Hey everyone! I’m the CEO of Vue.ai (Mad Street Den is the parent company). Our AI products automate a range of workflows within retail organizations using image, video and text recognition.
We help brands save costs and grow their business by automating time consuming workflows, helping them ship their products faster, targeting them at the right audience for higher growth. Think automated product tagging, digital models without photoshoots, AI based styling and personalization.
We’re an enterprise startup with a 200 member team across the US, India, Japan and Europe. I used to be a product designer at Intel in California for a decade. Then moved back home to Chennai - India, to build this company.
It’s been an exciting and equally exhausting few years given everything in this space is so new. But I’ve found a lot of joy in the challenges and discovery that comes with creating new vertical businesses on our AI platform. We’re also working to launch beyond Retail in the coming year.
I’d love to chat about:
Org building, nurturing & growing talent
Value generation with AI
Marketing when your category is relatively less known or new
Customer Success
Building Enterprise sales teams
Sanity & Habits in the time of Coronavirus
Or anything else on your mind
I’ll be here to answer questions on July 23rd at 3:30 PM IST.
Note: This AMA is closed for new questions, but you can check out the existing conversations below.
In this AMA, we had Ashwini Asokan — the founder and CEO of Vue.ai, a former product designer, and an advocate for women in tech — share her thoughtful insights on PMF and GTM strategies, building and selling in the AI space, the career ladder metaphor, and more. Dive in!
AMA Index (Ashwini’s brain pickings)
(founding insights, opinions, and observations; deftly examined and articulated)
Further reading/listening/pondering from the interwebz /
(Other insightful excerpts drawn from blog posts, interviews, and conversations)
On PMF and GTM:
“Understanding the user and iterating on the initial prototype or MVP, almost always happens as you sell to the first few customers. User research and iteration happens outside of a sale, only in academic and research settings. ‘User research’ in a startup happens in production, in pre-sale, in QA, across post-sale integration and more. This is why carefully crafting an early go-to-market process can be the difference between getting to PMF and feeling like you’re constantly one quarter away from the product taking off. And there are many ways in which this is revealed.”
Source: PMF & GTM: 2 sides of the same coin
On the career ladder metaphor:
“Thinking back about my own career, I can say this for sure — breaking down the metaphor of a ladder, allowed me to create spaces where I could bring my own strengths to create my own landing space. More often than not, this was not the top. And this ‘other space’ I designed over and over again, often made what others defined as ‘the top’ — entirely irrelevant. The top is an illusion. It’s not a single place that we’re all running towards.”
Source: Careers & Startups: Moving past the ladder metaphor
On the current crisis:
“The way I look at this is that this is an exercise in change management. That is what it is. And it is not just change management at the level of the work you’re doing for a particular customer or you know a product roadmap. But it’s change management at the level of an economy, at the level of humanity. The way people are going to be. The way countries are going to be. Entire professions and industries are going to be. And you cannot go through change management if you haven’t gotten to the bottom of exactly what is the change that’s happening, who are the people that are involved as part of the change, what is expected of each one of us as we go through this process, and what lies on the other side of change. And I want to call that out as step no. one in this process.”
Source: Hope is not a strategy
On the inherent complexity of building AI-driven products:
“…Okay it looks like this part of this original use case can be solved 100%, this part of this use case of this original problem that we were exploring can be solved at about 80-90%. And then as product people we gotta go back and figure out how can you actually construct the problem, now that you know what your sandbox is. So much of your work as a product manager is to identify that sandbox. And that iterative process, it’s pretty much the same. A similar product design process anywhere else in the world with any other product. You got to go through the same thing except, you’re not going to get out of this knowing the process is not simply for product development; it’s actually for defining the sandbox before you even get into product development.”
Source: Beyond the Hype on AI – Ashwini Asokan on The Product Experience
Stay in touch:
You can follow Ashwini to stay updated with her discoveries and insights:
It’s been great to witness Vue.ai’s journey over the years.
