Hello! I am Gary Saarenvirta Founder & CEO at Daisy Intelligence an enterprise SaaS company.

We are an AI platform for retail merchants and insurance investigators/adjustors. We are not a consulting company. We deliver to 25 customers in 5 countries. We deliver 100% plus net income growth for our customers. We are a team of 55 people and growing having just completed a series A raise. Our goal is to become the world’s largest and most respected AI company. We empower people to achieve their best at work by using autonomous machine intelligence to make decisions that are beyond human capability.

Gary is one of North America’s preeminent authorities on artificial intelligence, has over 25 years’ experience working with leading global corporations to deliver revenue and profit growth. He founded Daisy in 2003 to bring autonomous machine intelligence to clients in retail, insurance and healthcare.

Follow Gary on:

AMA Transcript

You just raised $10m - was it super dilutive? Why’d you need the capital?

We need the capital to grow … investing in sales and marketing and customer success to increase retention. Our goal is to reach $100MM in 4 -5 yrs.

Can you comment on the ‘information asymmetry’ that fundraising founders face? In the sense that VC’s have seen hundreds/thousands of deals, term sheets, business models, successes/failures - whereas founders and operators have only often ever seen a few.

Yes the VC’s use the weight of numbers against you … fortunately we are enterprise SaaS so there was not as much data as they usually have. We had outside advisors to inform us if the VC offers were fair and realistic … we also did a lot of research ourself and looked at deal flow that was publicly available

So $6m in revenue today, you just raised $10m. Can you share general valuation multiple range you were able to get? 5x? 10x?

We got a multiple of about 5.5X

Thanks Gary! It is hard to get traction with enterprise level clients. How long did it take Daisy to get one? What were the metrics prior to the very first enterprise deal?

Our metrics prior to first entperise deal was $25K MRR. We then starting buying customers at $5K MRR to get logos and have been upping our prices since then. Recent customers pay between 25K and $100K MRR

Hey Gary, thanks for taking the time to do this AMA! Lots of people in here are in much earlier stages with their companies, whats the biggest single piece of advice you’d give to those just hitting growth stages (or possibly even earlier)?

Focus on delighting your customers … retention is so important to investors .. delighting customers means you get to product market fit sooner. Losing customers is a warning sign to many VCs.

Gary can you explain a bit on how your AI works? Is it real AI or just “marketing AI”?

We use reinforcement learning … same technology that NASA and US Airforce use … what engineering and the sciences have done for 50 years… autonomous cars = autonomous enterprise … we are the latter. Steering brake and gas in a car … product price and inventory in retail. True AI is what we are.

Thanks for taking the time to be here @Gary Saarenvirta 🙂 I’d like to ask, how do you scale your leadership as company grows? As the CEO you obviously know the business best but at a certain point you can’t continue to be in the weeds of every problem/decision.

Building a leadership team is critical … you have to be a storyteller as a founder/CEO … get people to believe in your mission … set a big goal and get people excited … fortunately I am agood storyteller and want to change the world and have a great ELT. Find leaders who match up to your weaknesses … your ELT should fill all the needs as a group. We do behavioural profiling … $1K per person

Understand your product is based on reinforcement learning, how do you forsee the underlying tech changing as new state of the art RL research papers and approaches are released seemingly every week?

Our approach is that of physics … theory first … then data … our RL is based on a set of partial differential equations like the laws of physics. For complex systems you cant learn the dynamics from data alone … retail and insurance are too complex to learn from data. The whole world has seemed to turn upside down.

Sorry if I missed this, what is your pricepoint for enterprise? And if it is large what do you use for billing/accepting payments? (We have large size clients - $25K/ARR and it is becoming too cumbersome without a billing system. Invoices from quickbooks + ACH/Wires was fine with 6 clients, not 60 and growing)

Our pricing is based on client size, number of Daisy nodules, granularity and freq of decision making … clients pay between $15k and $150K MRR

So at LTV $2.4M over 120 months, the average customer is $20K/mo, and you have ~16 customers. I would think it would be a very custom software/productized service model more than SaaS business model. How do you handle request for custom features, are you building them to keep the customer or focusing on making a product that works for all of them (but not necessarily each getting exactly what they want)

We are a product … very little customization .. we deliver decisions … the goal is to do this autonomously with no human in the loop on client side … we have no human in the loop at Daisy … our SW is autonomous. AI empowered autonomous business decision making for problems beyond human capability. We do do some customization but only if it is good for all customers.

@Gary Saarenvirta how would you feel if someone offered to acquire you right now for $30m? That’d put nice $$$ in your pocket but you couldn’t take the deal because your new investors would block it.

I would not sell the company now … our mission is to be this generations tech giant … Facebook and Google and apple being last generations … I want to change the world.

With your new round assumedly going towards a hiring blitz, how are you sourcing data scientists and data engineers? There’s a scarcity of top AI talent

We don’t have many data scientists … we hire parallel computing SW developers. They implement the math designed by a few Engineering Control theory experts … I think the focus on ML computer science people is slightly off base. My background include data warehousing … we are very strong in data engineering on our team, I agree data engineering is critical.

Could you explain the journey to get to MVP - how long did it take, how many iterations, how many pilots, did you work with a specific customer in defining MVP, etc. given it is an enterprise pdt.

I built the tech over 12 years … doing 1 few pilots at a time … as a professional services company … then we pivoted and become 100% MRR … we did lots of feature enhancements with first 10 customers to get to what we considered MVP.

Gary how much cash did you sink into building your MVP before you had your first dollar of revenue?

$8MM of company profits when I owned 100% of company … using government tax credits, grants and revenue from prof svcs. I ran $30MM in prof svcs revenue and sold that business … and we are now 95% recurring.

Gary how much do you have to grow your ARR by before you feel like you could go out and do the next round of funding?

Our B round will be at around $25MM ARR. I want to keep capital effcy at or near 1 prior to each round raised.

How long do you think it’ll take you to get to $25m in ARR up from your $6m currently?

2 years, we will close this year at $8, next year at 16 and be at 25 in fall 2021…

How many engineers are on the team out of the 55 total?


What opportunities / threats are you most excited / cautious of at this point in DI’s life?

Opportunity to spend money to get the message out and scale internationally … I am hyper concerned about speed and creating a culture of urgency and we need to get a forecastable sales process … enterprise SaaS being very lumpy.

Favorite business book?

Built to Last … Habits of Successful Companies … old one … new one is Play Bigger and The Traction Gap

How much upfront, bespoke customer data integration is required for companies to use your platform? Do you account for this as part of your CAC? Can companies do tests with small datasets before the ‘rubber truly meets the road’?

We require 2 man months of mapping client data to our system … we get it done in one month … its pretty straightforward

simulation based reinforcement learning … and optimal control … invent the theory first and find data second. create partial differential equations that explain how your complex system works … engineering has done this since the 50s … called optimal control

This was super awesome … thanks for all the great questions … hope it was helpful!!