Episode 8
Vulnerability Is a Strategy Too
with Shael Kerbel
Shael Kerbel has spent a decade across FP&A and corporate development before landing at Oliver & Bonaccini, one of Canada's leading restaurant and catering groups. His view of finance was shaped by deals and diligence that taught him the numbers are only as useful as the story they tell.
Today he is building AI-powered tools that connect POS data, HR data, and financial planning into a single operational picture, and pursuing the harder personal work of becoming more comfortable being seen. This conversation covers both: the technical infrastructure needed to trust AI output, and the surprisingly simple practice of opening doors by being a little more human.
Thursday, May 7, 2026
.png)
Making Cents with Charlie
Podcasts /
Vulnerability Is a Strategy Too
The "Two-Cent" Advice
"Be vulnerable. Talk to people, open doors for yourself even when it is uncomfortable. And train your AI model. Take the time to learn about prompting, use the memory features, put in your instructions. You will get way better results."
- Shael Kerbel, Senior Manager FP&A, Oliver & Bonaccini Hospitality
Episode Key Takeaways
1. For numbers, use code. For text, use prompts.
Shael draws a clear line between the two kinds of data AI handles. Financial calculations need to run through Python or another repeatable, codified tool, not through an LLM that can quietly change its answer. Text-based output like research, summaries, and outreach angles is where prompting and training pays off, with the understanding that some variance is acceptable.
2. The biggest AI hurdle is not building it, it's making it work for everyone.
Getting an AI tool to work for the person who built it is the easy part. Making it production-ready, consistent, and adoptable by a broader team is where the real work lives. Shael is deliberately building his sales pipeline tool to be simple enough that the adoption problem almost solves itself.
3. Data that does not connect cannot tell you anything useful.
Oliver & Bonaccini has Vena for finance, a POS system for operations, Dayforce for HR, and OpenTable for reservations. None of it talks to each other today. Shael's longer-term goal is not to build a better dashboard but to connect those sources into one operational picture that moves a finance team from reporting the past to influencing the future.
4. Networking is not a career move. It is a life move.
The shift for Shael was realizing that building a professional network was not about advancing a title. It was about having people in his corner who actually understand the work. The CPA Ontario Innovation Leadership Accelerator opened that door, and the habit of showing up to unfamiliar rooms, a little uncomfortable, has compounded steadily since.
Charlie Liu
Hello everyone, welcome back. Today we have Shael Kerbel. Shael is a CPA, CA, senior manager of FP&A at Oliver and Bonaccini Hospitality, one of Canada's leading fine dining restaurant and catering groups, where he is putting AI to work on financial functions. He is building automated tools that turn raw POS data, GL details, and planning models into decisions the business can actually act on. Before landing in hospitality, he spent roughly a decade across FP&A and corporate development at PwC, Intact Financial, and Echelon Wealth Partners. Working on deals and diligence taught him that finance is less about the data itself and more about the story it is trying to tell. I am really excited for this one. Shael, you want to quickly introduce yourself?
Shael
Yeah, for sure. Charlie, thanks for having me. This is something different and new for me and I am really excited to discuss what we have been talking about for so long. Of course there is the hospitality and the AI angle, but really it is that personal growth side and just how to be a better person and be more comfortable in front of others. That has been a journey I have been pursuing and I know you have too.
Charlie Liu
We always like to start with some rapid fire questions and then we will come back to growth on the personal side. First question: Prosecco or champagne?
Shael
Can I say beer? Dark in the winter, sours in the summer. Not really a Prosecco or champagne kind of guy.
Charlie Liu
I am personally a fan of all stouts. I was in Ireland last year, Guinness straight from the source. Second one: brownie or cheesecake?
Shael
It has got to be cheesecake.
Charlie Liu
Michelin star or Google five star?
Shael
Depends on who is paying.
Charlie Liu
Straight from the finance perspective. Beach or mountain?
Shael
Definitely mountains.
Charlie Liu
Audiobook, podcast, or neither?
Shael
Depends where I am. I listen to audiobooks to fall asleep. On the TTC I am usually listening to a podcast. I like fiction for sleeping, mostly sci-fi. Not weird fantasy, more like Project Hail Mary. I loved that book before it was ever conceived as a movie.
Charlie Liu
What is your favorite AI tool right now?
Shael
Right now it has got to be Claude, Opus 4.7. But I use all of them. I default to Claude most of the time in our line of work because I find it is the most capable, but I will reach for whichever one fits the task. I also use Perplexity because it has access to different models and it is cheap. And the Comet browser from Perplexity has been a really interesting use case. I had it build a full Google Maps itinerary for my trip to Calgary and my upcoming trip to the UK, not just points on a map but actual routing day by day.