So really excited about the session!
My question is about decision-making concerning data, something that’s fundamental to AI-driven businesses of course, but also increasingly for more and more enterprise offerings in general.
— Given the need to have that initial data corpus unlike system-of-record software, how differently should founders think about navigating the early months/years of balancing research and validation? And ultimately, chasing product-market fit?
I’ve got a broad one in mind. I’ve long been curious about how different companies tackle unknowns. Given the nascent — not enough playbooks for teams, especially for B2B — premise of a product like Vue.ai, there must have been an abundance of them. Looking back, what are some of the key, early operating decisions/processes that have helped you the most as a founder?
Hey Krish! Glad to be here. The question is fairly broad, so super high level answer here, more principles and guidelines than anything else. If you had specific areas you’re looking for answers to, holler!
KNOW YOUR DATA
Your data = people or the things they use or content they interact with or goods they buy. Your data is subjects with stories and experiences before it’s statistics. And before you embark on any journey to build AI enabled use cases or automate workflows, the first thing to do is to ‘know your data’. It’s important that you understand your data, your subject before you decide to automate. All our product people, AI teams, customer success teams are trained on interpreting data before they build or talk to customers. How you train your data will depend on the patterns and inferences and assumptions you make, so this is very important.
MAKE YOUR DATA TEACH YOU, ASK QUESTIONS
Just like 1, its important to cut the data a bunch of different ways by asking several questions, relating to the workflow you want to automate or the use case you want to integrate AI with in some way. You’ll be surprised at the patterns you discover. Big part of the research process is to let your data sing and then understand it in the context of the subjective / anecdotal / qualitative observations you do.
HYPOTHESIZE, EXPERIMENT, ITERATE
Now you’re ready to use the data to train new models for whatever use cases you’re looking for. Like with any new feature building, experimentation, iteration is super important. There’s so much hype around AI. If your AI is not adding meaningful value to your customers, don’t do it. Have a hypothesis on what you want your AI to do, test and evaluate with customers and focus on the data emerging. You can setup several experiments in parallel with different control groups and test new ideas. Evaluating usefulness or value of AI features is a topic I’ve been interested in for years now. Depending on whether you’re simply looking to automate / increase speed / throughput vs. trying to change human behavior, the outcome can happen as quickly as immediately or can take as much as years to have an impact.
Hey Rajaraman! Thanks to your team for having me here.
I absolutely love this question because it’s probably one of the top things on my mind 24/7. Everything about AI has largely been unknown. There’s been so much hype and noise about AI. Very low signal:noise ratio honestly. And while there have been a lot of new businesses coming up in the space, we rarely actually hear about the ones solving real problems in verticals. Here are some of the challenges I’ve seen:
Lack of an actual problem to solve. AI is often a hammer looking for a nail. A lot of companies I’ve seen in this space, fail because they fail to use AI to solve a real issue.
The category is non-existent and the job of the company is to create the need and change human behavior. This is a gargantuan task. And one that I’m deeply familiar with because we’ve been going at this for years at Vue.ai. I’ll get back to this in a moment. There is a huge component here that is entirely about educating the audience. It took us years of educating, engaging with the audience before we could crack them on scale. And we were acutely aware that the industry we were dealing with was largely kitted out with legacy tech.
Helping humans who are doing these jobs get over the fear that their job will not exist in a few years. This is a very real issue. People in companies using these systems are very suspicious of AI systems and very unforgiving, as well. Often with good reason but it’s a challenge, regardless.
Lack of understanding on RoI & emotional impact of AI on users - Our users often said things like “ya the machine failed this once when I really needed it, so we might as well do this manually”. This meant they overlooked the fact that productivity and throughput tripled when compared to non-AI based systems or manual labor but focused on the one time that they looked bad in front of their senior management.
