AiThority Interview Series With Alain Briancon, VP of Data Science, Cerebri AI
User interfaces are ripe for the impact of AI. Smart dashboards in cars, automatic real-time voice translation in my earpiece, recognition of subtle motions for control.
Know My Company
Tell us about your interaction with AI and Data Science? How did you start at Cerebri AI?
I have known the CEO of Cerebri AI, Jean Belanger, for close to 20 years. He has tried to recruit me in his previous start-ups (all successful), but I declined in order to pursue my own corporate then entrepreneur career. Jean asked me to first join Cerebri AI as an advisor, helping part-time and over the weekend. I agreed and started to understand the technology and go to market.
When he asked me if, this time, I would join him, I said “Yes!”
Because this was and continues to be too cool and powerful to refuse.
Cerebri AI has the potential to entirely change the way customer experience is managed by leveraging real data and commitment. This is the fourth startup I have done where data science has played a key role. I was twice a CEO and twice a CTO. I have seen data science used for IoT, diets, elections, and now helping the Fortune 500 take customer experience to the next level without bias or fear. Cerebri AI is, bar none, the most cutting-edge and rewarding.
What are your interactions with other intelligent technologies in your daily life, including Alexa, etc?
Siri, hello Siri?
At times, I try to get my Apple watch, iPhone, and HomePod to talk at once. I say “attempt” because I am not impressed (yet) and wonder if the absence of a well-documented and used a markup language such as Amazon AMRL (Alexa Meaning Representation Language) is impacting Siri’s use in new domains. I experience the Amazon recommendations too often, for my wallet. Netflix is my testbed for AI. I maintain five personalities on my account and I am always fascinated by how different the recommendations are for each one. It is worth trying at home.
As a data scientist, how do you see the raging trend of including ‘AI in everything’?
AI is inside so many things, I don’t see it as a trend anymore. AI is a fait accompli. It certainly is for the online world. I take it for granted that all online interactions, that I encounter, have some AI model behind them, especially for recommendations.
What I am starting to observe is the move to define products on the fly. It starts with pricing (airplane seats, batteries on Amazon). I am still waiting for key breakthroughs. User interfaces are ripe for the impact of AI. Smart dashboards in cars, automatic real-time voice translation in my earpiece, recognition of subtle motions for control. In the area of system optimization, I am waiting for smart cities and traffic systems driven by AI.
Fixed light patterns are unacceptable, especially in DC. Healthcare is the area where I am the most enthusiastic, yet anxious about AI’s impact. The ability to have sensors continuously record your wellness status, not solely at the doctor’s.
To have a continuous risk, behavior assessment, recommendation for drugs will impact our lives greatly.
As a mentor in the tech industry, how should young marketers and sales professionals train themselves to work better with AI, Machine Learning, and virtual assistants?
How do you get to Carnegie Hall? Practice, careful practice. I would try to go to meetups and see how startups and emerging companies pitch their products and services. You often have the opportunity to ask questions to the creators after a presentation. That is a great way to understand what, why, and how. The rule for meetups is, some free advice here, get to the pizza quickly, and don’t be ashamed.
How do data scientists engage with business leaders and emerging research in the AI/ML space, and how does the Cerebri AI platform solve related issues?
The job of data scientists is to create business options that create value, increased engagement, and better KPIs. They must think about describing their technological achievements in terms of business impact. At Cerebri AI, every model’s performance is measured along 11 to 13 technical performance (we are way beyond AUC, F1, recall) metrics and also lift tables in terms of improved KPIs.
I think this is the way to go. If you can’t put a $ measurement at what you communicate to the business unit (or customer), you have not done enough. You need to also explain the ‘why’. I believe model accuracy, impact, and interpretability are the three essential legs of the data science stool.
Deep learning and context recognition seem to be the main wave of research.
The Cerebri AI platform takes a unique view of the creation of engagement. Firstly, we deliver our solution through a SaaS model, where our platform operates behind the corporate firewall. In working with a corporate or government customer, Cerebri AI imports data from both internal and external sources to build and visualize comprehensive customer journeys for the tens of millions of individual consumers. We resolve (to a large degree) the issue of onboarding, through the use of vertical market ontologies organized along key dimensions (event, customer, product, time, etc.) and also extensive feature analysis.
Read More: Will Business Development Ever Be Automated?
Cerebri AI then applies Machine-Learning modeling to quantify the impact of each event in the journey for every customer.
