Contrary to popular belief, I think that at some point, AI solution architecture is going to get easier. As that happens, talent is going to rise in the AI domain, which is a good thing. But for us, it means the high-end consulting we do will be less profitable. The process of creating multi-model AI will also become easier thanks to libraries like keras, that will allow both transfer learning from pre-trained models, and automated deployment of the most common types of neural network models (see here). As an AI company, we need to innovate or die.
So what have we been doing to make money? Below is a high level breakdown of where our consulting revenue came from so far, by sector. It seems we do mostly electronics and cloud AI stuff. No license fees and no recurring revenue. Most engagements in our first year were were for $15K to $30K.

Our pivot to products and larger contracts has been painfully slow, but we are finally there. Our hesitation to pivot was rational. We were not interested in writing RFPs that mostly go nowhere. Instead we were happy as a clam, focusing on creating value for clients, flying around to cool places, and growing our team. We were generally not interested in working on projects with long lead times between the introduction and a signed contract. We are a bunch of highly educated nerds that came from a medical devices background, where innovation moves slowly, and so we wanted to seize the day. We created a funnel, filled it with projects, and plowed ahead. I wrote more on how we got here in this past post.
We had a great time building this business, but everyone has to grow up.
"Grown ups are just kids that got older"
We can’t just keep on running in the consulting hamster wheel, as fun as it may be. We decided on the features of our pivot to products and larger scale services.
What we won’t do from now on:
- Basic science: Awesome technology with no application. We are senior engineers. We make things. We want to make $$, not get a Nobel.
- Basic coding: Trivial technology with an awesome application. No barrier to competition. Why do I have a PhD in ML if I’m going to make "stupid" code? We are kind of over-educated to be tinkering with front-end stuff. Even though we used to do GUIs, we have moved to a stance that front-end work is contracted out, whether for us internally or for clients, we don’t think we are adding value by writing GUI code. Quick recommendation: Marco DeLuca. We don’t mark up design work like this. We just pass it off to people we like. My nicest GUI is not as nice as Marco’s ugliest one.
- Moonshots: There are 2 steps to date a supermodel. Step 1: find supermodel. Step 2: date supermodel. The problem is the details. We don’t want to spend a lot of time doing something that ends up being a special case of proving P=NP. So, we restrict ourselves to the universe of things we believe we can actually make. Not to say that we don’t take on big challenges. We just don’t take on stupid risks.
What we will do from now on:
- Simple User Experience (UX): We are building easy interfaces to awesome AI. When something works like that, it seems like magic. I think all AI should feel like magic. You should design the system to work in a way that users understand but didn’t expect was possible.
- Deep Learning Neural Networks: Our niche market is deep learning neural nets. That’s where our value lies. If we build generic software, we might as well work in a software factory. So, we focus on the crazy high-end stuff.
- Recurring B2B Revenue: Businesses are happy to pay for deep learning, and they hold the most interesting datasets. They also have a lot more money to spend than consumers. We are working on APIs, subscription-based services, and software libraries that generate license fees. These three new avenues open up some cool possibilities for us to scale.
- Big Data and Cloud: We have done too much work on physical boxes in datacenters, and physical hardware components (boards, circuits, enclosures, soldering). Our digikey account is way more active than an AI consulting firm should be. So, we are focusing on our cloud game. Specifically deep learning on the cloud. This is where we want to deliver results to clients that many data science teams can’t achieve.
- Mobile: We don’t make mobile apps, but we work with developers who do. This capability is a nice to have feature. We all use smartphones, so having mobile capability is important. Off the top of my head, I would recommend underlabs.ca Their CTO is a buddy of mine from university, and their tagline says it all: *"We develop cool sht"**
This is not some sort of manifesto. We are actually doing this stuff, and making the pivot right now as I write these words. We will be at the Toronto Machine Learning Summit this week. Here is a glimpse of what we have been cooking up, beyond our usual consulting gigs:
- SageTea Software partnership (Text to Software Deep Learning). Our first software license sale is already in the bag. Follow-on sales interest is pretty hot.
- Another partnership in discussion ([censored] Kit Deep Learning?) is pending legal to sign off.
- Genrush.com is coming along nicely. Our goal with this project is to generate recurring revenue by providing leads to sales teams. The whole approach uses machine learning to figure stuff out. It’s still a work in progress, but it’s definitely moving in the direction of making the first $.
So, we are less and less a start-up. We are growing up, and doing long-term thinking and planning. We are making the magic happen. Exciting times!
Before I go, a few things caught my eye this week. The new p3 instances in AWS are out, and the Raspberry Pi 4 will support faster deep learning with an add-onboard from Google. Everything is moving.
If you enjoyed this article on how we are refocusing our Artificial Intelligence consulting, then please try out the clap tool. Tap that. Follow us on medium. Share on Facebook and twitter. Go for it. I’m also happy to hear your feedback in the comments. What do you think?
Happy Coding!
-Daniel [email protected] ← Say hi. Lemay.ai 1(855)LEMAY-AI


Other articles you may enjoy: