How UrbanStems turned AI friction into faster, smarter customer support
How cleaning up the “brain” behind the AI helped UrbanStems increase one-touch resolutions by 52% while protecting SLA performance during peak floral season.
UrbanStems is a high-growth floral and gifting brand operating in one of the most operationally intense corners of ecommerce. Seasonal spikes (especially around Valentine’s Day and Mother’s Day) create huge support surges, where even small inefficiencies can quickly snowball into queue backlogs, slower response times, and frustrated customers.
The team already had AI in place through Gorgias, but the experience wasn’t delivering the level of efficiency they expected. Too many conversations were still escalating to human team members, especially low-complexity “Where Is My Order” (WISMO) inquiries that should have been fully self-serviceable.
Boldr partnered with UrbanStems to audit the existing AI setup and develop a roadmap for optimization across knowledge base structure, AI guidance logic, and handoff flows.
UrbanStems implemented the recommendations internally, helping create a smarter, more accurate support experience that reduced friction for both customers and team members.
The challenge: when AI exists, but still creates work
UrbanStems wasn’t starting from zero. The AI agent was already active inside Gorgias, and customers were engaging with it regularly. The issue was that too many of those conversations were still ending in handovers to human team members.
At the time of the audit, 68% of AI interactions escalated to the support team. More than half of those handovers were tied to WISMO (“Where Is My Order”) inquiries: simple, repetitive questions that should have been resolved automatically.
In the world of time-sensitive deliveries, it was imperative to the UrbanStems team to get these questions answered quickly and accurately. The excess of WISMO queries created a frustrating cycle; customers weren’t getting answers quickly enough, while team members were spending valuable time on low-complexity tickets instead of focusing on more sensitive or revenue-impacting conversations.
A major contributor was the structure of the knowledge base itself. Over time, the help center had accumulated what the team described as “knowledge base debt.” Articles were too broad, titles contained multiple questions, and information was bundled together in ways that worked for humans skimming content but confused the AI trying to retrieve precise answers.
In practice, the AI had access to information, but struggled to surface the right information consistently. The stakes were especially high given the nature of the business. With each support interaction tied closely to customer retention and repeat purchasing behavior, delays, inaccurate responses, or AI “hallucinations” carried meaningful business risk, especially during holiday surges.
As Laura McDonald, Director of Customer Happiness, explained:
“The biggest headache, both during the holiday push and more generally, was volume, specifically one-touch tickets and easy wins eating up team member time that should have gone to more complex or sensitive cases. Deflection has been the most meaningful gain from our changes; it's taken pressure off the queue and let team members focus where they have the most impact.”
The solution: optimize the AI’s brain, not just the AI itself
Rather than recommending an entirely new platform or automation stack, we focused on improving the systems already in place. The core philosophy was simple: AI performance depends heavily on the quality and structure of the information feeding it.
Boldr’s Technical Services and Solutions team began with a technical audit of UrbanStems’ Gorgias AI setup, knowledge base architecture, and handoff logic. From there, Boldr developed a roadmap of recommendations focused on improving accuracy, clarity, and scalability.
One of the biggest recommendations centered around cleaning up the knowledge taxonomy. Instead of long, bundled articles covering multiple scenarios at once, Boldr recommended a “single-topic” structure with simplified titles and clearer categorization to make content easier for the AI to index and retrieve accurately.
Boldr also identified opportunities to refine the AI guidance layer itself. Rather than allowing the bot to “guess” its way through conversations, the recommendations focused on creating more intentional guidance structures, including collecting critical information like email addresses and order numbers before escalation, while redirecting simpler inquiries toward more specific FAQ flows.
WISMO flows received particular attention. The audit uncovered several points where customers were dropping out of automation and forcing handovers simply because instructions lacked clarity or context. By tightening those flows and removing ambiguity, the AI became significantly more effective at resolving straightforward delivery questions independently.
The timing of the work also mattered.
Because UrbanStems operates in an intensely seasonal industry, the recommendations included a strategy for temporary and seasonal FAQ content tailored to major gifting holidays. This helped prepare the AI for high-volume questions unique to Valentine’s Day and Mother’s Day, including time-sensitive delivery concerns and gifting expectations.
Importantly, the goal wasn’t to remove humans from the support experience; it was to make sure team members spent their time where they added the most value.
As Laura put it:
“The tickets that do come in now arrive with more complete information, so team members can resolve them faster instead of going back and forth to gather context.”
Impact
After implementing the optimization recommendations, UrbanStems saw measurable improvements across both AI performance and operational efficiency.
Key results
- 52% increase in one-touch resolutions
- 6% decrease in messages per ticket
- 6% decrease in median first response time
- Maintained sub-4-hour email SLA during peak seasonal volume
- Reduced unnecessary handovers for low-complexity WISMO inquiries
- Improved ticket quality and context collection before human escalation
The operational impact extended beyond metrics alone. Because the AI was handling simpler conversations more effectively, the human support team could focus on emotionally sensitive, time-critical, and revenue-impacting customer interactions instead of repetitive queue-clearing work.
That balance became especially important during holiday periods, where accurate and timely delivery is of the utmost importance.
Looking ahead: building smarter support systems, not bigger queues
UrbanStems continues to evolve its AI and automation strategy with a strong focus on scalability, operational efficiency, and customer experience quality.
The partnership reinforced an important lesson: effective AI support isn’t just about deploying automation. It’s about designing the systems, knowledge structures, and workflows behind it thoughtfully enough for the technology to succeed.
By treating the knowledge base as the engine powering the AI experience, UrbanStems and Boldr were able to increase support capacity without scaling headcount at the same rate as ticket growth. In high-volume ecommerce environments, that kind of operational leverage matters.
Ready to make your AI actually useful?
AI tools alone don’t solve support problems. The real impact comes from how well the workflows, knowledge systems, and customer journeys behind them are designed.
Boldr helps high-growth brands optimize support operations, improve automation performance, and create customer experiences that grow without sacrificing quality.
Want to build smarter support together? Let's talk.