Charlie Liu
You mentioned you are starting to test AI at work. You have been building dashboards and some kind of integration. How has that been?
Shael
We have some Power BI dashboards. But what I am really working on is combining our finance data from Vena with our POS data, our Dayforce HR data, and eventually OpenTable and Google data, into one mega operational tool. The biggest challenge is not conceptualizing what I want it to be. It is turning that into a reality.
Charlie Liu
What kind of early success or learnings have you had?
Shael
I have what I would call an alpha of a tool that helps build the sales pipeline for our events and catering business. Oliver and Bonaccini does a huge amount of catering across Toronto. With FIFA coming to Toronto, for example, there are a lot of private dining and corporate dining opportunities we could go after. The tool refreshes upcoming events, searches LinkedIn and websites for key contacts, and gives you angles on how to approach them depending on whether you want the main venue contract or just a private dining room for executives. It is a one-click refresh, basically.
Charlie Liu
Talk to me a bit more about Oliver and Bonaccini for the audience.
Shael
So Oliver and Bonaccini, known as O&B, has a number of restaurants across Toronto under different names, different concepts. French, steak, British, Latin. Then we also have the Concord Entertainment Group in Calgary with more concepts out there. Over forty restaurants between the two. On top of that there is a significant events and catering business. We own venues, we have strategic partnerships with other venues where we are the dedicated caterer, and within every restaurant there is private dining space that operates almost as its own revenue stream. So restaurants are a big part of the business, but certainly not all of it.
Charlie Liu
What is the biggest hurdle to actually operationalizing the tool you are building?
Shael
There are two hurdles. The technical hurdle is getting it from working for me to working consistently for other people. That will be the next step. But the adoption hurdle, for this tool at least, I do not think will be that bad because I am building it to be simple enough that anyone can use it. The real challenge is making it production ready. That is the hardest part. Building the features is not the issue anymore.
Charlie Liu
How do you think you will get to a place where the infrastructure you are building can consistently give you trustworthy, accurate output?
Shael
There are two kinds of data. Numbers, mostly financial but also operational, and then what Charlie called blobs, the text-based stuff. For the numbers side, you never want the LLM doing the calculations. You want it using Python or some other codified tool that is repeatable and consistent. That part needs to be hard code. For the text-based side, prompting and training your AI as well as you can is important. It is still not perfect, but it helps a lot. And the time you save already makes up for the occasional miss. There still needs to be a human in the loop.
Charlie Liu
That bonded us, I think, was also the focus on growing and facing uncomfortable parts of life. What triggered that for you?
Shael
I have always been an introvert. I do not love starting new conversations and small talk can be awkward. The more you do it the easier it gets, like anything. But really it was the CPA Ontario Innovation Leadership Accelerator program. Through that I heard about an event, and that is indirectly how I ended up meeting Charlie. Even joining that program, I was on the fence. I was long done with school. But the biggest thing it showed me was how important it is to have a network of people who actually understand the work you are doing. My family is wonderful but they have no idea what I am talking about when I describe a professional problem. Building that network of people at a similar level with similar interests, that was a huge eye opener. That got the ball rolling.
Charlie Liu
You have done quite well at your career stage. Was there a specific trigger where you said I need to change?
Shael
It was more that I just realized I have been who I am for so long, and not that I am unhappy with who I am, but I wanted more. It was not about advancing a career title. It was about personal growth. I joined the ILA program out of curiosity, not with a big plan, and it sparked something. That is actually the advice too. You do not have to have a grand intention. Just find people that have similar interests. It does not have to be a formal program. Join a running group. Show up somewhere. Be friendly, be hospitable, and you will see doors open that you never knew were there.
Charlie Liu
Shael, this was a lot of fun. Before we close, what is your two-cent advice for the audience?
Shael
It is really two different cents. The first is be vulnerable. Talk to people. Open doors for yourself even when it is uncomfortable. The second is train your AI model. Take the time to learn about prompting. They all have memory now. Put your instructions in, tell it the tone you want, tell it what you need. You will get way better results. It does not have to be perfect to start. You can even use AI to help you write better prompts.
Charlie Liu
I love it. On the personal side: be vulnerable. On the AI side: take advantage of memory, train it, build your instructions. Thank you so much Shael. Pleasure having you.
Shael
Yeah, pleasure. Thanks, Charlie.