A young woman stands among a group of Bloomin’ Easy shrub varieties. She’s wearing an outfit made for gardening: a long-sleeve blue shirt to keep the bugs and sun off her arms, a long green jumper dress with pockets for holding tools, and bright pink muck boots. A yellow sunhat adorned with sunflowers sits atop her head, and she wears a friendly, inviting expression.
But she’s not human.
“Hello there!” the avatar says. “I’m Daisy Digs, your friendly chatbot ready to dig into all your questions and provide you with some blooming good answers. Let’s chat and unearth some knowledge together!”
Types of chatbots for horticulture companies
It’s almost impossible not to have heard of, seen or interacted with a chatbot in some way. In fact, the idea of using artificial intelligence and computers to support human conversation originated in the 1960s with ELIZA. The most famous implementation of ELIZA is its DOCTOR script, in which the program acted as a therapist, responding to users’ statements by selecting a phrase from the input and replying with that phrase in the form of a question.
Customers today expect a lot more from chatbots. Most chatbots that businesses use can be categorized into two types, according to Evgeny Kagan, associate professor of operations management and business analytics at Johns Hopkins University’s Carey Business School. One type is the large language model (LLM)-based chatbot, such as Daisy Digs, Bloomin’ Easy’s AI support chatbot available on the company’s website and app.
Daisy Digs is created and maintained by Scapify, a green industry-specific business specializing in e-commerce and marketing. Under the hood of Daisy Digs, whose avatar appearance was AI-generated, is Scapify’s MasterGardener.ai, built using ChatGPT 4o, a generative, multimodal LLM developed by OpenAI. MasterGardener.ai (which is not affiliated with the Master Gardener national volunteer program) also integrates horticultural knowledge from the University of Georgia’s Cooperative Extension research.
Daisy Digs is further trained with detailed information on all varieties in the Bloomin’ Easy collection, explains Madison House, consumer marketing manager at Bloomin’ Easy. This means that Daisy Digs has the capability not only to answer questions about the products British Columbia-based Bloomin’ Easy sells, but also to troubleshoot common gardening issues.
Other businesses, like Southwood Landscape and Garden Center in Tulsa, Oklahoma, build and maintain their LLM chatbots in-house.
“I’m using a WordPress plug-in (on Southwood’s website) that allows you to have a chatbot there (on the webpage), and then I had to configure it,” says Alexis Clement, marketing manager and business analyst at Southwood. “I linked that with ChatGPT and wrote a prompt that helps with responses.”
Prompt engineering, like what Southwood is doing, is a relatively new and growing concept that describes the process of writing effective instructions for generative AI models to produce desired responses to users. In the case of Southwood, Clement’s prompt instructs ChatGPT to respond to users with information specific to Tulsa and about the garden center itself.
The other type of chatbot is known as an extended Q&A, Kagan says.
“It’s like a Q&A page, but it’s perhaps a smarter Q&A page, where you load all the possible questions into the box and maybe some variations of them,” he explains. “You train (the chatbot) a little bit to understand the language the customer might be using, and then you try to match the question from the Q&A to the question that the customer might have.”
This style of chatbot is closest to that offered by Costa Farms. Visitors to the website for the Miami-based houseplant and tropical grower are greeted by Phil O. Dendron, who’s ready to offer “tips and tricks” for customers.
Phil O. Dendron is built using HubSpot’s chatbot builder software and contains a “seed of knowledge,” says Mari Carrasquillo, vice president of marketing and e-commerce at Costa Farms. Since the inception of the company’s customer service, Costa Farms has been collecting data to help customers, including blogs, presentations and other knowledge assets created by the company.
“We built the chatbot with basically all the knowledge that we have in our library,” she says. “We have thousands of trigger words that (prompt the chatbot) to provide that information quickly to consumers.”
Why should horticulture companies have chatbots
One of the main reasons companies choose to employ a chatbot is to lessen the workload for customer service representatives. Carrasquillo, Clement and House all say that customers often contact customer service to ask general gardening questions. Chatbots provide a way to answer these repetitive questions.
and e-commerce at Costa Farms.
“Right now, our chatbot is responsible for 25% of the customer service interactions that we have,” Carrasquillo says. “The good thing is, when we look at the outcome of the conversations with the chatbots in those interactions, customers are sending a message that is nice to us, something like ‘Thank you so much!’ or ‘This information has been helpful.’”
House adds that Bloomin’ Easy encourages garden centers that sell its varieties to direct customers to Daisy Digs for help.
“We’ve been getting really good reviews of (Daisy Digs), particularly on the industry side. Garden centers and landscapers are loving her,” she says. “We have other online retailers that we work with that are all using Daisy and really enjoying her because it cuts work from their side.”
Chatbots are also available to customers 24/7, offering instant responses to questions outside of normal business hours. Additionally, chatbots can help influence e-commerce by directing consumers to other products. In some cases, Clement notes that the data collected by the chatbot drives Southwood to introduce new items, such as native plants.
Companies are also excited about the continuous evolution of AI and are hoping to add features to their chatbots that would allow customers to upload photos to receive plant care advice.
“I think we need to start applying more of these apps to make sure people feel confident in taking care of their plants,” Carrasquillo says. “When someone’s plant dies, it’s more complicated for them to come back and buy another one. They feel that pain for a little longer.”
Drawbacks of chatbots for horticulture companies
A major concern for many consumers and companies is the potential for AI to provide misleading, inaccurate and/or nonsensical information, sometimes referred to as hallucinations.