Here are some of the things we’ve done over the years to counter some of these:
PRODUCT MARKETING FTW:
We invested heavily in product marketing from day 0. This was perhaps one of the best things we did. Whether it was demos, FAQs, use cases, new ideas - our goal was to constantly demonstrate all the ways in which our systems could help people reimagine their work, reimagine new workflows with a 10x / order of magnitude growth in either revenue or savings on the expenses/cost side.
CUSTOMER MARKETING FTW
We invested heavily in customer marketing from the time we were at $1M ARR. A big portion of category creation is showing that the category is being adopted by the industry, saying stories about that behavior, RoI and more. People constantly want to know what everyone else is doing. You’d be surprised but the #1 question we get on sales calls is: “What are other retailers in this space doing”. A lot of this process had to do with normalizing, familiarizing the category and making it less ‘other’, alien to their daily work.
TEACHING THE AUDIENCE ABOUT RoI of AI
This one was huge because it helped us show people we were not there to put them out of their jobs but also help them produce 10x better work. The language took us a while to iterate and learn but eventually we realized it was not about how AI was the star of the show, it was about how the AI made those users the star of the show.
HELPING THEM TEAM REIMAGINE THEMSELVES AS EDUCATORS NOT COOL AI PEOPLE
The early years of a startup’s journey is exciting filled with adrenaline rush and a lot of ups and downs and emotional outbursts. I can tell you when we crossed our 3 year mark, for me it was all about helping the team undo, unlearn, rethink, remake and rebuild themselves not as people building bleeding edge tech but about helping them think of themselves as people who have a responsibility to help the world around them become AI-Natives. I can tell you this is one of the best things that’s happened to the org as a whole, maybe we should have done it sooner. But the responsibility that comes with enabling people with AI is a very real one and that can help counter the fear and the high threshold for change.
Thank you, Ashwini. I have a follow-on question and please feel free to answer whenever you get time. We see the evolution of OpenAI systems and so much chatter about GPT-3. If you were an early stage founder who is building a SaaS product in applied AI space, how would you think about this? What do you think would continue to be a value added service that customers will pay for to build a moat for your business vs. probably of getting disrupted quickly?
Thank you so much for doing this AMA!
Really cool company. I’m wondering what sorts of patterns you noticed across your customer base that has helped you automate different parts of your on-boarding process.
Coming from a non-retail background, what helped you in defining the use case you’ll solve for in your early stages clear & crisp?
In context of 1, how did you figure out who within your ICP will be your product’s early adopters? Any tools or strategies that helped you?
Apart from the tech leverage, how much does AI help in creating explicit value for customers? Or in other words, what role does AI play in your selling strategy?
I was just looking up Private AI - you guys are doing some much needed work yourself! Very cool. I might want to connect offline at some point about your work here.
We don’t actually use automation in onboarding. Most of our offerings are high end enterprise solutions and this often means white glove kind of onboarding - very human centric It is a bit ironical that a complex set of AI requires a lot of human interactions with decision makers on the other side to get them onboarded. While we have productivity tools to help customers tell the AI if they want something changed or edited or to override decisions, the onboarding happens entirely offline.
We deliver all our products through API integration with systems/platforms retailers run on, so this up front part of it is very hands on integration. We’ve invested heavily in customer success and delivery teams to enable a smooth up front week or two’s integration and onboarding. I’m not entirely sure if this is what you were referring to. If it’s something else, happy to discuss.
AppDynamics’s founder, Jyoti Bansal talks about how there are two phases of product-market fit. One, of course, is to identify a target group that finds the solution most resonant. Second, is to arrive at the ideal sales model. He even calls it the product-market-sales fit.
What helped you the most in arriving at the latter? And given that its enterprise-focused, how do you think about the balance between customization and productization as well?
Thanks for the questions. These are questions that I’ve spent a lot of time reflecting and thinking about over the years I’ll answer a couple and come back to the rest.
Not only did we jump into retail as complete retail noobs, we also jumped into enterprise with no background there either. Feels like a double whammy in retrospect. To be honest though, we were blind to both and it might sound weird - but it was a good thing in a way because we definitely lacked the fear we now feel, as people who know so much more about this industry than when we started.
