AI for good (staff training)
Restaurants get lots of feedback. Acting on it is harder. “There’s this question of, ‘Okay… now what?'” — Rachael Nemeth, co-founder and CEO, Opus
Artificial intelligence in the restaurant business has long promised better data analysis. Think: A solution to, “We have all this customer info, but we don’t know what to do with it.”
That promise is finally being realized as restaurant tech companies introduce tech forward and useful updates. Like this one!
Opus, a training operations platform for restaurants, recently added an AI-driven product that turns guest reactions into immediate action. It links up with customer feedback platform Ovation (just Ovation for now) to analyze feedback and turn it immediately into targeted, location-specific training recommendations and courses. In early testing, the new, AI-driven product is, frankly, killing it, reducing reoccurring customer complaints within the first month.
In an interview, Rachael Nemeth, Opus co-founder and CEO, told me, “No one has to interpret anything or decide what to do, it just happens.”
Staff training is one of those deep behind-the-scenes restaurant operational pieces that’s maybe a little boring or in-the-weeds to the average restaurant-goer. But its results (or lack thereof) show up readily in service, including pickup and delivery orders.
Opus helps multi-unit chains train and retain staff while standardizing service across multiple locations. It creates mobile-first digital training modules designed to be accessible and understandable to desk-less, frontline service workers. It works with hundreds of brands, including white-hot Craveworthy Brands, which tested this product, and the José Andrés Group.
I spoke to Nemeth, a Union Square Hospitality veteran who is always thoughtful about adding new tech to hospitality training, about the new feature and her thoughts on AI in restaurant software, plus some early surprises from Opus’s newest feature. That conversation continues below the paywall. (Thanks, as always, to Expedite’s paid subscribers who keep this newsletter afloat.)
It’s a practical example of AI that helps, not replaces, an hourly workforce.
“The goal is that we turn feedback into action,” Nemeth said, “instead of just more data that everyone's ignoring.”
Our conversation has been lightly edited for length and clarity.
Expedite: This feels like such a smart use of AI technology. As in, it’s actually useful.
Rachael Nemeth, CEO, Opus: “Restaurants get all this customer feedback through great digital tools. But then there's this question of, okay, now what? Someone has to read through it; they have to figure out what matters, then create some sort of training for it. By the time that happens, it’s like a week later and the problem is still there. We thought, ‘What if the system does it automatically?’
“I was speaking with an operator the other day and I asked what they expected of store managers. He told me they expect managers to respond to every single negative guest review — that’s 50,000 per week for a restaurant group with over 100 locations.”
That seems… untenable. How does the system know when it’s time to implement a new training?
RN: “We pull data from the guest feedback platform — Ovation in this case — and using AI, the system automatically interprets the data against the store location, the role, the franchise, and from there it will deliver training recommendations in seconds. Depending on what trends are forming, it could mean that there’s a new recommendation in the dashboard every week. From there, the person running Opus —usually a training or ops director — looks at the recommendation and chooses to accept it or not; they have control over what sort of corrective action they want to take. This could happen in literally minutes, and the action could be reassigning an existing training to reinforce something. Or it could identify a knowledge gap, then generate a new training with AI and deliver it. Or it could be something as simple as sending managers a message about what’s happening.”
And this is customized down to the store level?
RN: “Yes, it’s very granular.”
I know it’s early, but has any group discovered immediate trends?
RN: “Three words: Sauce. In. Bag. We’ve all ordered food for delivery and the restaurant forgets the dressing or sauce. But this is a common challenge for restaurants, so we tested against this with one of our restaurant partners. Early numbers are really encouraging — we were able to reduce that repeat customer complaint by 60 percent in 30 days.”
That’s significant.
RN: “These issues might seem trivial from a training perspective, because it’s not hard for an hourly employee to figure out that you need to put sauce in the bag, but it’s about reinforcement and ongoing training — that’s the key insight for us here. Restaurants think about new-hire training and onboarding, but ongoing reinforcement is completely ignored — not for any other reason than these restaurants just don’t have time.”
Has anyone discovered anything less expected than not forgetting the dressing?
RN: “I wouldn’t necessarily say less expected, but here’s a broader anecdote: With this new feature, we’re not just measuring engagement. We’re measuring fewer angry customers. I think every large multi-unit group has franchisees that it only has so much exposure to. And now those franchisees have access to this data and be able to take action. What we didn’t expect is how much this has surfaced franchisee-by-franchisee issues. The AI isn’t just pulling company-wide data, it’s signaling against the region or the sub-brand or the franchisee.”
I appreciate you’re so thoughtful about AI and its uses. We’ve talked about this before in this newsletter— AI is so buzzy, but how have you been using it at Opus?
RN: “We take a pretty different angle on AI from other training platforms, I think. We’re moving toward training that is relevant, to help solve issues that are actually happening within the business — training that actually knows what’s broken.
“The reality is that most frontline workers get maybe an hour of training per month, if they’re lucky, and you can’t waste that on generic topics that don’t fit into their day… but that’s the status quo right now. For too long, feedback tools and training platforms have been completely separate, so it makes sense to have data that shows what’s going wrong and automatically informs what people need to learn and do next. How we think about AI is, let’s take the information that’s already out there and help you process it and do something with it. It shouldn’t be guesswork, especially in the context of guest feedback. Your customers are actually telling you what’s wrong. You should actually do something with it.”
It sounds so obvious when you say it like that.
RN: “So many of the multi-unit brands we work with have the same issue, which is that they’re training employees but there are no results. Frankly, that doesn’t get solved by a standard training platform. We think about ourselves as a training operating system, though we’re not officially this yet. How can we get beyond the content management system and into this world where training — which is already this embedded activity in your business — becomes the central nervous system of your hourly workforce?”
The narrative around AI is so often, ‘replace workers, reduce workers’ and this is not that.
RN: “There is a component of replacing… it’s replacing the things you can’t afford to do. It’s replacing the huge gap of that restaurant group that’s saying, of course our managers aren’t responding to 50,000 pieces of feedback a week. But how can that manager use AI to respond and then collect the insight and deploy the training.”
Understanding this is still new, what’s the success metric? Better customer feedback?
RN: “It’s about reducing repeated issues. If you have lower repeated instances of customer complaints, that means you are getting more revenue per hourly employee. It’s really as simple as that.”