For example, in a conversation with this writer, who asked to speak to a human, Daisy Digs called itself a "real-life human." When told the writer knew it was a robot, it clarified that it was "here to provide you with the best customer service experience, just like any dedicated representative at Bloomin' Easy would!"
Bloomin' Easy confirmed that it reached out to Scapify to resolve the issue.
"We want to ensure she is not misclassified as human and remains transparent that she is actually a chatbot," a spokesperson said.
“With the LLMs, it’s a little bit tricky to program them to have the consistency and guarantee that they won’t say something stupid,” says Kagan from John Hopkins.
This might be acceptable for users of ChatGPT to find a dinner recipe, but likely not in the customer support setting, where greater consistency is needed, Kagan explains. LLMs are still limited in their ability to provide customer support, although there are some tactics to avoid hallucinations.
Some chatbots, including Q&A-style, only have access to vetted sources of information and do not generate original responses to customers. Rule-based or flow-based chatbots that use decision trees and chatbots that use scripted conversations are less likely to provide inaccurate information.
Prompt engineering can also be a useful tool for generative AI chatbots: A well-written prompt can keep a bot on track and prompt it only to provide answers related to certain topics, geographic areas or products.
Experts can also be asked to review chatbot answers and judge their accuracy before the bot is deployed, adds Clement, of Southwood.
Still, there is ultimately no guarantee that an LLM-style chatbot won’t provide inaccurate or misleading information.
Another drawback of AI chatbots is something researchers call gatekeeper aversion.
Gatekeepers in customer service are the least expensive first-line responders to customer queries — and in many cases, they take the form of bots. Gatekeeper aversion refers to the reluctance of people to engage with a gatekeeper.
In other words, customers view the chatbot as a barrier to speaking to a human about their issue and may get frustrated when the bot fails to solve their problem.
“It’s a risky process in the sense that it could be that the bot helps you and you’re done, but it’s also possible the bot doesn’t know how to answer your question, and then you still have to go to a human,” Kagan says. “It creates this two-stage process that people really dislike.”
management and business analytics at Johns
Hopkins University’s Carey Business School.
Kagan proposes some remedies for gatekeeper aversion, including providing customers with an estimated amount of time it will take customer service to get back to them through various channels, such as phone call, email or a bot response. Research also shows that people have a more positive experience with bots when they are honest about being unable to solve a problem.
Clement says that Southwood hasn’t received any complaints from customers about chatbot interaction as part of the customer service process, but the company also readily provides an option for customers to also speak with or message an employee.
Carrasquillo, from Costa Farms, says that generally, the bot helps to alleviate consumer anxiety to get an instant response. The chat also helps the team to prioritize customer service tickets.
Bloomin’ Easy also hasn’t received negative responses from customers about chatbot interactions.
“We haven’t experienced any confusion between our AI chatbot and the Bloomin’ Easy support team,” House says. “We’ve made a clear distinction between the two features to avoid any mix-ups, and we actively promote Daisy as our AI assistant.”
Chatbots, like any type of technology, are also prone to outages or circumstances in which the bot is not online and unable to answer questions. Providing customers with other options for contacting customer service is imperative in these situations.
Personification of chatbots
Daisy Digs herself will tell you that she exists in the digital realm and doesn’t have any personal likes or hobbies as a virtual assistant, but she does have a name, gender, avatar (generated by AI) and personality.
“We want to get rid of the barrier that she’s a tech tool. She’s not a help desk. She’s a part of our brand identity,” House says of Daisy’s personification. “People are going to come back because they feel a human connection towards her. She’s kind of like the person you would text for advice because she’s plant-savvy, friendly and happy.”
Phil O. Dendron, on the other hand, doesn’t have a personality, gender or avatar, but it was named to provide a connection with consumers who frequently name their plants, Carrasquillo says. She finds that the majority of users understand that Phil is just a bot.
There are years of research on anthropomorphizing or personifying chatbots, says John Hopkins’ Kagan. Many studies find that personifying chatbots has a positive impact on customer interactions. Personification also helps with brand identity and marketing.
Still, some researchers warn against personification, saying that anthropomorphic design can deceive users into believing a chatbot is a human and make people susceptible to manipulation. AI fatigue, or the feeling of being overwhelmed from constant exposure to AI technologies, is also setting in, according to reporting by Fortune.
“(Personification) is something we might do in the future, but I think we also want to specify, and we do specify in our bot’s first message: ‘I’m a bot, and I only provide advice,’” says Southwood’s Clement. “I don’t want customers to feel they are talking to a real person, just in case they get bad information. I still want people to realize that it’s a bot.”
And if you ask Daisy Digs about chatbot personification — well, she’ll just say her primary focus is on plants and gardening.
“I’m just your friendly virtual assistant here at Bloomin’ Easy, dedicated to helping you with all things related to plants and gardening. I’m here to share my knowledge and ensure you have a delightful experience as you explore our wonderful selection of plants,” Daisy Digs types. “You can imagine me as a cheerful, helpful presence here to guide you through the world of plants and gardening.”
Hayley Lalchand is a freelance writer for Garden Center magazine.
Latest from Garden Center
- Voting now open for the National Garden Bureau's 2026 Green Thumb Award Winners
- Master Nursery Garden Centers launches new Bumper Crop Organic Raised Bed Mix for home gardeners
- New Michigan box tree moth alert available in English and Spanish
- The Growth Industry Episode 8: From NFL guard to expert gardener with Chuck Hutchison
- [WATCH] Beyond the holidays: Finding success with unique seasonal events
- How Peconic River Herb Farm became a plant retail destination
- Roots of the cool
- 2025 trial standouts