We did 3 things really well:
One, we knocked on a lot of doors and spoke to a lot of retailers who were willing to talk to us. We had a hunch about a set of use cases in personalization that our Computer Vision stack lent itself really well to. So we built a bunch of demos and took it around and did a lot of show & tell but more importantly, used it as a simple starting point for conversations around problems retailers had. We were way ahead of the market (I don’t say that in a good way) and that meant people often got excited and gave a lot of feedback and ideas but never really followed through. We were able to get people to open up their systems, stack and show us what they had going. Lesser the pressure to buy, the more the intent to share, sometimes, especially in cases where your category is non-existent but exciting.
Two, we invested heavily in product marketing & customer marketing and had some really good demos to help explain what AI can do for their business. We led our story with here’s what Computer Vision is, here’s what it can do for your business and here’s why you should care. Half our problem was to help the retailer see why they needed to move status quo. And you really need those first 2-3 customers in enterprise to help launch you.
Which brings me to three, without even knowing Customer Success is what we were doing, we helped our first few customers so much - white glove like I said - that they really stuck with us, are still with us, speak at our events and have become our evangelists. We invested a lot in building those evangelists.
On the ICP question - it was actually simple. There was no way the big box retailers or the brands were investing in AI in 2016 It was the e-commerce, online, tech first folks who were going to invest in any of these features. This also meant competing with internal teams but we eventually won. So we replicated our success in that segment and went after the biggest online retailers as our starting point. Exec teams mostly.
As a successful woman entrepreneur, what are your inputs to young women entrepreneurs to succeed around challenges and obstacles? please share your views.
Value generation & Marketing with AI - With limited spend allocated, organisations are bit hesitant to experiment new ventures in the field of AI, where most part of the sales cycle goes into educating or creating awarness. How do you see this can be best tackled to keep the CAC and sales cycles realistic?
Now that Vue.ai is expanding into multiple segments (grocery, electronics, and others), how does the notion of repeatability play out? As in, is there something that’s common between the routes to market for all these segments and that helped with the decision-making?
Or does the move come from having tested out multiple GTM (direct/channel/inbound) strategies over the years and now being ready for varied segments? Would be helpful to learn about the high-level thinking process here.
I was just looking up Locale. It’s an exciting space. Good luck!
In the category we’re working in, the need for AI enabled use cases means delivering in a way that the AI lands in some part of an existing workflow, which is either manned by a human or intersects some fairly old legacy system!
We realized fairly early on that most of our products would never work in a self serve model. Our challenges revolved around convincing people to trust the AI, trust value and trust that it won’t break the systems they already had in place. And that automatically meant we were going to pick the enterprise route for sales.
Then we had to figure out how to price the up front integration vs. the recurring API parts. We iterated a lot on the pricing. It was directly a function of the RoI and value we were able to generate for our customers. And then of course, the upsell and cross sell cycles continued once that trust was established.
We don’t do services. You can’t have armies of humans in the loop when you’re delivering real time personalization, data extraction, classification related use cases. But in the first few years, we had the luxury of customers who allowed us to experiment with them on use cases, to iterate and learn. This is very important for enterprise. Those early supporters, adopters can often be the difference between take-off and crash.
I speak to a lot of early stage founders and this is the singular question that keeps coming up - customization vs. productization! There is no right answer. Every business is different. If your goal is to build a SaaS company, don’t let the customization eat you whole, is all I tell them. Building in the ability to customize, into your product workflow can also be a great way to deal with this. Hope that helps!
I’ll share a few rules/guardrails/guidelines I’ve had for myself over the years.
Don’t take No for an answer, keep at it, iterate and come back
Punch 10x harder than any of your peers. The playing field is not level. There’s no other choice
You can differentiate and build a massive organization that can fire up bellies full of ambition while being thoughtful, empathetic and caring. These are not mutually exclusive.