Our timeline framework allows us to handle irregular events strings from multiple sources that ARIMA, CF, or traditional supervised learning models cannot address (It took us time to get there). We go one step further in monetizing every journey event, by valuing them in local currency terms. Our Cerebri Values system dynamically values the impact of subsequent events in a customer’s journey and value every customer’s commitment to the brand, and the products they use. We also provide a measure of why the commitment to brand goes up and down with time and events.
Creating a common denomination for the performance of elements of the system is one of the most important tools in managing the customer experience (I got sold on that concept during my consulting stage with Cerebri AI).
We include dynamic risk in a customer’s journey enabling a better understanding of which customers qualify for offers before those offers are sent to the customer. This has proven invaluable with an early banking customer. With the impact of every event valued in local currency terms, we can calculate the next-best-action for each customer giving the opportunity to optimize the company’s marketing and sales campaign.
Could you tell us more about Cerebri Actions, otherwise known as “Next Best Actions”? How does it improve marketing campaigns?
A next best action (NBA) is the mechanism through which our software provides a specific set of actions a company can take to improve specific KPIs. These recommendations match customers, who are most likely to convert, with the best tactic for achieving the desired outcome. We build NBAs based on the impact that similar actions have had in the past. Two customers might have two totally different NBAs, at the same moment in time, because of differences in their journeys. We are also able to provide key attributes for the interpretation of our NBAs.
A correct interpretation of a prediction model’s output is extremely important. It engenders trust, provides insight on how to improve NBAs, highlights new scores to develop, and supports understanding of the causality vs. correlation components of the customer journeys being leveraged. Because we can predict the impact of actions on a journey, we can eliminate actions that have a negative impact, like a bad marketing outreach. The one all-nighter I was involved in at Cerebri AI was one where we measured, with real customer data, the impact of sending a new-vehicle marketing email on service engagement.
For a lot of consumers, the impact was negative. They did not want to learn about new cars when dealing with repairs. Every customer was a different dot on the map, a different impact, a transformative night. Since then, we have developed the NBAs framework for loan adjudications, service plan purchases, car purchases, and more.
We are glad to have strong early customers, who share the results of the implementation of the Cerebri Values system with us, allowing us to compare real performance to predicted performance.
How do you consume information on AI/ML and related topics to build your opinion?
I read Medium, TechCrunch, arxiv.org outlines, and the blogs of key players as my source of inspiration. Based on what I find, I create Google Alerts and scan Google Scholars to find the more technical papers needed to understand the technology and framework better. I am lucky to lead a very talented team at Cerebri AI, which enables me to delegate some reading to a staff member smarter than myself, who turns these papers into knowledge-sharing sessions. The analysis from the team informs my opinion greatly. I consume a lot of information on my iPad. Being a light sleeper, when it comes to sleep aids there is often nothing better than reading a Ph.D. dissertation (P.S. I read the thesis of all applicants before recommending a job offer).
How companies can further improve their search functionality by using Natural Language Processing (NLP) and AI? What makes Cerebri AI a leader in this space?
Context is key to interpretation. I am following with great attention how the Alexa Meaning Representation Language, ontology language, and a graph-based approach will complement deep learning to further improve recognition. Cerebri AI uses NLP in a generic manner to interpret and taxonomize marketing creative and notes.
What makes understanding AI so hard when it comes to actually deploy them?
Scaling anything to global proportions is difficult. Sometimes we don’t give credit where credit is due, thanks to the massive complexity of providing goods and services of uniform quality, etc., on a global basis. So, enterprise-wide deployments are hard to do, but there is sufficient margin to accommodate that challenge.
How does an end-to-end solution with data capturing of online behavior help a company better compete with the likes of Amazon on Google search? How does Cerebri fit into this ecosystem?
If you know the history of your customer, you are better able to provide unique recommendations when interacting with them. All the pages on Amazon are unique to you. Amazon builds on interactions (clicks, purchases) that you make. We enable to do the same for Fortune 500 companies in the Telco, automotive, financial, and B2C business, as well as government (what we have termed “G2C”).
The key is to have access to the corporate data and create a data store that allows customers journeys to be built. Once this is done (typically in a month or two), measuring journeys at scale (unlike surveys) provides the key to success.
Elaborate on your playbook for company-customer interaction and the role of AI/ML in making it human-friendly.
Agile design and quick interactions are key to making it human-friendly. Develop a modular widget-like approach and capture the analytics on how users leverage the system. Don’t be shy to announce you are wrong, when you are wrong, and right when you are right. The cost of iterating designs is so low, you have to use it.
How potent is the Human-Machine intelligence for businesses and society? Who owns machine learning results?
Corporations and public institutions own the results of their “Machine Learning,” but they have a responsibility towards consumers to not abuse the privacy of their customers and users. There is, rightfully in my opinion, a big backlash against social media and the like, where interaction and social data is abused. In the case of corporate interaction, where unlike social media, the product is not the consumer, the situation is less fraught with danger of abuse.
You have to be careful though. Multiple businesses each see a part of your journeys. Together, they can switch to a richer view of you and provide a better service (this is certainly true for healthcare). We see this as the future of journeys. This is where blockchain can be critical. Stay tuned.
Where do you see AI/Machine Learning and other smart technologies heading beyond 2020?
On the serious side, we will have the ability to synthesize drugs through software and AI will play a key role in the definition of the compounds and the way regulators approve their use. We will have the ability to assess and modify DNA to help with conditions. The world will change. On the fun side, actors will be rendered in a manner that is indistinguishable from reality.
We are getting close to the recent Rogue One movie but give it a couple years and it will happen. An actor is scanned, and they’re used and reused. This will make for fantastic contract negotiations. Of course, political campaigns will spoof their opponents. Imagine AI-powered fake news and we will look at today as the “good old days.”
The Good, Bad and Ugly about AI that you have heard or predict –
Good: AlphaGo Zero that self-taught itself and beat AlphaGo in 3 days.
Bad: The scraping and leveraging of Facebook data for illicit purposes. This has created a backlash that might spill over to (more relatively) innocent players.
Ugly: Cancer detection based on synthetic data. I am truly annoyed, I hope there is a mea-culpa and this is fixed.
What are your thoughts on the ‘weaponization of AI”?
This will happen, as the history of humankind teaches us that (alas) technology will be used for good and bad. I am not as concerned about weaponization in the battlefield, but that any weaponization will be in elections, discourses, news and the like. We are becoming balkanized in our information sources and recommendations, missing more and more common ground. That is scary.
The Crystal Gaze
What AI start-ups and labs are you keenly following?
I follow the companies that M12 (formerly Microsoft Ventures), Google Ventures and Intel are investing in because I trust they know what they are doing. Toronto-based accelerator Creative Destruction Lab has great graduates. MIT and Stanford spin-outs are also on my list. I use Feedly (an RSS reader) and Google Alerts to track them, parsing on keywords (never mixing with emails). More recently, I have been following the companies who joined the MOBI consortium and blockchain players. I also follow AI companies based on who contributes a lot to GitHub.
What technologies within AI and computing are you interested in?
Reinforcement Learning, Context Extraction, Latent Intent Analysis, Containerization, Blockchain, Ontology Generation.
We are using or starting to explore the use of these techniques at work and some are really promising. One of the critical challenges in providing one-to-one recommendations is scaling while minimizing the computational load and getting clean signals from the heterogeneous event types.
As a tech leader, what industries you think would be fastest in adopting AI/ML with smooth efficiency? What are the new emerging markets for AI technology markets?
Transportation and utilities should be able to adopt AI most easily. Telecoms should be next on the list. Bio-engineering and pharma are starting to emerge as key players (The DC area is rich with start-ups in this arena, so this might impact my worldview).
But, first and foremost, are financial services, which are traditionally never far from the lead on new technologies. This time around a surprising early entrant is automotive, where massive improvements are being made monthly in terms of autonomous driving and customer experience.
What’s your smartest work-related shortcut or productivity hack?
My most effective tool is having an empty mailbox every night.
I do not mean all messages read, I mean an empty mailbox. If I can’t resolve an action item, I forward it to OmniFocus, a personal task management system, where I capture all my action items and key team action items. I review my list twice a day — 6:30 AM, as I ride the subway and 7 PM, as I ride the subway (again).
Tag the one person in the industry whose answers to these questions you would love to read:
Thank you, Alain! That was fun and hope to see you back on AiThority soon.
Alain is a seasoned technology executive with a passion for building and leading teams to disrupt old thinking. A serial inventor, recognized innovator, and entrepreneur.
Leveraging the massive amounts of customer data recorded by Fortune 500 companies, Cerebri AI delivers actionable insights via its patent-pending Cerebri Values system, which uses artificial intelligence and machine learning to personalize customer experience (CX) at scale. Cerebri Values quantifies each customer’s commitment to a brand or product and dynamically predicts the next best actions for CX success. Headquartered in Austin with offices in Toronto and Washington, DC, the company has 50 employees who have been awarded over 130 patents to date.