There is discrimination, there is bias. It just is that way. Fight it by actually building the company you’ve always wanted. Change starts at home.
Ensure diversity bottoms up and top down. Lead the way and show everyone we can change the status quo. You can’t achieve this by having a diversity officer, you need a culture of diversity.
Surround yourself with mentors and people who will help you see possibilities beyond a world filled with bias.
We stand on the shoulders of so many who have fought this over decades. Pay it forward if you can. Nothing changes the status quo like enabling a 1 more founder like you to succeed in whatever little way we can help
It is a hard time for many industries. But people are buying ‘need’ based SaaS products more than they ever have before. In many ways, this is a great period for SaaS products that tackle must-haves, mission critical use cases. So more than sales cycles and CAC, this is actually a great time to think about product marketing and messaging.
We spent a lot of time thinking about how COVID impacted retail and iterated on our content to suit everything that’s changing in the market. Now more than ever before, its important people relate to you and your product. So invest there.
We’ve also been asking ourselves about who the winners and losers of this pandemic are. It’s an important one if you’re tuning your lead list. Chances are - you’re going to get a better chance of winning the guys who are survivors of this situation, than those who are not equipped to survive this market.
We’ve been wanting to expand for over a year now and been patiently waiting for the right time. Our AI has seen enough data over the years that it’s started to work really well in non-fashion categories too. We did some early tests in each of those other adjacent verticals and realized it was working very well. So we implemented it with a few existing customers who had multiple categories, measured the value and then launched.
Product and delivery wise, the teams are the same. The repeatability is very high there. On the GTM side though, it’s a whole other story. We have the same teams covering both today. And while it’s certainly getting a bit much, we’re just getting started here. And I’m a big believer in lean, highly ambitious teams that want to learn and grow. This new expansion has my team excited about the opportunity to learn more and they’re getting really creative. We’ll reach a point in 3 months when the demand outweighs the capacity of our GTM resources I’m banking on it and planning for it.
We’re planning to have separate inbound/outbound/sales teams for fashion and non-fashion. Some verticals behave very similarly - like beauty and fashion. So we’re using this quarter as a way to capture all these similarities and differences before we implement 2-3 separate teams.
Hey Krish
So I have some strong and unfiltered opinions about this
Big news in AI is always hyped. Real progress in AI happens in the real world, in the context of your day to day business. And it can take months or years to see RoI or real value. Don’t get me wrong, I’m not saying its not exciting. The reason everything is so hyped about AI, is precisely because of how exciting it is. But as someone who plays a big role in advising our customers about noise vs. signal, I can say learning to filter through the hype and taking away some actionable items for your business is where the emphasis should be. Along those lines, here are a few:
Focus on the use cases and problems you can solve: Ask yourself if you can use GPT-3 to build a new feature into your workflow? Can you improve the search you use? Can you improve customer support? Can you use this to generate automated emails and messages to customers? Copy on the website, perhaps? To filter out negative comments and reviews, so you can plan course of action?
Does it scale? A large % of AI that is out there, whether it’s tools or solutions being put out - does not scale in production. This is actually one of the main reasons we started Mad Street Den. Separate the hype and demo from the ability to scale what you see in the context of the use cases you see.
Experiment, iterate, deploy, learn. Like any other product development activity, if not even more - AI needs the care and time and cycle of experimentation and deployment and feedback - to see if it’s even relevant. We can fool ourselves into believing something is true, if the flash quotient is high I’ve noticed this doesn’t last. Customers move onto the next thing and stop using this. We have actively advised many of our customers to not use some of our initial features, despite their being a very high demand for it in a particular period. We stopped offering those features and let our customers to go shop for them elsewhere. 3.5 years in, I can tell you it was one of the best decisions I’ve ever made. Every little thing that doesn’t move you forward, sets you backward in some way.
/
and then understand it in the context of the subjective / anecdotal / qualitative observations you do.